AMD Going Dual-Core In 2005

gr8_phk writes "We recently learned of Intel's plans to go dual-core in late 2005. Well it seems AMD has decided to follow suit. It should be noted that the K8 architecture has had this designed in from the start. Will this be socket 939 or should I try to hold out another year to buy?"

Comments

Why not quad core?

[leonbrooks]

If more is better, why not proliferate cores like crazy?

Re:Why not quad core?

[Haydn Fenton]

2's company, 3's a crowd, and 4 is for the fat cats who wipe their ass with 50 dollar bills.

Re:Why not quad core?

[wmeyer]

Interestingly, in a review of P4 vs. K8, the K8 had a clear advantage at the 4 processor level and above, apparently because of reduced bus conflicts with their individual memory spaces. If AMD were to proliferate cores on chip, they'd wind up contesting for the memory bandwidth, just like the P4.

Re:Why not quad core?

[mp3LM]

heat

Yes..the evil of all machines
the reason why when the AC is not on in my house, and it is 90degrees outside, my computer resets
and of course..the reason why we're not going quad core


well..at least that's my personal opinion...as for the real reason...probally for profit...

Re:Why not quad core?

[k4_pacific]

This could ultimately lead to a reformulation of Moore's Law. Thus, I propose k4_pacific's hypothesis:

The number of processor cores doubles every eighteen months.

Re:Why not quad core?

[afidel]

That's funny Moore's Law says that the number of transistors per area will tend to double every (18 or 24 months depending on which part of Gordon Moore's career you listen to him). More cores per chip with better processes does nothing to stop this progression.

Re:Why not quad core?

[isorox]

6 is the right number, it divides up so easilly. 6*1, 3*2, 2*3, or any combination.

3 is just ugly.

Re:Why not quad core?

[AuMatar]

You mean 8. This is a computer, you're legally bound to use a power of 2.

Re:Why not quad core?

[ruiner5000]

actually there is plenty of bandwidth left in hypertransport to pull it off. also each cpu gets its own bank of memory. the design is superior to all others for SMP. even AMD's man CPU man says that at infoworld

AMD's dual-core server processors will share a single memory controller, Weber said. This won't create a bottleneck because a server with two Opteron chips, and therefore two memory controllers, already has more than enough memory bandwidth required to run that system, he said.

"It's always a juggling act to add a little more processing and a little more memory. Right now, we have plenty of memory and I/O bandwidth, so we're adding processing," Weber said.

The dual-core chips will work with current socket technology in motherboards that are rated for the specifications of the dual-core chips, Weber said. A BIOS change will be required, but otherwise the chips will work in the same sockets as single-core Opterons, he said.

Re:Why not quad core?

[stevesliva]

Right now, we have plenty of memory and I/O bandwidth, so we're adding processing,

Wow, that's a helluva significant statement to make. More bandwidth and cache never hurts... you just get diminishing returns. Either it's marketspeak-- "You need two cores, pay up!" --or it flies in the face of some pretty basic assumptions I have about processor and cache architectures. Perhaps he meant, "Our cache and bandwidth is currently large enough to support two processes without being detrimental, given two processes that play nicely?" That could make sense...

Re:Why not quad core?

[ruiner5000]

Are you not aware of how well AMD's K8 core line of CPUs scales. There are hundreds of articles on the web mentioning it.

Re:Why not quad core?

[stevesliva]

Regardless of whether they're accurate or not, the statement stands as being dumbfounding. How many times is "memory is the bottleneck" pounded into the heads of Computer Architecture students? AMD just said memory is not the bottleneck. You don't need a faster bus, or a faster cache, you need more parallelism. Usually that sweeping of a statement is qualified with a few caveats.

Re:Why not quad core?

[hxnwix]

The opteron (k8) has an integrated memory controller and up to three hypertransport links. In a dual k8 system, the cpus communicate over a single hypertransport link and are usually paired with their own memory bank. If one cpu needs data from the other's bank, it comes over the hypertransport link. Some cheap dual opteron boards save traces by pairing one cpu with all the memory banks - so every memory operation on the non directly linked cpu passes over the h-link.

The dual core cpu might have the pins for two seperate memory bank arrays or just the pins for one. Either way, the situation as far as dual k8s go is not really different from what we have already. Either way, it's a few steps above the p4 design: shared cpu bus to northbridge to memory. (yech! with a single proc, this introduces latency, with multiproc, you get contention and latency at every level)

AMD's cpu interconnect is so well thought out... it gives me the warm fuzzies pondering it:

A uniproc hammer needs one h-link for io.
A dually needs two per core: 1 for core to core, 1 for io (though all the io on all the boards I have seen feeds to only one proc's h-link... so that you don't lose PCI busses and such if you have only one proc installed, I suppose).
Quad and above requires three: each core links to two other cores, leaving one h-link per core for io. One could have a pci-e bus per proc, if one desired. But again, I haven't seen a design that doesn't feed all io into a single h-link.

Since no one uses the extra h-link anyway, a dual core package for a dual core system would need only one external h-link (saving some cash).

A quad core, dual package system would require three h-links feeding out of each package, though. But even then, the number of h-links laid out on the mobo is reduced and the whole shebang should be cheaper.

Intel's "one huge shared bus" + northbridge design is definitely being trampled...

Re:Why not quad core?

[MarvinIsANerd]

4 is a power of 2. 2^2 = 4.

Re:Why not quad core?

[HuguesT]

The answer is that it doesn't make sense for a desktop machine.

Windows professional comes with a license for 1-2 CPUs. Above that you need to purchase one of the server edition, and it starts becoming *very* expensive.

Soon 2 CPUs will be for the masses, they already are with hyperthreading in a way. However 4 and above really are for servers; multi-user environments, etc.

Also while it is easy to exploit 2 CPUs in a desktop environment (roughly speaking 1 for the O/S, the other for the applications) there are diminishing returns for more than 2 CPUs, except for parallel applications, and there aren't that many of those.

Re:Why not quad core?

[NerveGas]

Because the overall size of the die is a tremendous factor in the cost of a processer. Because of that, die sizes tend to stay relatively constant over the years.

As manufacturers are able to squeeze the transistors in more tightly, then you see more circuitry appearing. As they move to 90-nanometer production, they're going to be able to pack on more transistors, and using dual cores becomes an economic possibility. However, throwing FOUR cores on would make the die large enough to be an economic disaster. (Die size was one of the largest problems with the Pentium Pro.)

steve

Re:Why not quad core?

[lsdino]

While typically computers do use powers of 2, you can run SMP machines w/o powers of two. For example I've heard of 3-proc machines. I believe it's just a quad proc w/o the 4th processor. Odd, but true.

Re:Why not quad core?

[Anonymous Coward]

4 is not 6. Perhaps you missed the great-grandparent post?

Re:Why not quad core?

[RyuuzakiTetsuya]

Look at BeOS.

The threading was planned from day one to support multiple processors with out any special coding. It's been a few years, but I think i'm right.

If Microsoft is smart, they'll implement something like this for Longhorn and whatever binary executables are used.

Re:Why not quad core?

[Junior J. Junior III]

The upshot of this is that it makes Linux all that more attractive since there's no per processor licensing bullshit to deal with, and little if any distinction between a server and workstation install (mainly it's just a matter of what packages you choose include with the install).

It's pretty easy to predict that what will happen is Microsoft will be forced to amend or drop its per processor licensing scheme, or else deeply discount it, or else face even more widespread piracy. Then again, they might have enough clout to hold up AMD and Intel's release of the multi-core CPUs until such time as they make Windows ready for it.

Re:Why not quad core?

[Paul Jakma]

apparently because of reduced bus conflicts with their individual memory spaces.

Ah but with multi-core chips they can transduce their flux capacitors with the onboard trans-mogrification controllers. Seriously "reduced bus conflicts with their memory space", what does that mean?? That's gibberish.

P4, presumably, like the P6 GTL+ host bus is a shared bus (like most buses are). Only CPU can use the bus at any one time. If the bus does x GB/s, that's only to one CPU at any given time - effectively it is shared. Further, P6 and P4 do not have integrated memory controllers, and must access RAM via the (shared) GTL+ bus, if it is not in cache. Eg, a 4 CPU machine looks like:

P = CPU
MC = Memory Controller (part of the "northbridge" chip, also provides PCI host bus controller, etc.)

P P P P
| | | |
--------- GTL+ bus
|
MC--RAM

Also GTL+ is limited to 4 CPUs and one controller. To get 8 CPUs some controller vendors have invented a GTL+ 'bridge' to stitch 2 GTL+ buses together, but that just makes things worse really from a scaleability POV I'd imagine.

The K8 on the other hand uses a point-to-point (PtP) serialish, packet based transport, HyperTransport to interconnect CPUs and has onboard memory controller(s) (connected internally via HyperTransport links). A 4 CPU K8 machine looks like:

K = K8 CPU
HT = HyperTransport link

RAM--MC-\ /-MC--RAM
RAM--MC--K--K--MC--RAM
| |
| |
RAM--MC--K--K--MC--RAM
RAM--MC-/ \--MC--RAM

Each of the lines out of a K is a HyperTransport link. Each MC is integrated into the die itself. (you'll have to imagine interconnects and right-hand top/bottom MC's lining up with the K symbols, cause /.'s filter is chomping whitespace in some strange way on me).

Each CPU has 4 HT links, two to other CPUs, two to its (integrated on die) memory controller. For dual CPU setups, each CPU needs only link to another CPU obviously. Indeed the difference between 2xx, 4xx and 8xx AMD Opteron CPUs is the number of HyperTransport links. Indeed in large multi-CPU (ie 8+) SMP setups one need not attach a memory controller to each CPU, one might choose to have a central "cross-bar" of fully-meshed K8s who then connect to peripheral K8s which have memory controllers and hence RAM. Tis all down to the board designers I guess. And a bit of a fun computer science problem too in terms of designing optimal 'networks' of interconnected nodes with the best compromise of maximum node to node distance for lowest number of required interconnects.

The K8 is actually a ccNUMA (cache coherent, Non-Uniform Memory Architecture) machine, in SMP configurations. Ie, different memory is at different distances to different CPUs, or to put it another way, some memory is local, other memory is distant, some memory may be more distant than other memory. Eg, for the top-left CPU to access RAM on it's "local" MC is obviously potentially far quicker, in terms of latency, than to access "distant" RAM on another node, and to access memory on an adjacent K8's memory controller will have lower latency than to access memory allocated in the bottom-right CPUs RAM. A good OS aware of the issues can try ensure to keep processes on the CPUs to which that processes memory is "local" and hence maximise performance, but it's quite a juggling act (Linux has some NUMA support).

What AMD will do for multi-core we dont know. For certain the individual cores will be connected by HyperTransport. Most likely AMD will give each core their own dedicated memory controller, which would simply make a multi-core SMP be exact same in terms of architecture as the current dual K8 architecture (ie 2xx opteron), and hence no different in terms of bandwidth contention than for existing SMP Opterons.

It will make large SMP machines a lot easier to build though. Eg

Re:Why not quad core?

[Anonymous Coward]

Yeah, we bought a Quad Opteron system and when the vendor brought it in he dropped it and broke one of the processor sockets. It worked fine but only had three active processors while we waited for the replacement machine.

Re:Why not quad core?

[Paul Jakma]

Self-correction: Apparently it might be just _one_ memory controller per die, which may or may not itself be dual channel (I gather from other posts). Also, obviously each CPU potentially has additional HT links to connect to things like PCI bus controllers, AGP controllers, etc. (the basic block diagramme for the Tyan S2885 dual K8 board shows the AMD-8151 AGP controller and the AMD-8131 PCI-X controller wired to CPU0).

Re:Why not quad core?

[HuguesT]

I'm not sure I understand what you mean. If you mean that BeOS makes things simple for the developers by allowing use of multi-processors transparently then I think this is a bit of an optimistic view.

Developers would still need to write their code taking multi-processors into account. This is not so easily done with standard languages such as C/C++ or even Java etc, which do have support for multithreading.

Support for threading and transparent threading usage are two different things. Usually the former is easy to implement, hard to use well for developers but when used well it does provide significant benefits for the end-user, and the latter is hard to implement, easy to use for the developer but does not usually deliver as much benefits as the hand-coded parallel-loop approach.

As an illustration, Intel provides a parallel C++ compiler for Linux and Windows. If you let the compiler do the parallelization it does work but the benefits are not great in general, unless the developers take great care in writing their code, essentially laying out cues to the compiler. If the developers do the parallelization by hand using OpenMP or threading libraries, then the code is even harder to write well but often it delivers better results: better scaling, better load balancing, etc.

Re:Why not quad core?

[dragonknight831]

Another reason to not switch to a quad core is the expected yield of the processors fabricated on a single wafer. Currently, if one core has an error, that single core can be discarded; however, if there are 2 cores in each CPU, they both must be operational, and thus lowers the expected yield of a given wafer.

Re:Why not quad core?

[Anonymous Coward]

HyperTransport is getting upped to 2Ghz. That's 1Ghz each way. And in case you didn't know it's currently at 800 Mhz. That's a good jump.

Also I read (when they first announced this) that it will be socket 939. Or at least the AMD guy said "current motherboards" will be able to run these chips. Probably just need a BIOS update...

Re:Why not quad core?

[Hoser McMoose]

Of course it requires a few caveats, but it's mostly correct. The Opteron has a rather impressive level of bandwidth/processor and extremely low latency to that memory. The integrated memory controller REALLY helps here and I suspect that for over 95% of the cases AMD is correct that they could drop a second core into the picture and still get very good scaling without having to worry about memory bandwidth.

Ohh, and as for cache, that fortunately isn't a problem at all as each core comes with it's own 1MB of L2 cache (and of course it's own 128K of L1 cache as well). Combine that with the huge amount of I/O bandwidth they've got with Hypertransport and really the chip is well designed to accept a second core.

Re:Why not quad core?

[yuudoku]

Sure, but so is 1. 2^0=1

Re:Why not quad core?

[chip33550336]

I will have to back up the number 6. It is perfect.

Re:Why not quad core?

[3D Lover]

Some cheap dual opteron boards save traces by pairing one cpu with all the memory banks - so every memory operation on the non directly linked cpu passes over the h-link.

Care to elaborate? I'm researching dual opteron boards right now, and if some of them are "cheap" in the sense that they can hamper performance, then I'd love to know that ahead of time.

Re:Why not quad core?

[Anonymous Coward]

A question. I was under the impression that SMP (symmetric multi-processing) refers to a design where each processor has equivalent access to system resources, and this isn't true in a NUMA design. Yet you seem to know what you're talking about, and refer to the K8 multi-CPU setup as SMP. What's going on here? Has the definition of SMP changed or am I confused?

Re:Why not quad core?

[hxnwix]

here's one with all the banks on one cpu
here's a nice little overview

Re:Why not quad core?

[DMUTPeregrine]

Just like razor blades? What are they up to now, 8? The Octahypersonic razor!

Re:Why not quad core?

[Paul Jakma]

I was under the impression that SMP (symmetric multi-processing) refers to a design where each processor has equivalent access to system resources

Right, it implies a general symmetry in processing capacity.

this isn't true in a NUMA design.

It is, sort of. At least if you restrict your definition of IO to RAM. In general, in a NUMA machine, for every CPU the proportional difference in locality of RAM will be the same, ie it is still symmetric. If you want to define IO generally, then even the 2-way Opterons are not SMP - as i stated earlier the Tyan 2885 has PCI and AGP IO both going via CPU0, CPU1 only has HT links to CPU0 and RAM.

What will be interesting is 8-way. That conceivably could look something like:

IO-0--1--2--3-IO
| |\/| |
| |/\| |
IO-4--5--6--7-IO

Ie, if each CPU had 3 external HT links (as i gather the Opteron 8xx's have) one conceivably could have a system where a set of core CPUs (ie 1,2,5 and 6) had fully-meshed links, forming a 'crossbar' with a set of 'peripheral' CPUs (0,3,4,7) then having the spare links for IO beside their 2 links to other CPUs. And each 8xx K8 (I gather) additionally has 2 HT links to 2 on-die dual-channel memory controllers. See:

http://www.xbitlabs.com/images/news/2004-04/optero n_8p.gif

That architecture is pretty much what DEC were touting as the state of the art in the very late 90s and began selling at the beginning of this century in their Wildfire product (the big GSxxx massively SMP machines, 32 CPUs+). Except Wildfire was/is based on the EV6 bus, also point-to-point (which Athlon uses), but requiring dedicated EV6 hub controllers to stitch the CPUs and various IO together. HyperTransport is the logical generalised progression of that, which perhaps isnt really surprising considering AMD managed to acquire quite a few former Alpha engineers who really did not care to go work for paranoia-as-mantra intel when DEC got their rather odd settlement from intel for patent violations, including Rich Witek, lead architect for the initial Alpha AXP CPUs, who is now an AMD fellow.

So, is it SMP? I dont know, if you want to really pick nits, then I guess it doesnt quite meet the classical definition, but it is the logical progression of SMP, and the resources while not being completely symmetric in terms of locality are still uniform in terms of accessibility, which i think at least meets the spirit of the definition.

Re:Why not quad core?

[Anonymous Coward]

Thanks for the answer! I guess it's basically impossible too maintain full technical precision in that widely used terms, and I suppose these terms acquire consensus meanings among experts with an intuitive understanding of the field. Now, if only I had a project where I could justify a purchase of a toy like an 8-way Opteron...

Re:Why not quad core?

[CommieOverlord]

Well, Sun is planning on taking their UltraSparc processors to 32-core within the next couple years.

Re:Why not quad core?

[Anonymous Coward]

..and so is 3. 2^(ln 3/ln 2) = 3

Re:Why not quad core?

[stevesliva]

Four x86 cores.

Re:Why not quad core?

[Fulcrum of Evil]

You mean 8 [and not 4?]. This is a computer, you're legally bound to use a power of 2.

And this got a 5 insightful, how?

Re:Why not quad core?

[rrhal]

It's coming, be patient. AMD has to let your wallet recover from buying the dual core chip before they bring out the 4 core units.

"Honest dear, we need 4 2.4 GHz opteron processors to check our email - our e-mail client is outlook.net"

Re:Why not quad core?

[Donny Smith]

>The upshot of this is that it makes Linux all that more attractive since there's no per processor licensing bullshit to deal with,

Actually licensing for such commercial apps doesn't change - it remains per physical CPU, no matter how many cores the CPU has.

>It's pretty easy to predict that what will happen is Microsoft will be forced to amend or drop its per processor licensing scheme,

As above - if that does happen, the technology itself won't be the reason.

>Then again, they might have enough clout to hold up AMD and Intel's release of the multi-core CPUs...

What? Xeon MP has been out for years and with Xeon MP-based servers, such as IBM x440, when you type "top" you see twice as many CPUs. These of course are supported in Windows as well - use Task Manager to see it in action.

Re:Why not quad core?

[Junior J. Junior III]

Do you have any information to back those claims up? A link to an official statement would be nice.

From what I understand, if the OS "sees" two processors, then for all intents and purposes for licensing you're going to have to obtain a license for two processors.

Ie, if you buy a 2-processor license for MS SQL, and install it on your 2-socket server which happens to have 2 dual-core procs in it, and the OS decides that it views the 2 dual core procs as 4 procs, then you're only going to be able to use 2 of those detected processors to run SQL, and would have to shell out dough for it to run on the other two procs (as seen by the OS).

I wonder what would happen if each core on the dual-core chip supported Hyperthreading as well... You'd have 8 "processors" in 2 sockets.

Re:Why not quad core?

[AuMatar]

Damned if I know. I was aiming for 3 funny or unmodded.

Re:Why not quad core?

[TheLink]

Coz AMD already spec'ed for dual cores at the start.

Look at the thermal design stuff where AMD said 80+ Watts for their entire range of chips no matter what the frequency. Some silly people thought it meant the Opterons/Athlon64s ran hot and said dumb things.

What AMD actually was doing was making sure that as many PC/Server makers as possible would make designs that worked with 80+W chips.

With current chips dissipating from 30W-55W, "drop in" dual cores are definitely feasible. Make a 2 x 40W dual core chip, and it'll work.

If AMD's move to the next manufacturing process actually creates cooler chips that'll even be better. In Intel's case the move doesn't seem to have helped - see the Prescott. However it could be that all that heat is because Intel is trying to avoid using some stuff that it doesn't have patents for (e.g. IBM's stuff).

AMD has a deal with IBM for Opterons. Will be interesting to see how far that goes.

Hope AMD stays alive - so far they're making things quite interesting and in good ways too. Whereas Intel is doing stuff like Itanium, Prescott and BTX.

Re:Why not quad core?

[mike_c999]

I guess 'cus a lot of people here believed him.
Damn thats worrying.

A lot more info over at anandtech...

[MarkWPiper]

linky linky!

Re:Just get...

[hawkbug]

No I won't. I prefer a system I can build - to get a dual core k8 all I'll need is a new board and a single processor that doesn't require 5 loud fans in the case to keep it cool enough so I can actually use it without crashing.

Re:Just get...

[DAldredge]

Why?

for additional AMD dual core story links

[ruiner5000]

you can find them all here. It seems news has gotten around, and that AMD's dual core will consume just about as much power as a single core CPU at 90nm.

End of moores law?

[sploxx]

Is this the end of moores law, at least in the form of CPU speeds doubling every 18 months?
There are essentially two CPUs, I doubt each of them will get 2x faster the next 1.5 years :)

Re:End of moores law?

[DAldredge]

Moore's Law has NOTHING to do with CPU speed.

from a google search.

Moore's Law /morz law/ prov. The observation that the logic density of silicon integrated circuits has closely followed the curve (bits per square inch) = 2^(t - 1962) where t is time in years; that is, the amount of information storable on a given amount of silicon has roughly doubled every year since the technology was invented. This relation, first uttered in 1964 by semiconductor engineer Gordon Moore (who co-founded Intel four years later) held until the late 1970s, at which point the doubling period slowed to 18 months. The doubling period remained at that value through time of writing (late 1999). Moore's Law is apparently self-fulfilling. The implication is that somebody, somewhere is going to be able to build a better chip than you if you rest on your laurels, so you'd better start pushing hard on the problem. See also

Re:End of moores law?

[slash-tard]

Moores law is doubling of transistors, not clock speed. This is an easy way to keep up, double the core, double the transistors.
http://www.intel.com/research/silicon/mooreslaw.ht m

Re:End of moores law?

[The Uninformed]

no, Moore's law covers the density of transistors on the CPU and not the speed.

Re:End of moores law?

[soundsop]

Is this the end of moores law, at least in the form of CPU speeds doubling every 18 months? There are essentially two CPUs, I doubt each of them will get 2x faster the next 1.5 years :)

There have been quite a few posts pointing out that Moore's law actually refers to exponential growth in transistor density rather than speed.

The posters are technically correct, but the term Moore's law has come to encompass any processor-related metric that changes at an exponential pace, including processor performance, clock rate, and power consumption. Of course, these metrics are directly related to transistor size and density, so it makes sense that they have changed exponentially.

For those with access to IEEE articles, Gordon Moore (Intel founder, who Moore's law is named after) wrote an interesting paper called No exponential is forever: but "Forever" can be delayed!.

Re:End of moores law?

[Titchener]

Moore's Law has NOTHING to do with CPU speed. Um, I think he meant Moore's Law as it is popularly conceived. Who's the dummy now, big shot?

Re:End of moores law?

[DAldredge]

Easy, the people who don't know what they are talking about but post crap as facts anyway.

Just because people think something means something, doesn't mean they are right.

Re:End of moores law?

[sploxx]

Maybe you think "it's easy to say that now", but I already know at the time of posting this that moore s law covers processor density.

I did a talk about the physics of field effect transistors, esp. MOSFETs just yesterday. So, I probably know what I'm talking about, at least more than the avg. /. crowd, but why bother... :)

Demise of processors predicted!

[ackthpt]

As the number of pins continues to increase the mass does also, at some point processors will achieve such a large mass they will collapse in upon themselves.

actually it'll probably be more like the processors gets so big that you just clip things onto the outside of it and it takes the place of the motherboard.

Re:Demise of processors predicted!

[ocelotbob]

The New AMD Blackholeteron processor! It really sucks!

Re:Demise of processors predicted!

[Anonymous Coward]

As the number of pins continues to increase the mass does also, at some point processors will achieve such a large mass they will collapse in upon themselves.

Nope. that's where the heat comes in. As they get larger they'll also get hotter. The heat will cause an internal pressure that will counterballance its own gravetational attraction.

Eventually fusion will occur and the power supply can be moved on chip also.

Re:Demise of processors predicted!

[ruiner5000]

funny, except that AMD's dual core is pin compatible with current motherboards. When you research your jokes punchline it becomes funnier.

Re:Demise of processors predicted!

[RyuuzakiTetsuya]

uh

read.

He means that some point down the road, the CPUs will be bigger than the boards themselves, thus the joke.

Re:Demise of processors predicted!

[Anonymous Coward]

It really sucks!

That's not news.

Re:Demise of processors predicted!

[JoeBuck]

Rather, processors are consuming so much power that they will melt when you turn them on. Pushing chip design forward is getting tougher and tougher, and the processor architects have decided that it's easier to put multiple CPU cores on a chip (without increasing clock speed) than to continue to boost the performance of individual CPUs.

This spells trouble for many programmers. A lot of people have gotten spoiled; they can keep adding features and bloat to their programs, making them slower and slower, knowing that Intel and AMD will put out faster processors every year. The paradigm is changing; instead of faster processors, you'll get twice as many. That means we're going to have to learn how to parallelize our apps.

Does dual core mean dual licensing costs?

[schwep]

I have seen some licensing schemes that apply to per-processor costs... 1 CPU = $1,000, 2 CPU = $2,000 etc.

How long will it take to argue that consumers with a dual core processor should pay 2x the price? I'm betting not long.

Re:Does dual core mean dual licensing costs?

[Fooby]

It's a lot simpler than that, really. If you have a single-CPU license the product will probably only utilize a single thread anyway. So that application will only use one core at a time and you won't benefit from parallelism in that application. If you buy a 2-CPU license and run two threads then it will take advantage of it. Unless you're talking about something like operating systems. Of course things get hairier if licensing is not enforced in the software. Personally, if I've got one chip with two cores I would say I still had one CPU, even if it looks like two from software. But I'm sure the bean-counters will clear this all up once these things get popular.

Of course the real solution is to use free software when practical.

Re:Does dual core mean dual licensing costs?

[Anonymous Coward]

Yep, that would be $2794 for my next dual x dual core Debian boxen. Unfortunately it will be available only next year and certain company will certainly not be arround by that time. I am going to be very confused to whom address the money...

Re:Does dual core mean dual licensing costs?

[NerveGas]

Forget dual cores. What about Xeons, with Hyperthreading - where the system "sees" two chips? Nothing like paying a dual-processer license so that you can use a single chip.

steve

Re:Does dual core mean dual licensing costs?

[Vaystrem]

Thats actually an interesting point in the consumer space even. WinXP Home is not SMP Enabled but WinXP Professional is. And historically we've seen that with Win9x vs NT based OS (Yes this was an architectural limitation as well but still based upon expectation of the computing needs of the average consumer)

So when, in one or 2 years, dual cores become more common possibly on upper end desktops - will Microsoft have to ditch their policy of charging more for SMP enabled operating systems for consumers?

Re:Does dual core mean dual licensing costs?

[Wesley Felter]

What does it matter whether the cores are on one chip or two? A dual-core system has roughly the same performance as a traditional dual-processor system, so the licensing cost should be the same.

Re:Does dual core mean dual licensing costs?

[scotch]

It's pay for performance now? Where's my big discount for runing XP on my 300 MHz K6-2?

No, I don't really run XP on a K6-2

Re:Does dual core mean dual licensing costs?

[Anonymous Coward]

not always i've had some version of 3d studio i think it was, i dunno it was a while ago that refused to run on a dual cpu machine without the proper license

Re:Does dual core mean dual licensing costs?

[Anonymous Coward]

I have xp running on THREE k6-2 machines....2 550 and one 533....runs fine.

Re:Does dual core mean dual licensing costs?

[snero3]

How long will it take to argue that consumers with a dual core processor should pay 2x the price? I'm betting not long.

unlikely to happen, oracle has a license structure that charges per CPU, so for dual core CPU's like the power4 oracle only counts physical CPU's same with the xeon HT(yes I know HT is not the same as dual core) etc...

Think about it this way, when you buy a machine from these guys you don't spec it out by the cores but rather by actual CPU's.

Re:Dual core for a PC?

[Anonymous Coward]

What the hell are you smoking?

What AMD is really doing

[bugnuts]

They're making the first Desktop Fusion Unit!

AMD K9 barks up the wrong branch (prediction)?

[Anonymous Coward]

I thought AMD won't have dual core CPU until K9. Looks like the company is barking up the wrong processing branch. :)

Re:AMD K9 barks up the wrong branch (prediction)?

[ebrandsberg]

If they waited for K9, then they wouldn't have any bark left in their bite. You have to throw them a bone for putting Intel in the dog house.

Re:Just get...

[Anonymous Coward]

You will be wrapped in the moist warmth of the SJRDF.

939 is now

[aka-ed]

Why would you wait a year?

http://www.theregister.co.uk/2004/06/01/amd_939/

Re:939 is now

[Anonymous Coward]

...because if the dual-core isn't socket 939, then he'd have to upgrade his mobo AGAIN when the cpu came out.

Re:939 is now

[AKnightCowboy]

...because if the dual-core isn't socket 939, then he'd have to upgrade his mobo AGAIN when the cpu came out.

Isn't that fairly obvious though? I've always factored in the cost of a new motherboard and new memory to go along with any new CPU purchase because I know the crap I'll have will be obsolete in a year or two when I go to upgrade the CPU. I have NEVER upgraded to a faster CPU by swapping it out and keeping the same motherboard because then I feel cheated as I'm left with a CPU with no home. I then invariably feel I need to go buy a motherboard and processor to keep the old CPU company and I end up building a whole new (old) system to act as a Linux server. It gets expensive to upgrade!

Re:939 is now

[JWSmythe]

In a year, there'll be something bigger and faster (and probably incompatable) out anyways. It's always like that, if you haven't been paying attention. :)

Even if you bought a motherboard today, in a year the top of the line processor probably won't work on it anyways.

Re:939 is now

[aka-ed]

Exactly. A little restraint peoples! When the dual-core comes out, snap up a 939; when quadruple core is released, get the 939 dual core. Stay a step behind the bleeding edge and you can afford to upgrade more often.

Re:939 is now

[JWSmythe]

I've always bought close to bleeding edge. Like machines that won't break the bank, but are fast.

My first x86 was a 486/33, which I got just after the 50's came out. Whoohoo.

My next machine was a AMD K5/133+. If I remember right, retail on the CPU alone was $200. That machine is now running a friends personal website. :)

Right now, my machine is a 1.1Ghz Athlon. It was screaming fast for the day, but now I'm looking at my girlfriend's 2800+ saying "damn, more than twice as fast as mine."

Even our mail server, a bleeding edge Opteron 242, is slow now. We're getting parts in now for a pair of Opteron 848's.

After setting up Fedora in the mail server, I'm looking forward to an Opteron workstation. The first time I set up the mail server, I installed X and all, just to see it work, then I reinstalled with as CLI only.

Re:939 is now

[Paul Jakma]

I have NEVER upgraded to a faster CPU by swapping it out and keeping the same motherboard

I've done this a few times actually. Years ago I upgraded my machine from a P133 to a K6 200 and only recently I upgraded it again with a K6-II 450MHz someone had spare. And the other day I upgraded my current desktop's 600MHz Athlon to an 800MHz which again someone didnt want.

So you're obviously talking crap! ;)

Re:939 is now

[Anonymous Coward]

just swap motherboards with your girl friend

P.S. - don't tell her

Re:939 is now

[JWSmythe]

I can't. She uses the enemies software (XP Home, for her games), and it won't allow that. :( At least without begging some Redmond support minion for permission to do something so daring.

I'll be so happy when more authors are writing games for Linux, or Wine is much farther advanced.

I'd rather pay the Wine people money, than pay Microsoft, and have to ask for permission to change my own hardware around. Hell, this is *MY* equipment, and I did pay for the software, if I want to swap parts between my machine all day, isn't that *MY* business?

A year?

[aardvarkjoe]

"Will this be socket 939 or should I try to hold out another year to buy?"

You're planning on waiting more than a full year between computer upgrades? Are you sure you're on the right website?

Re:A year?

[Anonymous Coward]

Slashdot is divided into two groups - those that upgrade every two months and those still running a 486!

Re:A year?

[Anonymous Coward]

>Slashdot is divided into two groups - those that upgrade every two months and those still running a 486!

Also known as 'the winners' and 'the losers', respectively.

Re:A year?

[Anonymous Coward]

Also known as 'the winners' and 'the losers', respectively.

Actually, they're all known as losers.

Re:A year?

[Anonymous Coward]

pffft, I haven't upgrade in 2 years (so far). My 2.2 Ghz P4 laptop is doing just fine.

However, I wouldn't mind upgrading, I just haven't found "the next thing" yet. I want something low-heat (low-power, I guess) that runs really freaking fast. SMP would be OK, but if a single proc is fast enough I wouldn't need SMP (and could save some heat generation). And it needs to be cheap.

Opteron is looking good but they are pretty damn expensive for anything other than the super-slow chips. Athlon64's seem way too hot and no SMP. Xeons are out-of-date (might as well go P4 with big cache). P4's can't do SMP and are starting to get out of date. I don't know much about the P4-M but it seems promising; the clock speeds don't seem like it will compete with the Opteron or G5.

My next machine must be able to compete with the top Apple machines (I will not be buying an Apple though; $$$$$). AFAIK, those new dual 2.5 Ghz Apples are pretty much top of the heap right now, even compared to the fastest Opteron.

Re:A year?

[gr8_phk]

Dude, I'm still running an Athlon 700 (slot A). I write CPU intensive ray tracing software (see rtChess). I don't give a hoot about a simple doubling of speed. I told myself that I'd buy when they tripled which didn't happen until the 2.1GHz or about 2800+. By that time AMD64 was overdue, so I figured I'd wait for that. Then it was a question of which socket and a stable 64bit OS (which is here now). I was going to buy 2 months ago and then realized that socket 754 is going away. I'd like to be able to upgrade my CPU this time. The 700 runs everything fine, and for realtime ray tracing I can just reduce my image size to get a feel for what it would be like on todays hardware(not fast enough anyway). Patience is key, and it seems like there's always something great just around the corner that will require a whole system upgrade.

shall I invest in heatsink business?

[cyfer2000]

I could see a big future of heatsink business in Intel and AMD's plans.

Re:shall I invest in heatsink business?

[hackstraw]

oh really.

Re:shall I invest in heatsink business?

[evilviper]

What, you mean you're just now seeing the writing on the wall? It's been obvious since the first Pentium that continual heat increases are the future.

I don't think heatsinks are much of a business, since there are very good heatsinks very cheap, and there aren't going to be leaps and bounds of improvements that warrant paying lots more for them.

Re:Just get...

[ruiner5000]

Sure, if you are happier not only with liquid radiator cooling, and also having to have copper heatpipe cooling. That is right as I have discovered here Apple has had to implement not one, but two separate cooling solutions for their 2.5GHz PowerMac G5. What were you saying again? You do realize don't you that you will be able to swap out a single core dual Opteron system with two dual core CPUs and have Quad CPU power don't you? And that makes the G5 an advantage how?

You'll need a new motherboard.

[filledwithloathing]

Will this be socket 939 or should I try to hold out another year to buy?"

You'll need a new motherboard.

The DDR memory interface appears to wrap around both L2 caches, meaning that it looks like both cores have their own 128-bit memory interface; whether or not both memory controllers will be enabled is another thing, but if this is true we have a number of implications to talk about. If dual core Opterons do indeed have two memory controllers, the pincount of dual core Opterons will go up significantly - it will also make them incompatible with current sockets. AMD is all about maintaining socket compatibility so it is quite possible that they could only leave half of the memory controllers enabled, in order to offer Socket-940 dual core Opterons. AMD isn't being very specific in terms of implementation details, but these are just some of the options.

Re:You'll need a new motherboard.

[Penty]

No you won't need a new motherboard. Look in this article at Inforworld.

"AMD's dual-core server processors will share a single memory controller, Weber said. This won't create a bottleneck because a server with two Opteron chips, and therefore two memory controllers, already has more than enough memory bandwidth required to run that system, he said."

Penty

Re:You'll need a new motherboard.

[hattig]

Read up on the Opteron die layout.

This is NOT two ENTIRE Opteron processors plunked on the same die.

AMD have designed the ability to connect to TWO cores into the SysReq part of the processor since the beginning.

The SysReq connects on the other side to a crossbar that connects in turn to the HyperTransport Controller and the Memory Controller.

A dual core processor will still only have a 128-bit memory controller.

AMD have stated that the processors will be socket compatible. This also suggests that S939 and S940 are already future proofed for DDR2.

Re:You'll need a new motherboard.

[MBCook]

Nope. Sorry.

I understand your reasoning, but according to this article (I found the link on Ace's Hardware) the dual core chips will be compatible with current motherboards and sockets with as little as a BIOS flash (to recognise the new CPUID I assume). The downside of this is that the two cores will SHARE the dual channel memory bus. But because the bus is so effiencent, each core will probably STILL get more bandwidth than most P4s. At worst it shouldn't be much worse than having two single channel Athlon64s (which also are often faster than the P4). I think this is FANTASTIC news. For one thing, this means you could put FOUR CORES in that dual opteron SFF PC that was revealed a short while ago.

Really, it only makes sense. A dual channel processor has 939 pins, a single channel has 754 pins. So while some are power, you're looking at about 190 pins for the second memory channel. So that would mean that to have two cores on one die with their own memory channels, you'd need 1120 pins or so. That's a LOT of pins.

Instead of that enginering nightmare (you'd probably need 7 layer mobos to support that), we get drop in replacements that meet the same thermal requirements. Just think. You're dual operton not cutting the mustard any more? Buy two processors, drop 'em in, flash the BIOS, and now you've got FOUR processors without a new mobo or anything. All you'd have to worry about then is software licenses (unless of course you don't use any software that requirs that, for example you're all open source).

So to answer the grandparent's question, I'd say buy now. That said, I'm not sure if socket 939 will get dual cores or if it's only for 940s. I assume 939 will get them too.

Speculation: I'd like to know if the dual channel memory controler is shared by the two cores (like some kind of cross-bar architecture thing like nVidia used to promote) or if each core got exclusive access to one of the two channels. My guess is the former.

More speculation: Will there be a socket 754 dual core? That'd be cool, and I don't think the performance would be too much of a problem memory wise, unless you were doing memory intensive tasks. For CPU bound tasks I think you'd be fine.

Re:You'll need a new motherboard.

[ruiner5000]

No you won't. Infoworld got it right. Anand should have researched before he put up his story.

AMD's dual-core server processors will share a single memory controller, Weber said. This won't create a bottleneck because a server with two Opteron chips, and therefore two memory controllers, already has more than enough memory bandwidth required to run that system, he said.

"It's always a juggling act to add a little more processing and a little more memory. Right now, we have plenty of memory and I/O bandwidth, so we're adding processing," Weber said.

The dual-core chips will work with current socket technology in motherboards that are rated for the specifications of the dual-core chips, Weber said. A BIOS change will be required, but otherwise the chips will work in the same sockets as single-core Opterons, he said.

Re:You'll need a new motherboard.

[ameoba]

Nice job. You quote the previous poster, write 5 new words and plagarize a paragraph from TFA and get modded to 5.

Re:You'll need a new motherboard.

[filledwithloathing]

No you won't. Infoworld got it right. ....A BIOS change will be required, but otherwise the chips will work in the same sockets as single-core Opterons, he said.

Opterons don't work in Socket 939 as per the OP's question, only S 940. This implies that if he got a 939 motherboard there's a good chance his motherboard won't support them.

Could AMD design them for S 939? Sure. But if they're going to I wonder why they specifically phrased it as working with Opteron boards.

Re:You'll need a new motherboard.

[JDevers]

There would actually be more than 1120 pins, the 190 pins is just for the SECOND controller...so you could say that the memory controller on a dual channel system is ~360 pins, so in a dual channel, dual controller system 720 pins would be used for the memory controllers alone. Obviously this is not 100% accurate, but is as close to real numbers as we can get without looking at a white sheet ;) That would mean that a dual core, dual controller, dual channel opteron (2x2=4 channels) would have ~1300 pins.

I'll agree with the rest of your statement, just saying that it is even MORE unlikely than you point out...a 1300 pin CPU isn't coming to the desktop or low end server market any time soon (I don't know of ANY CPU with that many pins, but I guess it is possible in a market without extremely tight budgets).

Re:You'll need a new motherboard.

[MBCook]

Now how did I miss that. Good point. Either way, I think we can all agree that having more than 1000 pins in just HIDEOUS. I think we'll only see pin counts decrease from here for most people. We've seen many things go to serial with things like HT (an example would be the chipset connection). And if something is done about memory (say switching to that FBRAM or whatever it was called awhile go that was normal ram with an "interface" chip on it so you could switch out RAM techonologies and not have to change the chipset/processor/mobo) we'll probably decline or at least hold steady.

The only reason that I can see to add more pins would be for power. Like I said, you'd probably need 7 or 8 layer mobos for such high pin counts as 1300, so your board would cost $$$.

Re:You'll need a new motherboard.

[Hoser McMoose]

The dual-core chips are, at least initially, going to be targetted at the high-end (and high-$$$) market, ie the server and workstation market where the Opteron currently competes. At least according to this roadmap though, AMD will also be bringing out a dual-core Athlon64 FX chip. This chip could very well be in a socket 939 format (presumably all Athlon64 FX chips going forward will use that platform).

Re:You'll need a new motherboard.

[fatphil]

You're of course right, but do remember that IBM have had 5000+ pins on their chips since last year.
(~half are power.) And yes, these systems do not come cheap. However, AMD would not be breaking any new ground by heading this way.

FP.

Re:Just get...

[FirstTimeCaller]

So what's the point of having a fast CPU? Video encoding? Photoshop?

Answered our own question, we have.

What we really need...

[rewt66]

is dilithium cores!

Re:What we really need...

[sharkey]

Yeah, go ahead, jump on the latest incremental speed increase. I'M waiting for trilithium cores. Transwarp computing, here I come!

Re:What we really need...

[cwg_at_opc]

is a berylium sphere!

In answer to poster's socket question:

[MarkWPiper]

From the article. "If dual core Opterons do indeed have two memory controllers, the pincount of dual core Opterons will go up significantly - it will also make them incompatible with current sockets. AMD is all about maintaining socket compatibility so it is quite possible that they could only leave half of the memory controllers enabled, in order to offer Socket-940 dual core Opterons. AMD isn't being very specific in terms of implementation details, but these are just some of the options."

Re:In answer to poster's socket question:

[MBCook]

As I posted elsewhere in this thread (link), that speculation seems to be wrong, which is good. Source for that info is here.

Re:In answer to poster's socket question:

[Too Much Noise]

Well, they seem to go against what AMD has been telling so far: there's one memory controller designed to service core0 and core1; only the core0 link is used so far, core1 will be enabled when an actual second core is available. This is part of what they mean by 'designed for multicore from the beginning'. They pretty much want the dual core to be a drop-in replacement (+ BIOS update, provided that the motherboard supports that - for one thing, higher power consumption per socket would be expected).

Re:In answer to poster's socket question:

[Hoser McMoose]

Whoever wrote that article over at Anand was a bit confused or simply didn't bother reading up on this or even looking at their own pictures. If you look at the die photos it should be fairly easy to tell that there are *NOT* two memory controllers, just one. The memory controller and hypertransport controllers go around the outside of the main processor core(s) on the die. There's just one memory controller, 128-bits wide, just like in the single-core Opterons.

That memory controller might be updated to support DDR2 memory (this would fall into AMD's timeline for supporting DDR2), so we might see some extra memory bandwidth in that sense, but there will be no need for extra pins. AMD should have no trouble making this fit in Socket 940 as they have mentioned several times in recent interviews.

Hmmmm....

[spikev]

Will this be Sockett 939?

It seems highly unlikely that it will, seeing how the extra pin on 940 is supposed to deal with/control multiple processors.

Mod parent frickin' funny.

[Anonymous Coward]

Every time anyone says "Apple is moving to Intel" it's frickin' hilarious.

should I try to hold out another year to buy?

[polyp2000]

To be perfectly honest, it depends how rich you are. At the end of the day when it comes to buy now, buy later; the state of technology generally speaking is that in most cases (particularly with computer hardware) after only a short period of time , whatever technology you invest in becomes obsolete.

From my own personal point of view, my dual athlon 1.5ghz is still holding out beautifully. When the cash comes my way Im banking on a powerbook. Truth is I dont need another desktop just yet. However if i had a stupid disposable income, and one that predictably would hold out till these dual cores come out id proabably get one now, and get one later.

When I built this machine I bought the highest spec parts I could afford at the time and I havent upgraded for 2 or 3 years aside from upgrading the graphics card. The rule I live by is get the best available that you can afford at the time and it should keep you going for a good while.

Im running gentoo box; faster processors would be very nice for source compiles but I gave up on churning out seti blocks a while ago and dont have a massive reason for further processor power ...

Re:should I try to hold out another year to buy?

[mcrbids]

The rule I live by is get the best available that you can afford at the time and it should keep you going for a good while.

Which is perhaps the most expensive way to get what you need.

I take a look at pricewatch under "hard drives", here's the matrix:

CAPACITY PRICE PRICE/CAP

300GB $232 $0.77
250GB $158 $0.63
200GB $101 $0.51
180GB $100 $0.56
160GB $77 $0.48
120GB $58 $0.48
100GB $58 $0.58
080GB $48 $0.60

Notice that the price starts at a high of 77 cents per GB, then falls almost 40% in price per unit down to $0.48. It's quite a bit cheaper to get two 160GB drives than a single 300.

The price rise (per GB) you see going from 160 down to 80 is change from "best bang for the buck" to "cheap and works".

So, unless you really HAVE TO HAVE that top-end part, it's best to shoot for midline. You'll end up with a system for quite a bit cheaper that still plays all the latest games and does all the latest stuff, and you can spend the money you saved on your significant other!

Re:should I try to hold out another year to buy?

[ceswiedler]

and you can spend the money you saved on your significant other!

Signifigant other...you mean my laptop?

Re:should I try to hold out another year to buy?

[boots@work]

I think an even better rule of thumb is to buy one step down from what you can afford. You're likely to suffer a bit less of a novelty premium and it will still last almost as long.

So for example at the moment (in AU) 250GB ATA/SATA drives are affordable, but going down to just 200GB is 35-40% cheaper and only 20% smaller. Either one will likely last two-three years, at which point I'll probably replace it with something in the 500-900GB range.

(Insert tentacle porn reference below this line:)
-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o-o -o-

Re:should I try to hold out another year to buy?

[Anonymous Coward]

I think this is a fair comment. Though it really depends on how you interpret the word "Afford" in my case what I bought probably wasnt the top of the line anyway. When pricing a system you can "Afford" generally speaking one has a ball park figure and you have to make conservative judgements on which bits to spend how much on.
etc...

Re:should I try to hold out another year to buy?

[WuphonsReach]

Out of curiosity, what dual-athlon motherboard are you using with Gentoo Linux?

I've been looking around for one, but it's always good to know a specific one that someone got working.

Re:Just get...

[cyfer2000]

garage band of iLife

Re:Just get...

[Anonymous Coward]

Games. Video encoding. Photoshop.

The only three reasons anyone would ever want to use a computer. You heard it here first folks: alienw has surveyed all computer users and found that those who use computers for anything else are an insignificant minority.

Mod me (+1, Aghast).

Re:Just get...

[shundi]

So is this a figmant of my imagination?

Really nice alternative to dual processor systems

[Vario]

Dual cores processors seem to me like a pretty good alternative to a dual processor system. You don't have the hassle of 2 huge coolers blowing out hot air, the mainboards are don't have to be overpriced and it is already supported by all OS.

Some years ago I was thinking about getting a dual processor system. Alone the motherboard was two times as expensive as a similar single processor one, applications did not support it all and so on. I hope newer applications are ready for dual cores. Quake III was the first game I know that used two processors and finally I can consider that animated desktop background.

Is there a list which applications can effectively use dual cores besides obvious things like webservers?

Re:What about Apple?

[Anonymous Coward]

If anyone is prepared to move to dual core it is IBM with there POWER4 Chips, which the G4 & G5 are based off. If apple was to ask for it, I'm sure IBM could make a dual core PowerPC chip. Afterall the POWER4 chips ship as dual core chips as standard.

Re:Just get...

[Gyorg_Lavode]

PORN

Re:Really nice alternative to dual processor syste

[rebelcool]

Anything multithreaded. Which is just about any modern GUI app.

Socket 939!?

[Temporal]

Wait, socket 939 is real!? I thought the concept of a 939-pin CPU was some sort of hyperbolic joke!

Re:Socket 939!?

[MBCook]

Socket 939 is Socket 940 minus one pin.

They are identicle in features (dual channel) except the 939 lacks one pin. That pin just happens (wink) to be a HyperTransport link that was removed. This means there are not enough links to support multi-processor setups because you can't have the links to the other processors because you don't have enough. This is basically a marketing move to segment the workstation market from the desktop market. There is no (techincial) reason they couldn't have used socket 940 for dual channel desktop processors.

Also, from what I've heard, the pins are layed out differently so you can't put a 939 chip in a 940 board or chip and pin off a 940 chip and do the opposite.

Side note: Can someone clear this up for me? Opterons have three HT links. One for the chipset, and two for connecting to other processors (this allows for configurations of 4+ processors). The single processor desktop parts have one HT link to connect to the chipset. So if you chop a HT link off the Opteron, you get two. That should be enough to connect to the chipset and one other processor, right? Shouldn't a 939 be technically capable of dual processor configurations (although AMD has certainly disabled it)? Just wondering.

Re:Socket 939!?

[CaptainPinko]

Nope. Regular opeterons use 940 pins. They took one pin off going from registered to unregistered RAM. No joke.

Re:Socket 939!?

[NerveGas]

That's still not exactly exhorbitant. The UltraSparc-IV's have about 1400 pins.

steve

Re:Socket 939!?

[Paul Jakma]

The third (or fourth) link is internal, to the memory controller(s).

eg, a 2xx opteron has 2 external HT links, 4xx has 4, etc.

Re:Socket 939!?

[MBCook]

I was thinking there was one to the memory, but since the memory controller was on core I dismissed it. I forgot that HT links can be internal to the chip. Why wouldn't they be able to.

Thanks for the answer.

Re:Socket 939!?

[Temporal]

That's... not exactly what I meant. Just the mere concept of having over nine hundred pins on my CPU scares me. :)

Re:Socket 939!?

[Hoser McMoose]

They are identicle in features (dual channel) except the 939 lacks one pin. That pin just happens (wink) to be a HyperTransport link that was removed.

That is not at all correct. The two sockets are actually significantly different from one another. You can find a comparison of the two layouts of the two chips bellow:

Socket 939 description.

Socket 940 description.

As you can see looking at the pin-outs (pages 26 and 22 respectively) there are a LOT more than just one pin that has changed.

This means there are not enough links to support multi-processor setups because you can't have the links to the other processors because you don't have enough.

Dual-core chips don't use any external hypertransport links anyway, so this is a non-issue. Everything is handled inside the chip. As such, a dual-core Socket 939 chips is DEFINITELY possible, though whether it will actually happen or not is another question.

This is basically a marketing move to segment the workstation market from the desktop market. There is no (techincial) reason they couldn't have used socket 940 for dual channel desktop processors.

Actually there is a good technical reason why AMD introduced Socket 939 for desktop processors, they have a LOT more power and grounding pins. This allowed motherboard makers to design 4-layer motherboards using this socket while Socket 940 required more expensive 6 or 8-layer motherboards.

So if you chop a HT link off the Opteron, you get two. That should be enough to connect to the chipset and one other processor, right? Shouldn't a 939 be technically capable of dual processor configurations (although AMD has certainly disabled it)? Just wondering.

If you have a look at the pin-out diagrams you'll see that AMD has completely removed all the pins for two of the three hypertransport links and replaced them primarily with power and grounding pins on Socket 939 vs. the old Socket 940. As such there really and truly is only one hypertransport link on this new socket and it is not capable of classic SMP configurations.

However, as mentioned above, these dual-core designs are handle all inside the chip and only require a single HT link to the outside world, so they could certainly work in a socket 939 product. Whether such a product ever makes it to market or not is another question, though there are some hints that AMD might release a dual-core Athlon64 FX chip in socket 939.

yes

[StandardDeviant]

MSFT, Oracle, and others already argue that.

The commercial software vendors may be slovenly
about keeping up with security patches, but they
jump on extra money like rabid ferrets on raw meat.

Re:What about Apple?

[HiredMan]

Dual cores have been in the IBM PPC pipeline for quite a while - of course the (now old) Power4 arch has been multi-core all along.

In all probability the PPC little brother of Power5 (rumored to be called the 975) will debut at 90 nanometers and the next chip will be a ~60 nanometer dual core version possibliy called the 976.

Which if these will be called the G6 is left up to the reader as an exercise. My money is on the 976. Either way the PPC has some serious legs.

=tkk

Re:Just get...

[Anonymous Coward]

The reason people buy apples isn't to churn out SETI data blocks, it's for stlyle, and always will be (now). Apple is sort of like the Mercedes Benz of computers, they look nice, work nice, but aren't the power hungry rice rockets PCs are nowadays.

This really sucks.

[HotNeedleOfInquiry]

Now I'll have to pay SCO $1,149 instead of $699.

Yeah, right

Re:This really sucks.

[Sponge Bath]

Now I'll have to pay SCO $1,149 instead of $699.

The other day Darl demanded $699 from me for a Linux license.
"I'm not going to give you $699!", I replied.
Then Darl said: "How about tree fiddy?"

It was at that point I noticed Darl was really the Lock Ness monster.

Well

[Sycraft-fu]

Seeing as the G5 is, more or less, a sinlge core from the larger IBM Power4 processor, I'm not seeing that it would be a large problem to make dual core chips.

I highly doubt Apple will switch to x86, it's a pride thing if nothing else. Also, at this point, a switch would upset everything. It could have been done, potentially, with the OS-X switch. Since software was having to be ported to a new OS, a new architecture port is just one more thing. Now, however, x86 Macs would be binary incompatible with PPC Macs. That means emulation, which isn't very efficient.

I think Apple is pretty much stuck on PPC for good.

Re:Well

[canon006]

Kinda OT, I don't think I'm alone on this but I think OS X on x86 would kinda suck. The best part about Apple machines is the software/hardware integration. That'd be nearly impossible on x86 unless they managed to come up with some kind of proprietary x86 technology and then what's the point? That's not to say I wouldn't welcome it, I just don't think it'd be the same experience.

Damnit, they had to come out with this

[foidulus]

Just when I thought I had saved up enough money between upgrades to splurge on those fancy ramen noodles, you know, the one with the dried peas, this comes along.
Hey, Wal-Mart brand noodles are only 8 cents!

Re:Damnit, they had to come out with this

[o'reor]

I once tried emailing a spammer, got a message back from spamcop...

You did a "reply all", now, didn't you ?

Re:Just get...

[ps_inkling]

Must... not... feed... trolls....

Diablo II, Starcraft, Warcraft
Unreal Tournament 2004, Neverwinter Night, Dungeon Siege, Civ III
Myst, Riven, Exile
Medal of Honor and expansions, Battlefield 1942, Ghost Recon
Ghost Master
Quake III, Beyond Castle Wolfenstein
Escape Velocity Series, among others

There are plenty of other games for the Mac platform as well, check the Apple website for a larger list.

Re:What about Apple?

[Fooby]

No. The PPC architecture has the RISC advantage which makes engineering them that much easier. It would be easier to make multicore PPC than multicore Intel.

Apples are the only RISC-based consumer desktop platform, it would be tragic if they moved towards Intel with all its legacy baggage.

AMD was first, nitwit.

[Anonymous Coward]

AMD was the first to announce dual core. Intel had to re-adjust their roadmap to pull dual core in from 2006 to 2005.

Re:Really nice alternative to dual processor syste

[Vario]

Only if the application is doing time consuming stuff in at least two threads. You say any modern GUI app, so is Firefox rendering a page multithreaded? What about my DVD Player Software, Games, TeX, Maple?

Re:Just get...

[ruiner5000]

I'm fairly certain I like the style of my gaming box, my Shuttle SN85G4 with the black case and the mirrored front finish. That and I like the cost, and the fact that I can fit three of them in a PowerMac G5 case. Now what will be even cooler is when I can make that a dual core box, and not have to upgrade the cooling. Even better yet is when I get that Iwill dual Opteron small form factor in the Summer from those guys, and am able to write an upgrade story on it a year from now when I have dual core Opterons from AMD. Yeah, I will get to enjoy style, and power. Sweet, two patties!

Longhorn

[colonslashslash]

"We recently learned of Intel's plans to go dual-core in late 2005. Well it seems AMD has decided to follow suit."

Its amusing to watch the chip manufacturers scramble desperately to meet the recommend specifications for Longhorn in time.

Oh, c'mon don't look at me like that. A slashdot story without some kind of Microsoft snipe just wouldn't be the same now, would it?

Alright, fine. I'll pick on SCO or AdTi next time. Sheesh. /me crawls back under his rock

Re:Longhorn

[T0t0r0_fan]

As funny as it is, why can't it be true(partly, that is)?
Personally, I'm feeling more than happy with an Athlon1200, and I'm sure a lot of people would as well. So what's the perfect excuse to buy the newest-fastest CPU? To run WinXP/Longhorn/whatever else comes along! Surely you can't have satisfactory performance with only 1.2GHz!
IIRC that's approximately how it has been for the past several years.

Re:Longhorn

[ruiner5000]

what is even more amusing is that AMD began this work well before longhorn was announced. Even more amusing is that AMD announced before Intel. You know, it isn't funny when it isn't even factually correct enough to be so.

Re:Longhorn

[cens0r]

Accept for those weren't the recommended specifications for longhorn. Those were Microsofts opinion of what the average new computer was going to be like when longhorn was released. I'd have to say they hit the nail on the head.

3DFX anyone?

[imstanny]

multiple cores seems like a temporary solution to me... eventually the multiple cores will have to get smaller and deal with heat issues as well, so we'll be back at square 1.5

Re:Just get...

[JWSmythe]

We've had an internal joke in the office for years, that our in-house distribution should be named "Pornix". We think if it sold at $40, and included a one month membership to a couple of our adult sites, we'd make a fortune. :)

Re:Just get...

[Cyno01]

Screw the mirror finish front, finish the whole thing!

You response was half right and half assed

[AHumbleOpinion]

doesn't require 5 loud fans in the case to keep it cool enough

While I understand the desire to build your own and preferring not to be vendor locked, you G5 fan comments are quite ignorant. The Apple G5's are well designed and exceptionally well layed out to create thermal zones serviced by different variable speed vans. It is a very quiet solution. Do not confuse the G5 with some of the homebuilt Athlon abominations that have poor layout, poor airflow, and require multiple screaming fans. YMMV.

Re:You response was half right and half assed

[Lord Crc]

My case is crap. Yet, with a 3000+ athlon, my two WD hd's are "outnoising" my 4 case fans + cpu fan by quite a bit (I notice when they're powered down).
Cpu temps in the mid 50's C. Not what I would call screaming...

Re:You response was half right and half assed

[ameoba]

If you've been following things lately, Intel's P4 offerings have been cranking out far more heat than anything AMD's got comming of the line.

Re:You response was half right and half assed

[hawkbug]

The NEW G5 has quiet fans, the old one did not. I know a guy with one, it was twice as loud as my home built machines have ever been under load. So, Ignorant, no - talking about the old G5 - yes.

Re:You response was half right and half assed

[AHumbleOpinion]

The NEW G5 has quiet fans, the old one did not. I know a guy with one, it was twice as loud as my home built machines have ever been under load. So, Ignorant, no - talking about the old G5 - yes

Sorry, you are still making uninformed blanket statements. I am quite familiar with older G5s. The one that seemed loud was equipped with a Radeon 9800 Pro that was pumping out a lot of heat at the time. Same box with a 9600, quiet. It is not old vs. new G5, it is how loaded one of the thermal zones becomes.

Re:You response was half right and half assed

[hawkbug]

No, still not uninformed - if there were G5s being sold that got hot as hell because of the graphics card, and you just admitted it, I was right. Period. One of the "thermal zones" was hot as hell then, and the fans ran loud when on. Even a novice PC builder knows to find larger, low RPM fans to kill the noise - even when running at full power while reducing heat. Apple screwed up, and no fan boy can deny that.

Re:You response was half right and half assed

[AHumbleOpinion]

A high performance graphics card under load pumping out a lot of heat and causing a variable speed fan to up the RPM is not specific to Mac. This happens on both PC and Mac.

Novice PC builders may be able to install a replacement case fan, but not a CPU fan, that requires a little more expertise, and certainly not a power supply fan. They could manage to completely replace the power supply.

Your blanket statement about G5s being louder than PCs remains false. Grasping at straws and cherry picking Mac and PC examples will not change that. Your "name calling" is also counterproductive, not to mention it was yet another bad and uninformed guess on your part. I've built quite a few PCs over the years, in recent years optimizing for noise. I am quite familiar with both the PC and the Mac.

Re:You response was half right and half assed

[hawkbug]

Fine - but just remember, you made me do it:

http://www.insanely-great.com/news.php?id=2797

http://forums.macrumors.com/showthread.php?s=&thre adid=61377

http://forums.macosxhints.com/showthread.php?t=152 73

I believe that is what you were talking about...

Power supply fan affecting sound card issues here

http://www.osxaudio.com/forums/viewtopic.php?p=797 69

Finally, read the first sentence of this:

http://forums.macosxhints.com/showpost.php?p=12272 0&postcount=16

And there it is - it can be loud or quiet, depending on use - so when you called my comment half right and half assed, you were completely wrong. I was very informed on this topic, and I have no bias towards either PC or Mac as anyone who knows me could tell you. 8 to 9 fans in a system like this? That's insane. My PC has half that, and I run it under full load consistently (SETI, DNET, Movie Rendering, etc) in Colorado, where temps are in the high 80's to 90's all summer long, and I didn't build a machine that needs 8 to 9 freaking fans. I have a high performance graphics card, high performance memory, and a fast AMD cpu - and I'm willing to bet my life that my machine is quieter than any Apple G5 on the market. And, then there's your comment:

"Do not confuse the G5 with some of the homebuilt Athlon abominations that have poor layout, poor airflow, and require multiple screaming fans."

Trust me, I won't, and didn't. I compared a G5 to a machine I would build, or anybody could these days for that matter using very common cases, fans, and power supplies. My machine is not an abomination that requires screaming fans, I use Panflo low rpm fans, and my case is more than sufficently cool. Everything else in my system requires passive heatsinks and runs quit well thank you.

While you may have a quiet G5, many people don't - so you should check yourself before calling me half right and half assed buddy.

why go for CMP and skip SMT

[philipgar]

While the idea of dual core cpus is really cool, and will take over shortly due in part to the fact that we need something to do with all those extra transistors, I wonder why the focus of the industry is on chip multi-processors (CMP).

While CMP processors can give us rougly the same performance of a standard SMP system (somewhat faster due to interprocessor communication and shared memory, but also slower due to a larger memory bottleneck) I don't think that a CMP system would compete with a simultaneous multi-threading (SMT) solution.

While Intel's response to SMT (hyperthreading) has some benifits the performance of it is rather lackluster. The reason has more to do with their particular implementation. If you've read about the initial observations on SMT an 8-way SMT processor was shown to outperform a 4-way CMP processor. Now, I must note that the 8-way smt processor had more functional units then the cores in the 4-way CMP processor, but the overall area of the 8-way SMT processor would be much much smaller (far less structures need to be duplicated for SMT as opposed to CMP). For more information on this check out some of the papers at http://www.cs.washington.edu/research/smt/ .

What I don't understand is the insistance of the industry to use CMP first. From everything I've read, an 8-way SMT processor should take up less die space then a two way CMP processor. Even assuming that the 8 way processor contains more functional units. It kind of makes sense that a CMP processor is faster when there aren't enough threads to fully utilize a SMT processor (say only 2 or 3 threads that want full cpu usage). I guess SMT is a big chance in the model of programming and application development (I'm currently running research on the subject which is why I'm so interested in it). Is the reason to embrace CMPs simply because there's less new technology to add (they "just" have to interconnect two cores as opposed to adding the extra logic for SMT).

Does anyone else have any other opinions regarding this matter, or any idea why no one seems to be fully embracing SMT's potential.

Philip Garcia

Re:why go for CMP and skip SMT

[WeekendKruzr]

SMT is only needed if your execution units are having trouble remaining filled up, which was the problem with the NetBurst architecture due to the huge hits that it takes with a branch mis-prediction penalty. When a mis-predict happens the execution unit has to sit idling away and wait for the proper info to go be re-fetched. With SMT, the unit simply switches over to one of the other threads waiting in the wings which keeps the processor doing useful work instead of wasting cycles. This is why the software has to be re-written to take advantage of it so that the processor knows which threads to give priority to.

Intel stuck SMT into the Pentium in order to balance out the some of the negative effects the go hand-in-hand with a processor that has a LONG pipeline. AMD has a much shorter pipeline (especially when compared to the new Prescott) and therefore they don't suffer much of a penalty when a mis-predict happens. Also, if I remember correctly the Athlon was already known being extremely efficient in terms of resource allocation within the processor since AMD can't afford to just dump tons of extra cache onto the chip.

Both of these things taken together means that using up extra real estate on the die of the Athlon in order to get SMT isn't really worth it in terms of the performance it would bring. Even on the Pentium the benefits aren't all that hot and it's only in specific types of code that you see any impresive speed gains.

Re:why go for CMP and skip SMT

[NerveGas]

You really can't figure out why they're focusing on CMP? It's not exactly tough. They don't have to design a new architecture. That saves a LOT of money in R&D, and (more importantly) cuts a LOT of time off of time-to-market. It's also VERY easy - especially with the Opterons. Copy the lithograph, connect the HT links, and you're done. To top it all off, it's something that will fetch a good price premium

To summarize, it's easy, fast, and will (supposedly) make them more money. That's a lot more attractive than sinking gobs of money and time into trying to design something new that may or may not pan out to be all that attractive.

steve

Re:why go for CMP and skip SMT

[Wesley Felter]

You have to be careful when reading academic papers; many of them talk about processors that could not be built. Super wide-issue, wide-SMT sounds great, but it's really expensive to implement and area-hungry (compare the size of an Alpha EV8 to an EV6).

Anyway, IBM and Sun have already combined CMP and SMT; the rest of the industry may not be far behind.

Re:why go for CMP and skip SMT

[philipgar]

Granted wide-issue super scalars are area-hungry and more expensive to implement, that doesn't explain the question I had asked. It stands by obvious reasoning that doubling the number of functional units in a processors (and adding 100 or so extra registers to the rename unit) is less expensive then doubling the entire processor die. What may get tricky is the VLSI design of the extra functional units (as going from 1 to 2 cores in VLSI layout should be straightforward enough in terms up chip area, but I'm no expert in this matter).

From this simple fact it seems intuitive that a super wide issue processor would take up maybe 10-15% more die area then a conventional superscalar. Then you add another 5-10% of the die area to the required duplicated structures and extra rename registers, and you end up with a super wide-issue SMT processor that takes the advantages of a superwide issue processor, and adds the latency hiding features of SMT to get a speedup of well over 2 times.

I think the most accurate reason may have more to do with the simplicity of going the CMP route.

Phil

Re:why go for CMP and skip SMT

[Anonymous Coward]

Intel's Itanium 3 (that name is of course not confirmed yet), codenamed "Montecito", will be the first CMP *and* SMT chip for the Itanium line. It's interesting that Intel are going to introduce both CMP and SMT in the very same product. The only other chip which is both CMP and SMT is IBM's POWER5 (not really shipping just yet, but soon)

By the way, the SMT in Montecito is not "fine-grain" SMT like Hyperthreading, but "coarse-grain" SMT as in IBM's RS64 series of chips (as opposed to POWER series).

Re:why go for CMP and skip SMT

[philipgar]

For the record there is no such thing as fine grained or coarse grained SMT. You can have fine grained or coarse grained multithreading, but simultaneous multithreading is a form of multithreading that implies that it can read opcodes from multiple instruction stream simultaneously (although not neccessarily all the instruction streams at the same time). If I remember the architecure correctly the P4 can fetch up to 3uops from the first thread, and up to 3 from the second. It will issue all the instructions from the first one, and however many it can of the second for no more then 3 issues per cycle (this is one of the limiting factors of Intels SMT architecture).

While I'm sure there are differences between the SMT appraoch in the varying chips, it is not fine grained vs coarse grained multithreading (These imply only one thread can execute at a given time and the two are almost identical with the exception that finegrained can thread switch once per cycle).

Philip Garcia

Re:why go for CMP and skip SMT

[Anonymous Coward]

Sorry, you're wrong. The terms are already in widespread use and mean:

Fine grained multithreading: multiple instruction streams can be executing at any point in time

Coarse grained multithreading: not more than one instruction stream can be executing at any point

It's not a matter of finegrained doing one switch per cycle - a finegrained SMT executes multiple, _independent_ instruction streams _at the same time_. The coarsegrained SMT _must_ switch instruction feeds.

The idea, of course, is that coarsegrained SMT switches on long-latency events such as cache misses: it's amusing to note that the RS64 series consider an L1 miss just such an event (though not so amusing when you realise the L2 is in discrete SRAMS off the CPU core..)

Re:why go for CMP and skip SMT

[Anonymous Coward]

Sorry, but it's you who're wrong. The definition of SMT in the original Eggers paper, and ever since, has been Simultaneous Multi-Threading, specifically referring to simultaneous execution of multiple instruction streams without any switching. If you're switching from an instruction stream to another, you are surely doing coarse-grained multi-threading, but it isn't SMT since the execution of different streams isn't simultaneous.

Re:why go for CMP and skip SMT

[Anonymous Coward]

Oops, I meant 'MT', not 'SMT'. I think we're in agreement here.

Re:why go for CMP and skip SMT

[David Leppik]

One reason to do dual-core rather than better hyperthreading is clock speed. The clock signal needs to propagate across the entire core to keep the circuitry synchronized.

There's a fundamental limitation (the speed of light) and a more practical limitation: the propagation speed of the electrical signal. To boost the propagation speed, you need to boost the voltage, which in turn heats up the chip. And you can't go faster than light.

Let's see if I remember my high school physics. At 3x10^8 m/s (the speed of light), a 1-cm radius circular chip is limited to 1/(2*3x10^6) second propagation (signal travelling from the center and back), or 6*10^6 herz. That's six gigaherz-- and that's assuming a straight line from clock to the farthest point on the chip! I'm not sure how big the chips really are, but we're already halfway to a fundamental physical limit. (A more realistic geometry is to assume the chip is any shape, but 2cm between the farthest points.)

With a multi-cored design, each core can have a separate clock, so the propagation speed is less. Assume we replace your 3Ghz processor with four Pentium II cores in the same space. (Assume it's 1cm diagonal for simplicity.) If you make the (unrealistic) assumption that die size is the only thing that affects clock speed, we can twice the clock speed in each core: 6Ghz, with a 0.5 cm core.

In short, smaller cores allows for faster processors. If they continue to make sprawling, single-core chips, they can't boost the clock speed significantly because the universe won't let them.

Re:why go for CMP and skip SMT

[Alexis de Torquemada]

Why do they go for CMP instead of SMT? Simple. They're trying to sell desktop processors. SMT has the huge drawback that you can typically only run two threads of the same process in parallel. However, on the desktop, many programs aren't multi-threaded at all, or if they are, then they (being designed for machines with only one path of execution at a time) they do not parallelize their workload efficiently between threads. For example, in a CD ripping application, one thread may be responsible for all the MP3 encoding workload, while the GUI thread would just sit there, idling away SMT CPU clocks.

On a server machine, SMT may often buy you big benefits in terms of performance, since many server applications are massively multi-threaded and designed for load balancing between execution units. Not so on a desktop machine, where you're rather interested in a system that remains responsive although you're running multiple CPU-intensive processes at the same time (e.g. word processor, video/audio encoder, some Java applications...). The benefits of SMT are quite limited there in comparison to "true" parallelism. And note that AMD processors don't suffer as badly from branch misprediction as the Pentium 4 does, which means that plain old OoO-execution still works fine for them wrt making efficient use of functional units.

As a side-note, the two technologies (CMP/SMT) are not mutually exclusive, as can be seen in the new IBM Power5, which sports 2 processor cores with 2-way SMT each.

Re:why go for CMP and skip SMT

[Anonymous Coward]

Intel stuck SMT into the Pentium in order to balance out the some of the negative effects the go hand-in-hand with a processor that has a LONG pipeline.

It could also be said they went with a deep pipeline to make SMT practical.

Re:Just get...

[zors]

Well, not to say that Macs are gaming machines, but Halo, UT2K4, and others.

Re:What about Apple?

[AKnightCowboy]

Which if these will be called the G6 is left up to the reader as an exercise. My money is on the 976. Either way the PPC has some serious legs.

Except it's legs are too expensive and too slow. Sorry folks, but x86 won the war. The ONLY reason I bought a Mac was to play with MacOS X. If it was ported to x86 tomorrow I would switch in an instant. PPC sucks donkey balls when it comes to speed. My 800MHz G3 iBook is horribly slow compared to my PIII-600 Intel laptop.

Re:Really nice alternative to dual processor syste

[MBCook]

You also get benefits in multitasking. Sure your PC might be able to burn a CD, rip a DVD, play some MP3s, and run a ton of web browser windows now, but with two processors, things really seem smoother.

You'd notice the most difference if you had one CPU bound app and a ton of others that weren't. For example you were running some big simulation or POVray, and at the same time checking your e-mail and surfing the web. With two processors even if the prorams don't use them (they aren't SMP aware), as long as the OS is (Linux and Windows NT/2k/XP for example) things will be smoother because one CPU can do the heavy lifting, and the other can juggle the little tasks so you're not stuck waiting 100ms here and there for your interactive task to get CPU time.

It sounds a little odd and I'm sure I haven't described it very well, but trust me, things feel smoother on my dual PIII 600 even when heavily loaded than my PIII 933 when it's only mildly loaded. If you already have a 3.4ghz processor, the effect probably won't be as pronounced.

PS: Quake III did support SMP, but as I remember it didn't take full advantage and it didn't provide a huge performance boost. Are there any (big) games that DO take full advantage of dual processors? With HyperThreading and such, I would think that would be more common now.

And with all of that processing power....

[ShatteredDream]

I will finally be able to run Linux in VMWare with a VMWare instance running Windows98 running Bochs running BeOS emulating OSX with PearPC. Thank you AMD, you have guaranteed me alpha male status in the CS department for a semester.

Re:What about Apple?

[Anonymous Coward]

The G5 is a somewhat cut down Power4 IBM chip.
Has alitvec added, 2/3 of the cache and ...tadaaaa... one of it's cores removed.

I see no reason to worry about Apple and the G5. All their G5's are currently dual processor. Look for a real Power4 machine soon I think. :-D

Re:Just get...

[dennism]

Well, first off, I'm pretty sure that the G5 could be cooled via only convential fans similar to the P4 and Athlons. But, Apple has pretty much made it their mission to reduce fan noise on their machines.

Second -- actually, we don't know that we'll be able to swap out single core Opterons with dual core Opterons. They're not out yet. The G5 is. If later on it proves to be true, then you can say that you can swap them out.

Third -- the G5 gives you access to one of the better Operating Systems around, MacOS X. That has to give it a few advantage points.

BTW -- I happen to have both a Dell Dimension 8600 and a dual 1.8ghz G5 in my office at work. When the Dell is running, you notice it. It's quieter than the thrown together PC that's also in the office, but still loud enough to notice. On the other hand, the G5 is completely quiet. I never hear the fans in there at all. I can actually see one of the fans moving from the front, but it's moving at such a slow speed that you can't hear it at all. For some of us, that is a feature.

Re:Just get...

[Exitthree]

Say what you want about the merits of building your own box, but don't call the G5 noisy. It has multiple low-speed fans to keep it quiet. It has separate thermal zones with independent cooling systems to minimize noise. I have heard, or rather been near enough a G5 to know it is not a loud computer.

Re:Really nice alternative to dual processor syste

[Anonymous Coward]

Sim city 4 ?

Re:Just get...

[Tsiangkun]

Genome assembly, annotation pipelines,comparitive genomics, live video aquisition and encoding, image deconvolution.

Re:Just get...

[JWSmythe]

I agree.

I got to drive one of the nice newer Mercedes coupes,with a big V8 in it. They were bragging up the horsepower, so I was wanted proof. "Let me drive." I ran it hard. The owner, in the passenger seat, was impressed with the power I was pulling from it. Then asked the owner how much the car cost. Something around $100k. I handed him the keys to my car (2000 TransAm WS/6) and said "now drive this."

I paid about $25k for my car. New it was something like $30k. My car has better handling, better acceleration, better braking, and is faster. This was before I did any mods to it. The interior trim may not be as nice, but my car does have all the options including leather seats, and it turns more heads when I drive past, than a Mercedes does. It's comfortable enough for two people to ride in it all day (done that many times), and the back seats are just about as big.

Apple's are very pretty. I've used a few. I was happy that my girlfriend was on one using OS/X, but when that machine started acting flaky, we didn't buy a new Apple, we spent $1500 on really good parts. AMD 2800+, 1Gb RAM, 200Gb hdd, DVD reader, DVD writer, asus motherboard, high end video card, etc, etc.. What Apple does $1500 buy you? When we want faster, all we have to do is buy some faster components. When the G6, G7, or whatever comes out, well, you're buying a new Apple.

You can buy a new Mercedes at the really fancy store, or you can (could) buy a TransAm at any dealership. If I want more power, I grab Jegs or Summit, and start shopping.

You can buy an Apple at the fancy Apple store, or buy parts from a wholesaler whos "Will Call" area is the back door of the warehouse.

I still say "Pretty" every time I look at a Apple. I give them that. Then I hop back on my x86 based Linux machine and drive faster. :)

Re:Really nice alternative to dual processor syste

[MBCook]

I don't know. But I can tell you I've played SimCity 4 and while it's nice, it certainly COULD use SMP support, because it is a CPU hog like nothing I've ever seen. The trend between SC2k, SC3k, and SC4 seems eponential. At this rate, SC5 will require a bunch of Crays ;).

Will it at least inlcude a cupon to buy a Cray?

He he he.

Dual core...

[Uplore]

The architecture as I understand it also creates the ability to moderate CPU temperature by switching between cores as the temperature rises too much. So that both cores can be running flat out if you have great heatsink, but if the levels get to hot through insufficient heat dissapation or heavy CPU usage then it is possible to switch a core 'off'. Of course all this is controlled by the MB and CPU, leaving no opportunity for errors by the users.

Why not 8 x i486 cores?

[Nom du Keyboard]

Why not take an older processor (e.g. i80486) that already is basically single cycle execution -- or Pentium which has two execution pipes already -- update it to modern geometry which should increase speed and decrease power, and put as many as you can easily fit onto the die? After all, those older cores execute all the basic i86 code including MMX with a lot less transistors. How much does SSE, SSE2 and HT contribute verses a lot of cores just executing threads with little context switching?

Re:Why not 8 x i486 cores?

[Anonymous Coward]

Thought at first "cool idea" then realised that that means the gains made post 486 would be lost. pipeline gains, hardwired instructions etc would all be lost and would kill the 32 -> 64 bit change. :/

Re:Why not 8 x i486 cores?

[Bryan Ischo]

Because for programs which only have a single thread of execution (I would say the majority of programs fall into this category), your program is going to run on only one core at a time and at the speed of a 486 (or Pentium, depending upon which one you used).

You could theoretically run N instances of single-threaded programs (where N is the number of processors in your core) and all would run without speed penalty. Or you could run N threads from a single multi-threaded program, and you could get close to a speedup factor of N.

In other words, truly multithreaded programs with lots of concurrency might get somewhat close to N times the performance of a 486, but most programs would perform about as well as they would on a 486. It's just that you could run more of them at the same time without speed penalty.

I would expect that a processor with 32 Intel 486 cores would feel most of the time like a fast 486, until you ran that one perfectly parallelizable program, at which point it would finally feel very fast.

Also, having 30 or so of your 486 cores sitting idle most of the time would result in alot of wasted electricity.

Re:Why not 8 x i486 cores?

[Nom du Keyboard]

your program is going to run on only one core at a time and at the speed of a 486

Did you miss the part about shrinking it down to modern geomerty, meaning it would run faster on less power (read less heat) than the original? Sure a 90nm i486 isn't going to run at 3.6GHz like a P4, however I expect it would run a good amount faster than a 486DX2-66 once did.

Also, having 30 or so of your 486 cores sitting idle most of the time would result in alot of wasted electricity.

Modern power control circuitry shuts down cores not in use.

Re:Why not 8 x i486 cores?

[minator]

This is almost exactly what Sun are doing in their next generation CPUs. Taking multiple simple cores and stuffing 8 of them onto a single die. I believe they are based on UltraSparc II based cores with support for 4 threads per CPU added. On specific types of tasks the Sun CPUs are going completely toast any x86 system.

Re:Why not 8 x i486 cores?

[ArbitraryConstant]

That's better for stuff with lots of branches, and Sun is doing it. They have 4 cores, each with 4 threads. Very rudimentry pipelining, and not superscalar at all. Whenever a core waits for something, it flips to another thread which is ready to go. There was an article on Ars about it. Basically, they can get about 60% more instructions per clock cycle out of these things on workloads that have lots of branches, and they've got 4 cores. Shit for gaming, very very good for dynamic web content and databases and stuff.

The other side of the coin is high clock speed. That's why P4's are so good at benchmarks that are basically a little loop that just crunches through numbers. Athlons are somewhere in between.

Right now, the best way to do that number crunching has been one big core rather than a lot of little ones. But there's a limit to how many parallel instructions you can extract from a serial stream of instructions. So it's reaching the point where a different solution is the cheapest way to get faster.

Re:Why not 8 x i486 cores?

[barawn]

Did you miss the part about shrinking it down to modern geomerty, meaning it would run faster on less power (read less heat) than the original? Sure a 90nm i486 isn't going to run at 3.6GHz like a P4, however I expect it would run a good amount faster than a 486DX2-66 once did.

Unfortunately, nothing will beat the architectural gains which have advanced since the 486 era, and the "worst case" pipeline waits will keep your clockspeed at an insanely low level.

Let me try to explain. The 486 had a 5 stage pipeline - fetch, decode, dispatch, execute, and writeback. Now, each of those pipeline stages isn't going to take the same "minimum" amount of time - some of them are fixed by things other than switching latencies. So, say your execute stage is fine taking only 1 clock cycle up to, say, 2 GHz (a minimum latency of 500 ps), but your decode stage, simply from physical concerns, is going to take at least 5 ns to complete. This means that the maximum you can ramp the clock speed up to is 200 MHz, because each stage in the pipeline has to take 1 clock cycle, so if 5 ns is your minimum, you'll have a max clock speed of 200 MHz.

The solution, though, is obvious - break that "5 ns" decode step into multiple pipeline steps - say, 5 of them, each taking 1 ns each. Now your maximum clock frequency is 1 GHz. The problem is that your pipeline is now 9 stages long, and you have a new architecture - which is precisely what Intel did several times over to allow the clock speed to ramp.

And that's just the pipelining limitation. There are other architectural problems with "ancient cores" as well. One basic problem is that the x87 floating-point architecture is crap. It's stack-based, which means you can only do math with the "stack head". So in order to store things in the registers, you need to use the FXCH instruction to switch the stack head and one of the registers. Well, modern CPUs (the P3 and the Athlon) got around this by saying "we'll make FXCH be a zero-cycle execute when paired with an arithmetic instruction (and after the Pentium, screw it, they're free totally)". Since the modern CPUs can decode more than one instruction per cycle (3 for an Athlon), and the FXCH instruction only lives up to the decode stage, you're really not hurt, as the FXCH fills a pipeline stage that probably would've been left empty anyway. Now consider the P4, which was designed to try to encourage people to move away from x87: it does not have a zero-cycle FXCH, and its x87 performance is abysmal. (The 486 does not have a pipelined FPU, nor a free FXCH instruction. It would be even worse.)

And I haven't even mentioned register renaming yet, which works around the register limitations of the x86 ISA by creating registers that the software doesn't know about, but which the hardware can "cheat" and recognize certain compiler patterns which work around the register limitation.

In short - many core 486 CPUs would suck. Even many core Pentiums would suck. Architecturally, they're old, dead ends. The best designs for multicore processors would be the P6 design (PPro/PII/PIII/PM) and the Athlon design (K7/K8 - while the K8 is "new", it's about as new as the PM is to the P6 design). Curiously enough, Intel is likely to go with a multicore PM, and AMD is likely to go with a multicore K8.

It should also be noted that a 486DX had a transistor count of 1.2M transistors. A P3 had a transistor count of 9.5M transistors. That's an increase of about 8X - however, the P3 also has twice the data width (64-bit rather than 32-bit), 4X the L1 cache (32KB rather than 8KB), and had two instruction set enhancements tacked onto it, as well as massive architectural improvements, including, essentially, multiple versions of the 486 execute engines inside it. An 8X increase in size for those enhancements is not crazy at all.

Re:Why not 8 x i486 cores?

[Hoser McMoose]

FWIW if you're interested in how a very simple cores compares to today's copmlex P4 and Athlon64 chips, look for some comparisons with the VIA C3 chips. The C3 is essentially what you describe, a core very similar to the old Pentium but updated to the latest and greatest manufacturing process, a new bus (the old Pentium bus with it's off-chip cache just stank for scaling) and a few other tweaks and enhancements.

The size, cost and power consumption are excellent for this chip, but it's performance is rather low. However you can pack about 4 of these chips in the same die space as a single P4 or Athlon64 and you can get about 8 of these chips for the power budget of AMD and Intel's latest and greatest. For highly multithreaded applications this could be a better solution.

The problem is that many applications are not highly multithreaded. It's always a trade-off between good multi-threaded performance and good single-threaded performance. For the time being going beyond 2 cores probably doesn't make much sense for desktop and workstation users, there just aren't enough apps that can effectively make use of more than 2 threads to make up for the loss in single threaded performance. For servers, that's another story. What you're describing is pretty much the exact strategy that Sun is taking with their "Niagara" chip design, 8 simple SPARC cores on a single chip. Low single threaded performance but should do VERY well for highly multithreaded applications that are common in the server world. For it's intended use the performance/die size and performance/watt of power should be excellent for that chip.

No new sockets!

[WeekendKruzr]

Hector Ruiz already stated plainly in an interview that the dual-core Opterons will be socket compatible with the the current Socket 940. It should be noted that Socket 940 is for Opteron only (servers) not the Athlon64/FX and according to Hector he highly doubts that they will produce a dual-core chip for the desktop/mobil market. That's pretty understandable considering that 90% of all desktops/laptops go under-utilized nowadays anyway. Socket 939 is for Athlon64 desktops and the newly redesigned AthlonFX.

Re:What AMD is really doing - Sink

[Nom du Keyboard]

They're making the first Desktop Fusion Unit!

Can't wait to see the heat sink that goes with it!

The main issue will be memory!

[NerveGas]

The principle reason why the Opterons are kicking the trash out of Intel's Xeon line is the memory bandwidth - with each processor you stick in the board, you're getting another 128-bit memory controller. With a well-designed OS, that equates to enormous potential memory bandwidths.

I've been a little leary of the dual-Opteron stories. Yes, it'll let you pack more CPU's into your board: But will you be able to utilize the additional memory controller in that extra core? If so, it will rock trash. If not, then it's losing the key strength of the Opteron.

Now, it *could* be done. The existing Opteron pinouts provide for four DIMM sockets. With dual-core chips, that same pinout could provide *two* DIMM sockets for each of the two cores. If that's how they go, then AMD is going to continue their incredibly strong showing in the server arena.

(FWIW, I just bought a 4x848 Opteron system, and it is a *screamer* at database work!)

steve

Re:The main issue will be memory!

[Nom du Keyboard]

equates to enormous potential memory bandwidths.

No it doesn't.

Unless the new controller goes to an entirely separate memory bank, then it is simply contesting with the first controller to access memory. Regular ram chips only serve up one memory location at a time, and are more efficient when serving up a burst of consecutive locations, delaying the second controller even further. Since even a single processor is often waiting on main memory (figure 90% cache hit rate, and there is still a lot of memory reading and writing happening), a second controller can't even get to the memory.

And if you go to separate memory banks for each core, costs go up and getting the right data in the right memory bank becomes a problem. Yes, you ought to be able to cross-bar the memory accesses in the chip between multiple controllers, but the system costs go up beyond what many people would want for a desktop system.

Re:The main issue will be memory!

[NerveGas]

No it doesn't

Yes, it does.

If you're at all familiar with the Opteron architecture, you'd realize that each chip's memory controller does, indeed go to a new memory bank.

As an example, I just bought a 4-way Opteron. It's got four seperate banks of memory on it. Each processer has a 128-bit, DDR400 memory controller, all independent of each other.

If you have a program on each CPU, accessing memory tied to that CPU, the 4-way machine I mentioned would have a theoretical memory throughput of 25.6 gigabytes/second. The theoretical throughput of a dual-Xeon machine is 5.4 gigabytes/second. That's a huge difference.

You're right, it takes some intelligent work to schedule programs on CPUs that are close to the memory the program will access. If you hadn't been in a hole for the past year or two, you'd know that there has been a lot of work put into Linux to make it handle these NUMA architectures more intelligently. IBM has some VERY large NUMA systems, and has been pouring a lot of development into Linux.

As for system costs going up so much that it would be prohibitive for a desktop, think again. AMD's entire desktop line is transistioning to the Opteron architecture. Even the lowly 1xx single-proc Opterons and Athlon64's have nearly all of the features of the highest 8xx 8-way chips. The difference between a 848 and a 148 is just reduced cache, and fewer Hypertransport lines out of the chip.

steve

Re:The main issue will be memory!

[Nom du Keyboard]

If you're at all familiar with the Opteron architecture, you'd realize that each chip's memory controller does, indeed go to a new memory bank.

Yes. And in the 939 pin package, a single processor accesses dual memory banks in parallel.

However...

Most desktop units are not going to put in 4 or 8 independent banks of memory. For servers, yeah, spread 2 or 4 GB over multiple banks. But a desktop PC, I don't think so. The gain is not worth the price. You don't need that much memory, and buying less by spending on lower density memory modules is not cost effective.

So where 95% of us care about it -- i.e. on the desktop -- we don't see this gain because the price for it is too high.

And that is my point. Not that you can't build some screaming fast machine with outrageous memory bandwidth using a multi-core Opteron chip going out to a lot of independent memory banks, but rather for most of us users it isn't going to happen for cost reasons, and therefore it is nothing to get excited about. I save my excitement for the stuff most of us will actually own -- which may likely be a dual core Athlon-64 with dual memory channels running out of a Socket 939 in about a year.

Re:The main issue will be memory!

[Wesley Felter]

The existing Opteron pinouts provide for four DIMM sockets.

No, the existing Opterons provide for two memory buses. A bus can have a variable number of sockets, but the wires are shared; only one DIMM on each bus can be accessed at one time.

Re:The main issue will be memory!

[NerveGas]

Right now, there are dually boards intended for workstation/desktop use where each of the two processers has 4 DIMM sockets associated with it. If these dual-core chips were done ideally, then each chip would have two cores, each having two DIMM sockets. The real estate usage is the same. The total maximum memory is the same - but you'd have twice the potential bandwidth.

On single-socket boards, you'd still have the same 4 DIMM slots you have now - 2 DIMM slots per core. It wouldn't cost any more, and it wouldn't use any more space. Now maybe there's some reason on the pinout that it can't be done, and that would be sad. But if the pinout allowed it, the motherboard cost wouldn't be any greater.

You don't need that much memory

It's not about the amount of memory, it's about the amount of memory bandwidth. Do you need 26+ gigabytes/second of bandwidth for the desktop? No, you don't need it. But if it were possible with little to no increase in cost, then why not?

It wouldn't even force people to use more DIMMs than they use now. Shoot, you could run an 8-way Opteron on two DIMMs if you wanted. It would be kind of a waste, but it could be done. : )

steve

Re:The main issue will be memory!

[Nom du Keyboard]

On single-socket boards, you'd still have the same 4 DIMM slots you have now - 2 DIMM slots per core.

Which means that you would have to populate all 4 DIMM slots to run the thing. As opposed to the 1 or 2 DIMM slots you populate now at minimum on single or dual channel memory systems.

One gig in 1 DIMM, or 1 gig as 256MBx4 DIMMS. Which do you feel will cost less? And to upgrade memory you would have to replace it all.

This was never about the cost or space needed for the sockets on the motherboard. It's about what you have to put in them with multi-channel memory access.

Re:The main issue will be memory!

[sexylicious]

But why would you go out and buy a fancy-pants Opteron and not get it to its full potential?

That's like buying a Ferarri and putting a three-speed automatic transmission in the thing. It works, but you're not really ever using the full potential of your engine. Unless of course, you have a funky three speed automatic that can handle the horsepower, torque, and still work over the published speed range of your Ferarri.

It's also like putting together a parallel cluster of superfast machines, but you only use 10 Mbps ethernet to connect the machines. Why do it that way?

Re:The main issue will be memory!

[Nom du Keyboard]

You're not gong to run an OS on a Cell, it's not designed for it.

1: Cost.
2: Cost.
3: Cost.

The very same reason that super fast gaming machines sell to such a limited market. There's always a cost verses performance tradeoff. Most users probably trade at a different point than you apparently do.

Re:The main issue will be memory!

[Rich0]

Well, you would only need to populate as many banks as you have processors.

So if you only bought one CPU, you'd only fill two of the slots. If you upgrade to two CPUs then you'd upgrade the RAM as well.

If cost is the main issue then you probably aren't thinking about dual-CPUs anyway.

I'm running a 2GHz AMD64 at home with the neutered cache for just this reason. I'm sure double the cache would make it faster, but all the reviews suggested it wasn't worth the cost. I looked at cost vs performance and picked the point on the curve that made the most sense to me.

People considering multiple cores are also likely to want lots of RAM. After all, if you're at the point where you want 2 CPUs, would you want only 1GB of RAM?

Re:The main issue will be memory!

[Nom du Keyboard]

Well, you would only need to populate as many banks as you have processors.
So if you only bought one CPU, you'd only fill two of the slots.

Go back to the title of this entire discussion. We're discussing dual core processors, and single socket motherboards intended for dual core processors that might contain upto 4 memory banks to make optimal -- and expensive -- highbandwidth memory access available. Any single CPU discussion doesn't fit the parameters of what I'm talking about from the beginning.

SMP and Windows stability?

[HuguesT]

This raises questions regarding stability and Windows.

While I find that multiprocs settings under Linux improve things to a significant degree (although there are still outstanding issuess with NVidia proprietary drivers and SMP), I found the opposite true for Windows.

The last time I tried, which was about 2-3 years ago, many drivers didn't seem to expect true concurency under Win2k and I was experiencing significantly more crashes on my dual P-III than when I forced the system to only use one of the CPUs. Yet it probably wasn't the hardware because that same machine was very stable with Linux.

With the advence of hyper-threading, have things improved markedly with WinXP?

Re:SMP and Windows stability?

[DrHex]

So in essence we end up with systems that crash harder and faster, much like Steve Austin did in that 70s Series, The Six Million Dollar Man. How about coders get their overall quality of code production up before we hand them MORE computing horsepower to waste and continue to perpetuate the insane computational power increase while all the while not really addressing the CORE issues around a solid stable code base. Once again, Linux and the BSDs will leap ahead of Windows thanks to their more stable code base and we'll see the distance stretched between Mainstream and the techies.

Overall, this will just mean users will see the blue screen and crashes all that much faster.

As a friend of mine said a few nights ago when he fired up his new PC with WindowsXP. Why isn't this working better, they've had 3 years to fix this??? Why fix it when people keep paying for your crap?

Just a thought and maybe slightly off-topic.

Re:Just get...

[Anonymous Coward]

Apple is sort of like the Mercedes Benz of computers, they look nice...

Haven't seen the G5s, have you?

Re:Just get...

[Omega1045]

I am sorry. You need to prove to me how anything Apple is cheaper. To me (ignorant fool) Apple is for the rich liberals...

Re:This really sucks. - and so does your math!

[Nom du Keyboard]

Now I'll have to pay SCO $1,149 instead of $699.

Don't you mean $1,398? Or are you doing your math on an old Pentium -- in which case your really do need an upgrade.

HaHaHa!

[Anonymous Coward]

You're planning on waiting more than a full year between computer upgrades?
Are you sure you're on the right website?

You're surprised that a bunch of slashnerds who run Linux on their PII 400 with 128 MB of RAM don't upgrade every six months? I'm surprised that you're surprised!

Re:This really sucks. - and so does your research.

[HotNeedleOfInquiry]

SCO charges $1,149 for a dual license. Check their website, Darl.

Re:Really nice alternative to dual processor syste

[Jeremi]

Dual cores processors seem to me like a pretty good alternative to a dual processor system.

Sure, but given the option of dual-core processors, I want a system with two of them... I know, it's a fetish, what can I say?

Exactly my first reaction!

[PaulBu]

But then, the trick is that he did not mention memory latency, only bandwidth! Getting the latter is relatively easy -- just make memory bus wider (as given bus speed), trying to decrease latency will pretty soon make you run into speed-of-light limitation.

Maybe those processors do have enough memory bandwidth to load two of them completely doing SAXPY? Assuming 12 GFLOPS sustained (3 GHz, 2 cores, separate ADD and MUL on each) you need to feed input vectors at 12*8 bytes/double = 96 GB/sec, for, say 1 GHz memory bus it is translated into 96*8=768 memory pins only for input -- well, wider than I've seen on desktop PCs... ;-)

When you start doing anything else , the roundtrip time between processors and memory (latency) becomes more important than raw bandwidth.

Paul B.

Re:Exactly my first reaction!

[kryptkpr]

The Hypertransport serial bus runs at 1.6Ghz.

So signals take 6.25e-7 seconds to travel across the link.

The speed of light is 299792458 m/s (according to my calculator), which gives a theoretical hard maximum distance of 187.37m between the two endpoints of the HT link..

In practice, we won't even need 1/20th of that, and the speed of light should not be a problem.

Disclamier: I'm tired, and still an engineering geek in training. Please correct my math or my logic if I'm wrong.

Re:Exactly my first reaction!

[witichis]

Since there is about a magnitude difference between CPU and memory speed there is only one thing making sense to me - bringing them closer in geometry and clockspeed. see this approach with nice piccies

Re:Exactly my first reaction!

[ChrisMaple]

Your math is wrong by three orders of magnitude. 1/1.6 GHz is 6.25e-10 seconds, round trip. 3.125e-10 seconds each way. About three inches. But there's a lot more to take into consideration: speed of light on a PC board is lower, there are protocol and chip hardware delays, etc..

Re:Exactly my first reaction!

[kryptkpr]

Bah, hit the M key instead of the G key. That is indeed not very far.

HT can run at 800 Mhz as well though, possibly giving a further reaching but slower connection.

Re:Exactly my first reaction!

[PaulBu]

Yeah, also called PIM, for "processing in memory". Peter Kogge (of Kogge-Stone fast adder fame) is doing research into those in Notre Dame (disclaimer: I used to be affiliated with the same HTMT crowd designing ultra-fast superconductor processing elements for the petaflop computer, thus I know first hand how hard it is to match memory speed to pipelined processing speed... ;-) ).

Bringing processors and (large) memories closer to each other does not help much, as, as you mentioned, there is an order of magnitude difference between processor clock speed and memory access speed. The physical reason for this is that to do a certain operation on one pipeline stage in a processor you need to charge a clock line passing through a couple dozen to couple hundred gates; in memory case you have to charge the word line passing through sqrt(1G)=30,000 gates. It takes time (RC, unless one uses superconductors and forgets about R ;-) ) and power (CV^2/2).

The only rerasonable solution is, indeed, to make memory blocks smaller and closer to processor elements, making them essentially registers/caches, not RAM.

Oh, and, BTW, in the rather naive picture on the link you sent, the solution will not work that well if you have multiple processors -- you have to make sure that each can talk to other's memories (in SMP case) AND to each other.

Paul B.

Re:Exactly my first reaction!

[addaon]

Also, you don't need to round trip one signal at a time, you can have multiple signals in flight. Which makes this whole discussion rather moot.

Re:Exactly my first reaction!

[PaulBu]

Also, you don't need to round trip one signal at a time, you can have multiple signals in flight. Which makes this whole discussion rather moot.

Nope! Having several bits flying over the wire is a form of pipelineing (check-out "wave pipelining"), which increases bandwidth, but does not do anything to decrease latency. bringing us back to my original point, which can be illustrated by a fact that good weight-lifters rarely make good boxers... ;-)

Paul B.

Re:Exactly my first reaction!

[ckaminski]

And I wonder, is that simply because they don't spend enough time training to be boxers, or simply because their muscles are too slow and bulky? After years of martial arts training and weightlifting (not bulking, tho), I'd reason that it's the former...

Re:Exactly my first reaction!

[sexylicious]

The muscle fibers are actually different.

The shorter reaction times required of a boxer require shorter muscle fibers. Endurance-type training requires longer muscle fibers that are more efficient at repetitive tasks (spread the work over the entire fiber).

There is also the fact that you train your mind and body differently when boxing versus weight lifting. Weight lifting requires a one-time explosion. Boxing requires an explosion, then retraction as fast as possible. In boxing you imagine yourself landing a blow and following through, but rarely do you imagining yourself stopping after you've followed through. Same with most martial arts.

In weightlifting you tend to put all your visualization into getting the weight up just the once. This is the same for some of the punches and kicks in martial arts.

Oh and back to the muscle fiber thing...
Endurance athletes tend to have very long muscle fibers compared to sprinters and weightlifters. Endurance athletes' muscles are literally longer and less bulky than their less efficient counterparts.

Re:Really nice alternative to dual processor syste

[NerveGas]

Multithreaded and multi-process.

If Firefox is rendering a page, you've got Firefox doing the rendering, the GUI working with video drivers, disk drivers looking at/updating your browser's cache, kernel code managing disk cache, kernel code managing network activity, and perhaps even firewall code running.

Whether you use Linux or Windows, there are a LOT of things running that you don't see in normal process list.

Now, will dual CPU's speed up that render time in Firefox? Not to any significant amount. But having used a LOT of dual-CPU systems, I can say that under heavy load, the machine will be much more responsive. If that helps your workload, it might be worth it. If it doesn't, it's not worth it.

As an example, at work I have a dual AthlonMP 1800+. At home, I have a single AthlonXP 3200+. For what I do at work, the single-proc chip would suck rocks. For what I do at home, the 1800+ would not compare to the 3200+. It's all about your usage.

steve

Re:Just get...

[beakburke]

Umm let's see, I'm sure that the Unreal games are. Actually the game situation on the Mac is much better now than in years past. Most of the more popular games do work on the Mac. But not nearly as many as on windows, but that's to be expected I guess. It's just fine if you are only an occasional gamer and aren't super picky about your games.

Integrated memory controller

[charnov]

Because the K8 has the memory controller on die, as you add processors, you actually add memory bandwidth. It kinda stands the old logic on its head. Really the only thing that can be an issue on this core is latency can make a difference at 16 CPUs or more ;-)

Re:Integrated memory controller

[Ninja Programmer]

Hypertransport only supports up 16 nodes, and one of them has to be the southbridge. So you can't get to 16 processors anyways. :) Seriously though, I've only seen topologies of up to 8 processors at once. So quad-boards with two processors per core is probably about as high as they will go with this architecture.

Re:Integrated memory controller

[Anonymous Coward]

Because the K8 has the memory controller on die, as you add processors, you actually add memory bandwidth.

Huh? To talk to completely separate blocks of RAM, yeah, but then you're building a cluster not a multi-processor. And that's nothing new - it's called NUMA.

A second controller will only shuffle the bottleneck and locking issues around if you're trying to share RAM.

Re:Integrated memory controller

[fitten]

There are X number of data/address pins coming out of the chip. If they stay with the S940 or S939, that number of pins is exactly equal to the number of pins required for one CPU. Adding another core without adding another X pins means that the two cores will share the X pins to the memory. So, unless the memory controller currently has 2X the bandwidth required for a single Opteron to run, then you will have a bottleneck accessing that one bus off the chip in memory intensive applications as the bandwidth available off-chip will be shared by two CPUs instead of one.

Re:This really sucks. - my bad!

[Nom du Keyboard]

SCO charges $1,149 for a dual license. Check their website, Darl.

My bad! Who would have ever expected SCO to be that nice about anything?

Maybe we need to check their website more often. Yeah, all of Slashdot check their website every day to see if anything has changed. That would be good.

They already have dual channel memory cntrl

[charnov]

The current Opteron has dual channel controllers. There really isn't that much of a reason to go dual dual channel when in many situations, the single channel Athlon 64's outperform the Opterons because of reduced latency (no registered dimms).

Market segmentation

[charnov]

no that extra ping provides market segmentation...ha.

Actually, the pinouts are very different between S940 and S939. I think they originally were going to do all of them as 940's and realized the issues that would cause (the power leads are in different places, for instance).

Hold out another year...

[csoto]

By then, maybe IBM will have enough 3GHz G5s for your Mac!

Fully pin compatible -- get an opteron now

[wray]

I didn't see this anywhere else... sorry if a repost.

The dual-core opteron's will be fully pin compatible with the current ones. Have a dual-opteron? drop a couple new ones in, and its a quad.

Awesome.

Re:Fully pin compatible -- get an opteron now

[bersl2]

The question is, would anybody really buy two single -core Opterons after the dualies come out? I guess if it's dirt cheap...

Re:What about Apple?

[Anonymous Coward]

Ooops, all AMD and Intel cores have been RISC based for generations. Time to leave the 1980s methinks.

Re:Just get...

[Anonymous Coward]

MacOS X is one of the better operating systems? In what land is that the case? Certainly not in reality.

Pin compatibility implied

[ogre7299]

In a quote from AMD's CEO taken from http://www.theinquirer.net/?article=15605: "And he said that next year its dual chip Opteron-whatever will "shock the hell" out of everyone because it will be pin compatible, hardware compatible and otherwise compatible with existing motherboards."

So I'm reasonably sure that current Opteron/Socket 940 users will be able to use Dual-core chips as long as the bios supports it.

Advantages of multi-core

[chrysrobyn]

I see lots of conversation comparing this generation of processor to space heaters, wisecracks about Longhorn minimum systems (that actual article was about the predicted "average", not minimum). Not much about actual multi-cores. They're an interesting direction to go.

The current direction of single core CPUs is basically running into the most they can do with XUs, MPUs, caches, etc. Sure, you can decrease the pipeline depth below the 18FO4 that the PentiumIV supposedly has, and that can help you with serial data paths, and that makes simple XUs, MPUs, etc. faster, but the branch mispredict is still horrendous -- perhaps too high for a general purpose processor found in our PCs. The more complicated logic is possible to do, but there's only so much you can do with the data and sub-Angstrom logic.

Beyond the geek factor, multiple cores on a single die attack the same problems as putting SMP did in the first place (plus a few race conditions that otherwise may have been very rare), allowing much less manpower to design a processor that is still much faster in the end. A single threaded application will seem slower, and that will place more burden on the developers to see the light of multiple threads. Instead of allowing an XU to munge through and deal with a single thread at a time, which may be a misuse of incredible resource (like a thread that said "go to grocery store" and the XU was a race car), multiple die have correspondingly multiple XUs each with their own resources, so hard tasks can be spread across multiple cores, or simple ones can get executed in parallel with others (like a thread can take a Kia to the grocery store while another Kia goes to the Post Office). Of course, problems that cannot be divided into multiple threads do not see the advantage of multiple cores, but other tasks remain responsive without requiring a monster task to context switch.

I've read about multiple cores that share a single L2 outperforming multiple cores with dedicated L2s in specific tasks, basically one core essentially acts like a pre-fetch core under a workload and the second core can reap the benefits.

Re:Just get...

[Anonymous Coward]

three reasons?
emacs, emacs, and emacs, in that order.

Re:What about Apple?

[mattkime]

i doubt its horribly slow, and i can guess which one has better battery life.

heh ?

[kayen_telva]

Well it seems AMD has decided to follow suit.

It should be noted that the K8 architecture has had this designed in from the start.

Who exactly is following suit ?

ANYBODY READ THE 4 DAY OLD UPDATE?

[Anonymous Coward]

"UPDATE: A representative for Intel Corporation told X-bit labs the company had never released any precise details in regards the dual-core strategy. The information published herein should not be considered as based on official statements."

OR NOBODY KNOWS HOW TO READ ANYMORE!

Re:Just get...

[forkazoo]

Oh, and I suppose the reason I bought my current x86-64 box, or my SPARCs was because of all the optimised games available for the platform. Oh, wait! Shoot, I almost forgot, I got it to do real work with. 3D rendering, video encoding, programming... Games are just what I do while my boxen are all busy. Seriously, I've never understood people who will buy a high end system just for gaming! WTF?

Re:Really nice alternative to dual processor syste

[drinkypoo]

I think you're getting your terminology mixed up here. Even Windows 3.1 had multitasking, though it was cooperative. When you have multiple cores in use, you're not just multitasking, you're multiprocessing. That is to say, two processes can be active at once, because you have two separate contexts. Arguably this is true of the Hyperthreading (I refuse to call it HT, that's what I call HyperTransport) Pentium IV as well, since it also has two contexts.

Re:What about Apple?

[Anonymous Coward]

Your G3 iBook is probably slow because:

• Less L1 & L2 cache than the Pentium (this is the purpose of a PowerBook)
• You're probably not running with enough ram for OS X and your apps (512+ gives a good headroom)
• Did you buy the budget HDD for the iBook? This matters a lot.
• If the PIII is running Linux, obviously it will be faster.

You should actually try using a decent G5 for a couple days before you start slandering PPC.

Re:Just get...

[Anhaedra]

Umm... Let's see, how many of today's PC games are available on the Mac? That's right, zero. So what's the point of having a fast CPU? Video encoding? Photoshop?

Let me see...

Warcraft III
Halo
The Sims (all the expansions as well)
Starcraft
Escape from Monkey Island
Pool of Radiance
Fallout
Deus Ex
Max Payne
Medal of Honor Allied Assault
Prince of Persia 2
Quake III
Return to Castle Wolfenstein
Sin
Unreal Tournament
Unreal Tournament 2003
Escape Velocity: Nova
Baldur's Gate II
Icewind Dale

That is all I can think of.

Re:What AMD is really doing - Sink

[Hoser McMoose]

Can't wait to see the heat sink that goes with it!

Sure it can't be any worse than this Prescott heatsink/jet engine!

Baaahh!

[danalien]

why don't they just manufacture them with 'pins to spare'? ... like, for once, think 3 or 4 steps ahead of the curve.

/* what I mean is, make a socket, with like 4000 pins (I'm just `cat /dev/urandom`'ing here :)), and use it for the next 5-10 years or so .... then one would _truely_ have a 'backward compatible' solution .... */

PS. Baah, what am I rambling about this to 'greedy capitalist', they freakenzie want to squeeze as much 'juce' as possible (and 'updates/upgrades' is their 'golden goose' ...)

Re:Just get...

[ArbitraryConstant]

My officemate has a dual Xeon 2 ghz (ENVY ENVY ENVY ENVY) from Dell, and it's quiet as a mouse. I'm not saying G5's are loud, but there exist quiet PCs.

As for MacOS... I promise you, 99% of the dual core Athlons will go into gaming machines, and while there are a decent number of games for MacOS, the Windows version is either released at the same time or first, and there's always a Windows version. Sometimes there's no MacOS port. That makes the competition Intel and no one else.

Don't get me wrong, for desktop stuff MacOS is the only OS I can stand. But sometimes it's just not acceptable, either due to the requirement of Windows or the requirement of something else (usually Linux or Solaris).

As for memory bandwidth, most of these will likely be single core systems. While there are dual G5s out there (indeed, the current ones are all duals right now), the Athlons will have the same memory bandwidth because the G5s aren't NUMA. Maybe a bit more due to the on die memory controller*.

* Yes, I know how fast the frontside bus is, and I know every CPU has its own bus. But there's one memory bus and that means 400 mhz * 128 bits is the fastest it can access memory.

Re:What about Apple?

[HiredMan]

A tip of the hat, Sir. A most excellent troll. I was almost tempted into it until I read this part:

My 800MHz G3 iBook is horribly slow compared to my PIII-600 Intel laptop.

I realized that this argument is too stupid (even for /. holy wars) to ever be serious...
"My 91 Cressida so pwnz your Sux0r 92 240SX."

But all in all a very nice effort. I applaud the effort to look like a complete moron in an effort to troll.

Nice commitment,

=tkk

Re:AMD! STOP CHANGING SOCKETS!

[Hoser McMoose]

Err, these dual-core Opterons are going to be socket 940 compatible, just like existing Opterons.

This simple fact kinda of makes your rant completely pointless.

Re:Just get...

[mccoma]

Well, I'm a middle income non-liberal and I have an Apple. The new dual G5 systems are priced damn good when compared to Itanium-2 systems (still running at 1.5 ghz) and in the ballpark of the dual 64-bit AMD offerings (2.6ghz? available).

The big savings is in having a decent UI and time not spent messing with the computer to get it to work right. I value my time and my hearing (the Mac is pretty silent).

I guess Macs aren't for everyone. If you must build your own, then don't get a Mac. But, don't compare a Mac against a home built (not saying you are, but it seems to be a trend).

Re:Just get...

[evilviper]

It has separate thermal zones with independent cooling systems to minimize noise.

I have a question. Why is it that the "seperate thermal zones" is a big buzzword whenever talking about G5s? It's not a new idea, it's not revolutionary, it's not that significant of an improvement, etc. So, I'm wondering why that simple point always gets thrown out there, even though it's rather irrelivant...

Comments requested.

It's a bit of both...

[Kjella]

...the Intels have been putting out more total power. The AMDs have been putting out more power/die space. One leads to bigger case fans (remove heat from case), one to bigger CPU fans (remove heat from CPU). Overall, neither is quiet at least.

Kjella

Re:Why not 8 x i486 cores? Toast, well done please

[Nom du Keyboard]

UltraSparc II based cores with support for 4 threads per CPU added. On specific types of tasks the Sun CPUs are going completely toast any x86 system.

What I actually expect to toast everything around in a couple years, and at a surprisingly affordable price, will be the Sony/IBM Cell Processor.

Re:Just get...

[Anonymous Coward]

I have to say if its just the car with raw power, a new Ford Mustang Cobra can get to 60 in just over 4 seconds, but it, like your car, has one major draw back, its a lot less safe in a wreck, even at the lower speeds. People might contend with me about this, but there is a reason why the benz coupe is so damned heavy and why these cheaper cars can go so much faster with about the same engine. That extra steel does help.

Granted, having said this, I don`t care about safety all that much and this is why the next car I get will be that Cobra with that damnable governor chip removed(its jsut so annoying to be told 130 is the fastest, when the drive train and engine can get much closer to 190).

yes yes, completely off topic, but really , everything that can be said has been said about dual processors, at least everything I can say. anyways, the cobra is only 35k base, which for the speed increase, isn`t that much. For anyone who wonders, its faster than any benz on the regular market.

-Gordo3000

Just on time!

[Alexis de Torquemada]

We recently learned of Intel's plans to go dual-core in late 2005. Well it seems AMD has decided to follow suit.

Great! That will be just on time for Doom 3, Half-Life 2 and Duke Nukem Forever.

Re:What about Apple?

[Anonymous Coward]

The point is that AMD and Intel need to carry on a huge decoder circuit and handle lots of baroque complexities in hardware, that the PPC engineers never need to even think about. That's why it is cheaper to develop a true RISC design than x86 processors. Of course x86 makes so much more money that they can easily afford to hire an entire division of engineers to develop a top-performance decoder circuit, but the point is that the PPC can compete with less income and less resources expended.

More Cores

[thebdj]

Let me explain something, though it may have already been done for many of you. You joke about more cores, but both groups are surely already in the process of adding more cores to their architectures. Granted I heard my news through a third party but apparently they know a person at Intel who said there was development of upwards of 16 cores on a single chip.
The reason this works out as more is better is simply because we can. Think about how small the processes have gotten. Most will be over to .09 soon and there is technology to get that down even smaller. Before the limitations on the expansion of the speed of a chip were often affected by Cache size. Look at the crazy performance given by doubling cache sizes on a CPU. The problem is Cache is expensive to place on a chip, cores are not. Expect the new war in the CPU world to be more along the lines of more cores and not so much on clock speeds. This is part of the reason the companies are trying to break the traditional numbering schemes for processors and inventing convoluted messes of numbers that literally mean nothing.
My only concern so far has been on the usefulness of dual cores. I am sure they have made some sort of hardware method to allow current software to continue treating the chip as a single CPU, because otherwise it would be pretty useless to have what amounts to really having twice the CPU on the same chip space since most software isn't multithreaded to handle multiple chips. But I am sure they have taken care of this. Better stop before I look like I am rambling....

Re:Why not 8 x i486 cores? Toast, well done please

[minator]

Yes and No, yes the Cell processors are going to be shockingly fast (especially when set up for stream processing) but they are vector cores (8 per Cell), not general purpose CPUs. You're not gong to run an OS on a Cell, it's not designed for it.

Re:AMD! STOP CHANGING SOCKETS!

[Sloppy]

Get the new Osbourne64 when it comes out. They say it will be compatible with all future processors.

Re:Really nice alternative to dual processor syste

[snero3]

Some years ago I was thinking about getting a dual processor system. Alone the motherboard was two times as expensive as a similar single processor one

OK dual mobo's are more expensive but they are normally better built plus once you have use a dual machine as a desktop you won't go back, it just so much snappier.

applications did not support it all and so on. I hope newer applications are ready for dual cores. Quake III was the first game I know that used two processors and finally I can consider that animated desktop background.

When you buy and multi CPU machine for home/desktop use you have to think multiprocessing your work. Ok so some apps don't use the benefit of SMP natively, why do run two CPU heavy apps at once? IE play games + play mp3/ogg files without shutter? (ok not a good example but you get the idea).

Is there a list which applications can effectively use dual cores besides obvious things like webservers?

Anything is that is multithreaded or multi processed (databases etc) will take advantage of a SMP environment

Red Storm

[charnov]

Look up Red Storm and the other massive systems. You can use bridge chips and hyptertransport switches to bridge 8 - way nodes. You can go 4 way glueless, 8 way with glue, and up to a number limited by interconnect and latency (who knows how high that is...more than the 10,000 node systems proposed by Cray...and they should know).

Re:Really nice alternative to dual processor syste

[MBCook]

No, I'm aware. I was just saying that multitasking seems much much smoother when you have two processors than when you have one.

Re:Just get...

[Anonymous Coward]

I don't want to jump to conclusions here...and I can partially understand not wanting a speed governor as a matter of principal, but if you are driving over 130 on the highway...I look forward to reading about you in the Darwin awards. Keep it on the racetrack folks. Leave the stupid behavior to the Honda kiddies. If you aren't "that guy", cool...rock on. :)

As for the car being 35k...for that price difference, you can tack a supercharger on ALOT of cars and rival or beat the performance...which is basically all Ford did to do it. I am NOT a fan of most of Ford's current engine lineup. GM seems to be able to get more power and better economy out of a really well designed pushrod V8 that probably even weighs less. Those OHC engines tend to be kinda piggy on weight. The old 427, 351, and even the 302's where pretty light (freaky light in the case of the 302) and powerful in comparison.

Eh...consider this an attempt at an educational off-topic potential troll. Try modding that one, I dare you.

Re:Just get...

[Exitthree]

I know it's not a new idea or revolutionary. Houses have had separate zones forever. The reason it gets brought up is because it is relevant. There is always heat in the computer, but instead of having one fan on all the time cooling the entire case (because the heat isn't localized), you can have one fan in one part of the case cooling one component. This keeps the computer quiet.

The zones are mentioned every time someone calls a G5 noisy based purely on the fact that it has 8 (or 9 in the duals) fans. People call the G5 noisy a lot, because they have never seen one in operation and are ignorant of how quiet they really are. Therefore, the zones get mentioned a lot in an attempt to educate the ignorant.

Re:Just get...

[EnglishTim]

I'm sorry, but he's got a point.

When you try and prove the wealth of Mac games by listing several games that are five years old, you've got to realise that your argument is a little thin.

Yes, I'm sure that Macs are lovely and all, but you're not going to buy a Mac for gaming. Sure, games do come out for them which I'm sure is lovely for all those Mac owners who like to do a little gaming on the side, but if gaming is a big thing that you want to do on your computer, you're going to want to get a PC with Windows.

Of course, it's a lot cheaper just to buy a games console if you're not too bothered about missing out on RTS and the other more cerebal, point-and-clicky games. Console games do seem to work out more expensive though.

Re:Really nice alternative to dual processor syste

[Monkelectric]

You my good man are dead on. I have a PIIx300 system, and it is more responsive then my athlon 1700. not faster -- more responsive. Back in 1998 when that was a freaking awesome system, remember when burning a CD was a fidgety process and even having your screen saver come on would ruin your cd? Wiht this dual system you could burn a CD and play QuakeII At the same time, it was really great :)

Re:Why not 8 x i486 cores? Toast, well done please

[Nom du Keyboard]

You're not gong to run an OS on a Cell, it's not designed for it.

You're certainly going to run some form of OS on it. You couldn't use it at all otherwise.

IBM has already talked about Cell-based workstations, so they have some ideas on how to use it.

Other main reasons

[abb3w]

...are word processing (M$- or Open- Office), Web surfing, and E-mail... which are for 99% of the people using them more performance limited by the computer-to-network or keyboard-to-chair interfaces than the CPU. Pretty much all of those can be done quite nicely by a 333 Celeron or PPC750 233MHz chip, running your OS of choice. Games, Video Coding, and Photoshop are the primary CPU intensive operations... leaving aside constantly bloating OSes.

Which was sorta the original point of that idjit: computer gaming, video work, and photoshop are the most common reason for getting USE out of a high-end CPU. Of course, if you routinely use a java script to pop open all forty of your daily web comics, you can shock almost any CPU (and your DNS server) quite nicely. =)

Mixed SMP systems?

[42forty-two42]

Does this work with a dual-core and a single-core opteron installed to make 3-way SMP, or must both either be dual or single?

And...

[StarKruzr]

... how long did it take for those titles to come out after the PC version came out?

Most of these are *remarkably* dated games. Q3A? FALLOUT?? Starcraft! Come on, man.

Sounds familiar...

[gphinch]

Mach 3 vibrator vs. Schick Quatro

Re:What about Apple?

[AKnightCowboy]

Your G3 iBook is probably slow because:
* Less L1 & L2 cache than the Pentium (this is the purpose of a PowerBook)
* You're probably not running with enough ram for OS X and your apps (512+ gives a good headroom)
* Did you buy the budget HDD for the iBook? This matters a lot.
* If the PIII is running Linux, obviously it will be faster.

You should actually try using a decent G5 for a couple days before you start slandering PPC.

If I could afford a G5 or a Powerbook why the heck would I have bought an iBook? ;-) To answer your questions though, the cache issue could be very likely, but as for the memory, my Dell Inspiron 4000 has 256 megs of PC100 SDRAM and the iBook has 640 megs of PC100 SDRAM. Memory is not an issue as the machine is never swapping.

What exactly was the budget HDD for the iBook? There was only one option, 4200RPM 30GB drive which is the same speed as my Dell so that's a moot point. One thing vastly in it's favor though is that the iBook is virtually silent whereas the hard drive in my Dell reminds me of some old band-saw like noises my Sparc 5's SCSI drive used to make. It is very loud and very annoying to work with it for extended periods of time. Probably time to replace it. The PIII is running Windows 2000 by the way.

Now, don't get me wrong, the iBook is a nice little machine, in fact I wouldn't have sent it in to Apple for the second time within 7 months for repair (first logic board died, second time now the backlight is dead) if I didn't think it was worth keeping. I like OS X but it's very hard to justify the cost premium for the Apple hardware in order to run it. For the $2300 a 15" Powerbook would cost with a DVD burner I could get a very nice eMachines laptp with an AMD64 chip in it for $1500 with more features, bigger hard drive, faster processor, etc. Oh well.

Re:Just get...

[evilviper]

you can have one fan in one part of the case cooling one component. This keeps the computer quiet.

One fan in one zone cooling a section of a computer is no more quiet than one fan in an unsectioned computer, cooling it down.

Therefore, the zones get mentioned a lot in an attempt to educate the ignorant.

But the (lack of) noise has nothing to do with zones. It has everything to do with thermally-controlled fans. Zones don't make it any louder or quieter.

linky

[Anonymous Coward]

27 Dec 2003
Athlon 64 and Opteron dual-core explained

29 Apr 2004
AMD boss says dual core Opteron will plug into existing boards

Re:What about Apple?

[ZigMonty]

The point is that AMD and Intel need to carry on a huge decoder circuit and handle lots of baroque complexities in hardware, that the PPC engineers never need to even think about.

Actually, the PPC970 does decoding similar to that which an x86 chip does. PPC instructions are converted to a simpler, more RISCy instruction set. Most PPC instructions are 1-1 mapped with their internal counterparts, but many are cracked into 2 internal instructions and some are microcoded.

Most applications are not multithreaded

[sorbits]

You say any modern GUI application. This is far from my experience.

All I know avoid threads like the plague, since it introduce a lot of synchronization problems and makes debugging very difficult since the program will in effect behave non-deterministic.

It's true that a GUI application which will need to do some long calculations (like ray-trace an image, compile a source, translate a LaTeX document etc.) will often spawn a thread to do this, but a) how many modern GUI apps work like this? b) the main thread (running the GUI) will have close to no load and c) it spawns only one thread, so the task will not complete in shorter time.

Even an application like a browser could very well be single threaded and use select() to wait for socket activity, and then dispatch new data to the proper page-parser, image-decoder or similar, since the network is unlikely to provide data so fast that the cpu will get a 100% load.

Furthermore, many operating systems move an entire process, with all its threads, from cpu to cpu, but not individual threads (since the communication overhead would affect performance).

Re:Most applications are not multithreaded

[rebelcool]

All I know avoid threads like the plague, since it introduce a lot of synchronization problems and makes debugging very difficult since the program will in effect behave non-deterministic.

Maybe if you have no idea what you're doing or how to use threads properly... I will admit, one does need adequate training to use them correctly. Once you have this though, all other methods are downright ugly and much more bugridden.

It's true that a GUI application which will need to do some long calculations (like ray-trace an image, compile a source, translate a LaTeX document etc.) will often spawn a thread to do this, but a) how many modern GUI apps work like this? b) the main thread (running the GUI) will have close to no load and c) it spawns only one thread, so the task will not complete in shorter time.

In a modern GUI app, at the very least, input and display will be running in separate threads. That is, while its waiting for you to click on a button, and anything that causes a display change.
If you add different ways of input (such as network), that will also run in its own thread.

Even an application like a browser could very well be single threaded and use select() to wait for socket activity, and then dispatch new data to the proper page-parser, image-decoder or similar, since the network is unlikely to provide data so fast that the cpu will get a 100% load.

The problem of the network not supplying data fast enough is exactly why threading is needed. Would you want your browser to be completely non-responsive (even in such tasks as displaying a menu or manipulating configuration options) while you wait for a page to download and display?

Re:Most applications are not multithreaded

[sorbits]

[threads] Maybe if you have no idea what you're doing or how to use threads properly... I will admit, one does need adequate training to use them correctly. Once you have this though, all other methods are downright ugly and much more bugridden

I think this is a bit harsh. I have been programming on a multitasking operating system for 10+ years where I have occasionally used threads. Through my degree in computer science I also have theoretical knowledge about semaphores, monitors, deadlocks and similar (it was a rather large portion of the OS course which ended with us writing a kernel allowing user programs to spawn threads).

Still I do all in my power to avoid threads, because it requires a very good overview to design it correctly if threads need to share resources (and it needs a very good understanding of what is safe and not safe when multiple threads work in parallel, I'm tempted to say "on the instruction level"), and debugging is as said very hard, because you may be using the program for half an hour, and suddenly it deadlocks, or it crashes because some thread is releasing a resource that another one used or similar, and it's nearly impossible to reproduce these conditions because of timing issues.

I have seen a lot feel exactly like I do, who were all skilled programmers, and I see recommendations to use asynchronous OS functions rather than do ones own threading. So I do not think this is just learning to master it. Every programmer makes mistakes (I would at least think), and threads introduce a wealth of problems to solve, and if you don't do it correct, you may never notice, because the problem will only arise in 0.01% of the runs (because of timings), and that results in software which seems unstable or freezes for only 0.01% of the users, which I actually think is not an uncommon phenomenon ;)

(not that I blame threads on the non-deterministic instability found in some programs)

In a modern GUI app, at the very least, input and display will be running in separate threads. That is, while its waiting for you to click on a button, and anything that causes a display change.

I am not entirely sure what you mean by this. Generally the OS takes care of waiting for hardware devices, it gets an interrupt when there is data on a device, and it then sends a message to the proper program (which directly or indirectly is "waiting" for that device).

How the application then handles the input depends, but in most of the GUI kits I have worked with, there is one central event loop which waits for these IPC messages, and take the proper action. This "waiting" is done in the main thread, and all GUI work is also done in that thread.

So I do not see any multithreading here, nor do I see the advantage. It would mean extra communication, and there is a serial flow between receiving input and performing the action they prescribe.

There is generally one "input" and one "output" to a program. For a command it would be stdin/stdout, for a GUI application it would be IPC messages and a GUI-kit as output. So if there is a 1:1 mapping between input and output (i.e. press "e", see an "e" character in the active text field), there is absolutely no reason to use threads.

Only if e.g. pressing "e" will result in a fractal image being generated, which takes 2 seconds, and is then displayed, then it is an advantage to offload that to another thread.

If you add different ways of input (such as network), that will also run in its own thread.

Certainly not by default. Generally the program will receive a message about new input, and it makes sense to direct all messages to the same message port and then have one dispatcher.

The problem of the network not supplying data fast enough is exactly why threading is needed. Would you want your browser to be completely non-responsive

No, a network a

Re:Most applications are not multithreaded

[rebelcool]

I didn't mean to insult you, but I do have to point out that most of the people I see bitching about threads are usually older programmers used to the 1:1 input/output paradigm of yore.

I too have a CS degree and work in research. Nowadays, threading is commonplace and is taught much better than 10 years ago. More than just semaphores and monitors, proper overall design is (or at least should be) taught as well. Java's excellent thread implementation have helped in this respect.

Deadlocking isn't an issue if you design it right. Theres no good reason an app will deadlock .01% of the time if you know how to use your threading tools properly, and how to implement the threads.

The problem with putting everything in central loops is the OS is going to switch the process out in the middle of those extremely long loops. With threads, in many cases another thread will get the CPU time (assuming your OS is handling thread scheduling). This increases responsiveness of the program and potentially actual program speed.

And its a much more elegant way of doing things than giant monolithic distributor functions. What do you do when one part of the function needs to wait on another task to complete? I can only imagine the horrors of trying to deal with that in a consistant, bug free manner. At least with threading all you need to do is call your wait() and notify().

Simple, clean, and tight. There's no reason in the 21st century to still be using 1970s styled monolithic functions.

Don't fear the thread. Its your greatest ally in modern programming. Yes it can be hairy if you're not well versed in the design methodolgies behind it. But in the end, it is much cleaner and more reliable than what you've described to me.

Re:Most applications are not multithreaded

[sorbits]

I didn't mean to insult you, but I do have to point out that most of the people I see bitching about threads are usually older programmers used to the 1:1 input/output paradigm of yore.

Don't worry, I wasn't insulted. And I'm certainly not an older programmer ;)

[...] work in research. Nowadays, threading is commonplace and is taught much better than 10 years ago

It has probably been 6 years since I had the course in question. I don't know if things has since improved, but it's not like I have a problem writing a consumer/producer architecture (using threads) or solving the dinning philosophers problem. The problems arise in complex programs.

Working in research, how complex programs have you been involved with where threads were used? I am asking not because I have prejudices against researches (heck, I am very close to going back to a ph.D. myself, as I sometimes favor research over the practical work), but my aversion toward threads have come through practical work, on the paper I can also agree to how "beautifully" it solves a problem, but practice have taught me otherwise (and the paper doesn't always show the code-overhead required for proper synchronization, and code-overhead add to complexity).

Java's excellent thread implementation have helped in this respect.

Here I hole heartily agree (despite my lack of fondness for Java in other contexts), but remember, most modern GUI applications are *not* written in Java.

Deadlocking isn't an issue if you design it right. Theres no good reason an app will deadlock .01% of the time if you know how to use your threading tools properly, and how to implement the threads.

Well, nothing is really a problem if you design it right!

I don't know if you agree with me, but in my experience most causes of failure in programs (apart from off-by-one and other "typos") are, that a situation arise which the programmer did not consider (and thus did not explicitly write code to gracefully handle, or he did not prevent the situation from arising in the first place).

My claim with threads is that you basically get an exponential number of program flows. E.g. if you spawn 3 threads, and they each send a single message to the main process, then the order of arrival has 3! cases -- and thus the program flow quickly gets difficult to overview. Adding to this are the other subtle details, like a thread sending a shutdown message, but when that message arrive, the thread might actually still live, or when the application needs to quit, but it needs to wait for threads to shutdown, but some thread is waiting for a response from a main-thread service (which is no longer running, because the application is about to shutdown) etc.

Actually, I have still not found a good solution to the later problem, i.e we have a worker thread which frequently communicate with the main thread. When the main thread quits, it needs to ensure that the worker thread also quits (it may need to do cleanup).

This problem is also littered with implementation specific problems, because how an application quits depends on the framework used -- you can probably solve the above in an isolated Java program, but when it has to be written using ATL, Qt, Cocoa or another GUI library, then suddenly it has to play nice with these as well (and a quit signal might come as a ctrl-c (interrupt), a menu choice, a scripting command etc.)...

The problem with putting everything in central loops is the OS is going to switch the process out in the middle of those extremely long loops. With threads, in many cases another thread will get the CPU time (assuming your OS is handling thread scheduling).

Which OS handles thread scheduling? Those I know all let that happen in user space.

Further more, if they did handle thread-scheduling, how would