Slashdot Mirror

• AMD Opteron are server-oriented CPUs, they require the extremely expensive registered ECC RAM to work and special Motherboard, they use the Socket940 which has (surprise !) 940 pins. It's been designed to work in server environments and multiprocessor boxes
• AMD Athlon64 is the desktop version of the Opteron, the older generation ran on Socket754 and the current one, along with the AthlonFX CPUs (which are basically the high end A64) runs on Socket939. These chips are not compatible with S940, and run just fine with regular DDR Ram

Well I am not an expert on the latest AMD stuff (of course they had the two core design in mind from the beginning of the Opteron) but from what I have seen the two dies in the Opteron interface with a common crossbar switch (the each have an independent L2 cache so no real unification at that level).

So they are not interconnected via HT on die but by some cross bar interconnect that presumably allows some level of concurrent point-to-point (core to core , core to HT link, core to memory, etc.) transfers and at high clock/data rates.

see this diagram ... or in a nice pdf

Well I am not an expert on the latest AMD stuff (of course they had the two core design in mind from the beginning of the Opteron) but from what I have seen the two dies in the Opteron interface with a common crossbar switch (the each have an independent L2 cache so no real unification at that level).

So they are not interconnected via HT on die but by some cross bar interconnect that presumably allows some level of concurrent point-to-point (core to core , core to HT link, core to memory, etc.) transfers and at high clock/data rates.

see this diagram ... or in a nice pdf

I wouldn't say "rebranded" (because they are not exactly the same), but "Intel" motherboards are/were definately manufactured by Asus, Foxconn, Wistron and other Taiwanese companies. For example, see:

http://www.xbitlabs.com/news/mainboards/display/20 040702071453.html

And btw, AMD *does* make their own chipsets, for use on server (Opteron and Athlon MP) boards:

AMD-8000(TM) Series Chipset
AMD-760(TM) MPX Chipset

I wouldn't say "rebranded" (because they are not exactly the same), but "Intel" motherboards are/were definately manufactured by Asus, Foxconn, Wistron and other Taiwanese companies. For example, see:

http://www.xbitlabs.com/news/mainboards/display/20 040702071453.html

And btw, AMD *does* make their own chipsets, for use on server (Opteron and Athlon MP) boards:

AMD-8000(TM) Series Chipset
AMD-760(TM) MPX Chipset

I haven't had any luck finding the wattage of Pentium-M chips from Intel's site. AMD had some information posted.

Sempron - 27 Watts
Pentium-M - 21 Watts

AMD's product comparison.

I know that the Pentium-M chip lacked some features that have been added to the latest round of Pentium-Ms.

why not .. everybody else is doing it

You can also put 4 dual core processors in a normal 4 socket motherboard to make a 8-way computer. AMD has tons of information on this stuffhttp://www.amd.com/us-en/Corporate/VirtualPre ssRoom/0,,51_104_543~94936,00.html

The two processors are going to have seperate L2 cache, but share the memory controllers and HT links. So it will work just like todays chips except for it will have two cores.

Read more about it http://www.amd.com/us-en/0,,3715_11787,00.html?red ir=CPPA65

The AMD 386dx40s were also good processors. I had one, and it never failed me or otherwise caused incompatibilities, random crashes, or lockups. I don't know which x86 processor AMD cloned first(they were a cloner company in the x86 market up until the k6, really), but according to their own history they were cloning x86 processors as far back as 1980 or 1981.

Problem is, Intel manufacturing is so expensive, they can't afford to give Dell any more of a discount
Intel can manufacture cheaper than AMD. If you look at the 2004 financial numbers, Intel has much better gross margins than AMD.
Intel Revenue: $34.2B
Intel Cost of Sales: $14.5B
Intel Gross Margin: 58%
AMD Revenue: $5B
AMD Cost of Sales: $3B
AMD Gross Margin: 40%

"Arrgh... practically every point in the above post is misleading or wrong, and it get's modded to +5"

Ok, the two dies on one chip was true, or believed to be true when they first demoed:

Source: http://www.xbitlabs.com/news/cpu/display/200409151 34740.html

I'll concede that point to you - that Intel is now putting 2 cores on a die... however they were never engineered to work that way initially. They only have an 800 MHZ FSB, not 1066 like the newer P4's, so they have even less bandwidth to share. Want a source?? Here:

Source: http://www.anandtech.com/printarticle.aspx?i=2252

As far as my other points go, let's go over them, shall we???

"the new dual core P4s won't be compatible with a majority of Intel boards on the market"

Source: http://www.theinquirer.net/?article=21793

"The two cores use hyper transport to communicate with various system devices"

Source: http://www.amd.com/us-en/0,,3715_11787,00.html

It actually uses a cross-bar to handle the switching as well.

"Now for the best part - anybody with an existing Socket 939 AMD based motherboard will be able to use one. Worst case, you'll have to download a bios update to enable it, but it will work."

Source: http://hardware.gamespot.com/Story-ST-x-1583-x-x-x

"AMD designed the K8 core to be dual ready out of the box, so this whole thing about them having an extra year isn't exactly true - they've had much longer than that."

Source: http://www.theinquirer.net/?article=13344

Still think almost my whole post was wrong? About what you said:

"Re: sharing I/O bandwidth. Intel has to do this because they don't have a built-in MCH. It has *nothing* to do with "selling chips with 2 normal P4 dies on them"."

It has EVERYTHING to do with having 2 P4 cores in a single package - look at that anand article I posted above, here is a quote from it:

"The major issue with Intel's approach to dual core designs is that the dual cores must contest with one another for bandwidth across Intel's 64-bit NetBurst FSB. To make matters worse, the x-series line of dual core CPUs are currently only slated for use with an 800MHz FSB, instead of Intel's soon to be announced 1066MHz FSB. The reduction in bandwidth will hurt performance scalability and we continue to wonder why Intel is reluctant to transition more of their CPUs to the 1066MHz FSB, especially the dual core chips that definitely need it.

With only a 64-bit FSB running at 800MHz, a single x40 processor will only have 6.4GB/s of bandwidth to the rest of the system. Now that 6.4GB/s is fine for a single CPU, but an x40 with two cores the bandwidth requirements go up significantly."

Your post seems to indicate you are not looking at the entire Turion line. Indeed, there are chips with only 512K of L2, but there also others with 1MB. Also, the Turion has double the L1 cache compared to the Pentium-M, 64K.

Maybe the reason your last post was modded down was your lack of supporting evidence and merely speculative.

AMD data: http://www.amd.com/us-en/Processors/ProductInforma tion/0,,30_118_12651_12658,00.html

Intel data: http://www.intel.com/products/notebook/processors/ pentiumm/index.htm
http://download.intel.com/design/mobile/datashts/3 0526201.pdf L1 Cache size is stated on page 7, second bullet

they use 754 http://www.amd.com/us-en/Processors/ProductInforma tion/0,,30_118_12651_12658,00.html

All in all I see that Intel is going down unless they do something quick. And remember Competition is good for the Customer.

The Acer Ferrari 3400 uses an AMD Athlon 64 Mobile, which sits in a 754-pin socket (source here). The dual-core sockets have 939 pins. Also, your Ferrari would overheat with a non-mobile S754 - nevermind a dual-core S939.

I believe at the time their systems came with 1 year of free on site support. If their phone tech support determined there was a hardware problem, they'd send someone out to do the work.

It appears this option is still available as "At-Home Service".

Not sure what their support is currently like. They sent a guy to my house to replace my CD-ROM back in 1996 without asking too many questions.

Tech: What's the problem?
Me: The CD tray stops half way when I press the close button.
Tech: Ok. We'll send out a tech to replace it. Will you be available Tuesday from 1-3pm?
I think the conversation was a little longer, but it didn't take multiple calls or many hoops.

AMD lists several places to buy AMD64 systems, many of which offer the same optional on-site repair contract. Not sure who lives up to their claims, but for that I'll have to read the reviews as it sounds like few actually do these days.

For the rare performance critical parts it is however worth the effort to try various constructions to get the best performance out of the code. The most problematic issue is to identify the hotspots in the code and figure out which variables that should be declared as 'register' and those that shouldn't. Ordering of statements are also important in order to match the various performance improvments the CPU can offer. One very good document on this is actually found at AMD.

One code construct that I am using that I found is very useful is to place the matching '{' and '}' in the same column in the code. This eases the effort trying to find where a block begins.

Example:

if (a == b)
{

printf("Hello!\n");

}

Even a simple sequential read loop that exceeds the L1 cache can benefit greatly from the appropriate cache hints in the assembly (prefetchnta and its variants).

Toss in a second processor and a NUMA architecture, and everything gets even more fun.

For examples of the hacks you can do on Opterons that vastly improve simple C code speed, take a look at the stream.c source and see AMD's technical pubs no. 25112, 24592, and 32035.

--Pat / zippy@cs.brandeis.edu

Even a simple sequential read loop that exceeds the L1 cache can benefit greatly from the appropriate cache hints in the assembly (prefetchnta and its variants).

Toss in a second processor and a NUMA architecture, and everything gets even more fun.

For examples of the hacks you can do on Opterons that vastly improve simple C code speed, take a look at the stream.c source and see AMD's technical pubs no. 25112, 24592, and 32035.

--Pat / zippy@cs.brandeis.edu

Even a simple sequential read loop that exceeds the L1 cache can benefit greatly from the appropriate cache hints in the assembly (prefetchnta and its variants).

Toss in a second processor and a NUMA architecture, and everything gets even more fun.

For examples of the hacks you can do on Opterons that vastly improve simple C code speed, take a look at the stream.c source and see AMD's technical pubs no. 25112, 24592, and 32035.

--Pat / zippy@cs.brandeis.edu

More specifically, the chips cost $1,299 when they were first released.

Many people seem to forget that this was not considered abnormally expensive for a processor not all that long ago.

Or, I went to AMD's page here and clicked on one of the manufacturers listed. Where I found this dual opteron supporting 16GB ram. Took me all of 2 minutes.

If it can keep up with AMD on performance, power/heat requirements and compatibility is another matter, my vote still goes to AMD, especially since I suspect foul play by Microsoft on 64 bit support. Windows 64 bit will be released as soon as enough chips are available from Intel.

I've been using the pre-release versions of XP 64-bit for a while, and the only weakness is a lack of 64-bit drivers from hardware companies that don't provide source code to Microsoft (all the hardware with driver support in Windows works fine on the 32- and 64-bit versions). I'm just guessing, but my suspicion is that Microsoft are waiting for drivers before releasing the OS. I mean, if I had bought a final version (rather than getting a free RC version), I'd be an unhappy customer, since there's no driver for my 802.11g hardware.

* Dave Cutler is the most senior software developer at Microsoft, and headed the design of the NT kernel.

If it can keep up with AMD on performance, power/heat requirements and compatibility is another matter, my vote still goes to AMD, especially since I suspect foul play by Microsoft on 64 bit support. Windows 64 bit will be released as soon as enough chips are available from Intel.

I've been using the pre-release versions of XP 64-bit for a while, and the only weakness is a lack of 64-bit drivers from hardware companies that don't provide source code to Microsoft (all the hardware with driver support in Windows works fine on the 32- and 64-bit versions). I'm just guessing, but my suspicion is that Microsoft are waiting for drivers before releasing the OS. I mean, if I had bought a final version (rather than getting a free RC version), I'd be an unhappy customer, since there's no driver for my 802.11g hardware.

* Dave Cutler is the most senior software developer at Microsoft, and headed the design of the NT kernel.

Not really sure how you'd go about hobbying around with one. We bought 3 devboards (DbAu1100) directly from AMD, and IIRC they were over $1000 each. But that was over a year ago. Might have changed by now.

It's a shame there's not something like the LART StrongArm board for the Alchemy yet. They're really great processors. I had one running an Xorg X server and the IFS screensaver (which takes the CPU to 100%) all day, and it was still cooler than a cup of coffee. Barely even warm, really.

Via, which took over Cyrix, is building low voltage, low power CPUs. Both CPUs support x86 instructions. VIA Eden-N AMD Geode(TM) GX

Also of interest would be AMD's Geode line of processors, found here: http://www.amd.com/us-en/ConnectivitySolutions/Pro ductInformation/0,,50_2330_9863,00.html. It looks like these guys run on about 1 Watt and are x86 compatible.

They aren't. However, AMD does have a line of x86 embedded processors known as Geode. For example, I see here a 400Mhz Geode consumes 1.1 watts of power. This is part of their more general x86 everywhere plan...with x86 chips as cheap as $1.

The 1100 is aimed at that market. It seems most of the line is special use - routers & various appliances.

The 1100 Development Board looks like fun, though. Is it bad when you have the urge to say, "Ooohh, preeeeetty" even if you have no real use (or skill to work with) it?

If you want x86 processors AMD already realesed their Geode line for low-power mobile solutions.

People seem to think this is leaps and bounds above everything else, but they're missing the details. In order to obtain that much performance, you'll need a task which parallelizes well so it can be broken up into chunks for the 8 SPEs. Graphics rendering falls into this set of tasks, but a lot of general applications just don't gain that much from parallel processors. Even when you have a task that does parallelize, writing parallel code is quite a bit harder than writing code for just a single thread of execution.

I've seen a lot of hype about having the Cell in your laptop talk to the Cells in your desktop, microwave, and TiVo, but you have to consider real-world limitations. When you set up a network like that (presumably wireless), you're going to be limited to around 100Mbps. In computer clusters and supercomputers, one of the main limitations of performance is the communcation bandwidth available between processors, and the latency of the network. To build a "home supercomputer", you not only need a task that parallelizes well, but one that doesn't require so much inter-node communication that it's held back by a slow network. You can't work around this problem with hardware magic - if the task you're working on requires lots of communication bandwidth, you're going to be held back.

So how much beyond a modern PC is 250GFLOPS anyway? Not much! A GeForce FX at 500MHz does 200 gigaflops. An AMD Athlon's peak performance is 2.4 GFLOPS at 600 MHz... if we scale this up to 2.2 GHz (high-end Athlon), that's 8.8GFLOPS (note: As we're talking about theoretical performance, nonlinear factors like bus speeds can be ignored). Basically, if the Cell dedicates most of its power to graphics rendering, you'll have computation power in the same range as a fast PC of today. Given that we're not going to see any products based on the Cell for a while, this isn't going to be the end of the world for Intel and nVidia (let alone the fact that Cell isn't x86).

Consoles using the Cell will have the advantage of only having to render for TV resolutions - at most 1080 lines, while PCs will be rendering at up to 1600x1200, but if you look at recent history, you can compare the xbox to a then-good PC with a GeForce3 (which came out at around the same time) - the xbox looked better, but PCs did catch up and surpass it's performance and it didn't take all that long. Consoles have to be very high-end when they're released, because the platform doesn't change for 2-3 years, and they still need to be "good enough" after a couple years, before the next generation is released.

The market has already decided it doesn't give a shit about itanium. Only one market has shown significant interest - and that is in scientific computing. Chip for chip, last I checked it had the most FPU around, although the new POWER offerings may beat it there, too, I don't know. Of course, dollar for dollar, its FP performance was nothing special, but if you can't afford to build a system with umpteen processors, itanium has (or had) a purpose. But... that's about it.

I agree that GNOME and KDE would be nice to see across all their platforms, and I even think it makes sense.

Opteron is already designed for high speeds without an excessively designed pipeline, anyway, and since processors are getting wider (more functional units) the clock speed thing is going to not matter AS MUCH (it will ALWAYS matter) as it has in the last bunch of years. AMD has moved beyond the bus used on Slot/Socket A; Putting the memory controller in the processor, for example, means that they can use hypertransport for everything. That makes sense, of course, because it simplifies the chipset, too.

As for Dirk Meyer (third google hit, you lazy bugger) he already works for AMD and has since 1996. So, what do they need from the Alpha guys, again? Looks to me like they have it all already.

I remember using my old Athlon to warm my apartment on cool winter nights...

On a more serious note, I think that it's impressive that AMD has turned it around in the temperature department while still delivering great performance. The AMD64's kick ass when you want them to, and then scale back to be cool when you aren't running processor-intensive aps. I'm sure Transmeta or someone else thought of doing something like that before, but I think that this is where we'll see it more and mroe often.

I wonder if they will choose to go with the more obvious x86 based Geode or the very deserving MIPS solution, the Alchemy. Personally I would really love to see the Alchemy used.

I wonder if they will choose to go with the more obvious x86 based Geode or the very deserving MIPS solution, the Alchemy. Personally I would really love to see the Alchemy used.

I wish that someone would do another VIC-20. For $200.00

AMD already did it. It's not that cool.

Um...there is no x86 emulation mode. The AMD64 is x86 with 64 bit extensions. All x86 software runs EXACTLY the same with no emulation.

Umm... are you saying that the AMD64 64-bit mode instructions are a superset of the i586 instructions? This is not the case and for this reason there is a seperate 32-bit compatibility mode for running 32-bit applications under 64-bit kernels. (See citation below.)

I think the poster meant 32-bit compatibility mode, not "emulation mode". The differences between a "legacy emulation mode" and a "legacy compatibility mode" are mostly semantic. In any case, the AMD64 instruction set is not a superset of the i586 instruction set, hence long-mode 32-bit ring3 "compatibility mode", which causes the CPU to run in 64-bit mode in ring0, but 32-bit mode in ring3. The FPU programming model is no-longer stack-oriented in 64-bit mode and I've heard that the BCD arithmetic instructions are gone in 64-bit mode. However, I have no personal in-depth knowledge of the ISA changes. Some instruction encodings have definately been eliminated in 64-bit mode. (See citation below)

See "The AMD64(tm) ISA Value Proposition", especially the first sentence in the last paragraph of page 5, which reads "Additionally, some of the binary instruction encodings have been eliminated to support future instruction set expansion." Also see the bottom of page 9, where it mentions the outdated x87 FP stack is eliminated in favor of a more modern FP register file. Clearly, the AMD64 ISA is not a superset of the i586 ISA.

Compare the multi-level 16-bit/32-bit/64-bit AMD64 compatibility modes (including kludges like the A20 gate) to the POWER/PowerPC ISA, where the original ISA was designed to be 64-bit with a pre-defined subset for 32-bit implementations. It's a shame Intel didn't come up with a completely new architecture and create a software emulator to ease transition, a la Apple's switch from m68k to PPC. At least AMD is finally getting rid of the x87 FP stack (except in compatibility modes).

Maybe some day there will be an x86 chip that starts up in 64-bit mode and allow the boot loader to switch the CPU into legacy modes so OS writers can forget all of that legacy kludge (including using a spare pin on the keyboard controller to determine the functionality of the 21st memory bus address line... necessitating keyboard controller chipset-dependent code to change addressing reverse-compatibility...)

that 400$ pos isn't anything compaired to AMD's nearly identical project called the PIC (Personal Internet Computer) powered by the newly aquired Geode line of cpu's. (http://www.amd.com/us-en/ConnectivitySolutions/Pr oductInformation/0,,50_2330_12264,00.html/
might i add that in my humble geeky opinion, the geode looks much better, and its smaller!
..a wise man once said, its not the size its how you use it.

..so i bought an old room-size VMS! ;)

SOI is not patented by IBM. Only certain components of IBM's SOI technique are patented. Their most important patented SOI component is their SOI FET Design to Reduce Transient Bipolar Current.

Intel will most likely use their own SOI technology without needing "permission" from IBM.

Tom's Hardware has some good information about thermal loss. Notice that an idle AMD Winchester (SOI Athlon 64) loses only 3.2 watts, while the more recent P4 chips are losing > 34 at idle.

Notice that you compared the "Cool 'n' Quiet" versions of the AMD Winchester to the older P4s (D0 stepping) without Enhanced Halt State. The same page you referenced shows a 3500+ Winchester without "Cool 'n' Quiet" technology losing 11.1 watts when idle and a "more recent" 3.4GHz P4 550 chip (E0 stepping with Enhanced Halt State) losing 13.4 watts.

This number changes at load to 30 watts for the Winchester and 100+ watts for the P4.

Only for the fastest (3.8GHz) P4. The previously-mentioned 3.4GHz P4 550 (E0 stepping) loses 73.6 watts and a slower Prescott (3.0GHz) loses 59.3 watts.

I know the new Winchesters are a lot cooler than the new P4s at load, but you seemed to be greatly exaggerating in your comparison (especially at idle).

AMD Alchemy(TM) Au1200(TM) Processor
http://www.amd.com/us-en/ConnectivitySolutions/Pro ductInformation/0,,50_2330_6625_12409%5E12410,00.h tml?redir=PCAU04

Oh, wait...
Built-in decryption hardware for digital rights management (DRM)

Does that mean you won't buy AMD chips either???

AMD Alchemy(TM) Au1200(TM) Processor

In the print versions sold locally (e..g in the HCC magazine) it is even more obvious as you see the whole machine.

Dw.

Because a processor with 64 bits of memory address bits can access more memory than one with only 32?

The AMD 64 actually has 48-bit virtual addresses and 40-bit physical ones. Check it here. 64-bit refers to the size of the registers/data paths. Whilst it is unlikely you will often operate on data which needs to be 64-bits (except for double precision FP for accurate maths applications) it does for instance allow for good use of SIMD instructions (4 16-bit values packed into one register and operated on at the same time for instance). Other tricks are possible too. The upshot is it will speed things up but generally only if programs/compilers are written with exploiting the architecture in mind.

I think AMD will still continue bumping clock speeds - it was Intel (which is a full GHz. ahead in that department, without getting any more performance) that blinked.

AMD and IBM recently announced new strained silicon technology that should facilitate higher clocks.

They still rock in one spot, the mobile processor market. I just got a new laptop with their Centrino processor and it's awesome. Loads of power, and I can run for 3 hrs easily. I was looking for a similar offer from AMD but to no avail.

Yup. AMD has a good guide to power consumption here. It has, among other things, a formula for computing the power usage of the processor, taking into account the expected inefficiency of the voltage regulator. Combined with spec sheets available from the manufacturers of various parts, it can tell you if a power supply will work.

I did all the math for a server I was thinking of building, and I was quite surprised by the outcome. Despite the fact that my server should only use about 140 W of total power, some of the 300 W supplies were inadequate. They didn't supply enough on the 12 V line. It looks like the maximum current draw on the 12 V line is a better indicator of the quality of a power supply than its total maximum power.

I recommend the PSUs from PC Power & Cooling. The Silencer units really are near-silent, and you can pump a lot of amps through the 12 V lead on all of their supplies. (An improvement they made since I bought mine, unfortunately.)

Why wait? AMD has this now and it appears Intel is now following AMD in this direction.

AMD Press Release

http://www.amd.com/us-en/Corporate/VirtualPressRoo m/0,,51_104_543~91999,00.html OOPS, forgot proper coding

The 530 AMD Opteron processors, which run on a Linux operating system