AMD to Demo '8-socket' Dual-Core Opteron System

flynn_nrg writes "AMD will make the first public demonstration of a system built out of its dual-core processors today, the result of a strategy first made public almost a year ago. Two-core Opteron chips aren't due to ship until the middle of 2005, but AMD will have four of parts running inside an HP ProLiant DL585 server at its Austin plant later today."

Comments

Four of parts?

[Anonymous Coward]

Is that 7 of 9's ugly sister?

Re:Four of parts?

[StarfishOne]

Everyone knows that 7 of 9 is not dual core.. tskk.. calling it a sister

Re:Four of parts?

[RsG]

No "four of parts" makes perfect sense in binary. Wait, you aren't still thinking in wetware language are you? ;-)

Re:Four of parts?

[PingPongBoy]

That was supposed to be four of hearts.

Re:Four of parts?

[Mattintosh]

I'd love to see some h0t du4l c0r3 br0g pr0n involving 7 of 9.

Re:Four of parts?

[Jhan]

... I just can't help but wonder if that's a minor arcanum/suit for some sort of geek tarot or playing card deck.

The Four of Parts... The Eight of Processors... The Hanged System... The Processor Affinity, inverted.

The Ten of Consultants... The Two of Millions... The Meltdown...

Fate bodes ill for your server upgrade.

Speeeed

[Klar]

I didn't see any specifics in the article, so I was wondering if anyone knows how fast the Dual-core Athlon 64's and Opteron's will be running? Has there been any clue's? I'm just wondering how long my processor will seem fast for, lol..

Re:Speeeed

[TheRealMindChild]

Call me a troll, but I would gather, pretty close to the same as if there were two processors.

Re:Speeeed

[essreenim]

:Lift toilet seat: ..lifted. :Relaese wallet contents into toilet: ...well done

Re:Speeeed

[JDevers]

I think he meant the speed of the demo CPUs in GHz...but I may be wrong...

Re:Speeeed

[mjuarez]

Well, although there have been no specific comments on CPU frequency for dual cores, I'd bet that these babies are running somewhere between 1.8Ghz and 2.2Ghz. Remember, these dual core is from the very first batches of 90nm AMD products out there. It will take some months to squeeze all the bugs out.

OTOH, I fully expect a 2.4Ghz dual-core Opteron available for purchase by July 2005. Meanwhile, Intel has absolutely nothing to throw at this, except for vaporware.

Re:Speeeed

[Short Circuit]

Faster than my 750MHz Duron. 'Nuff said for me.

Re:Speeeed

[v1x]

Since the implementation is essentially two cores on one die, the speed would be limited by whats available with the existing 90nm line at present. If you were asking about their performance rating, I'm guessing it might be way higher than the existing line of Opterons.

Re:Speeeed

[Anonymous Coward]

"Has there been any clue's?"

Huh? What language are you speaking? Did you mean "Have there been any clues?" perhaps?

Re:Speeeed

[Epistax]

Don't knock that vaporware! I plan on running Duke Nukem Forever on that. (They sound compatible)

Re:Speeeed

[Christopher Thomas]

Call me a troll, but I would gather, pretty close to the same as if there were two processors.

Performance of a CMP chip relative to a dual-processor system depends on the load. On one had, you have shared L3 (and maybe L2) caches (depends on whose CMP implementation you're talking about), which means you have two (or more) processes trying to use one chip's worth of cache space. On the other hand, if you have loads that are not cache-bound, you get faster inter-process communication than with a dual-core system (as data the processes are sharing is in-cache instead of in main memory).

Several types of scientific load meet the footprint requirement. Rendering might or might not, depending on what you're doing (tends to be memory-bound).

Re:Speeeed

[Anonymous Coward]

Both AMDs and Intel's projected offerings will be dual cache ("its easier that way") making them pretty much literally two seperate processors in a single package. There might be tiny increases in speed due to the shorter bus paths (and AMD's hypertransport) but on the whole a single dual core processor will be the same speed as two single core processors.

I am sure both companies are working on a shared cache implementation of the concept.

Re:Speeeed

[Christopher Thomas]

Both AMDs and Intel's projected offerings will be dual cache ("its easier that way") making them pretty much literally two seperate processors in a single package.

If I recall correctly the Power4 has a shared L3 (which they call "single-transistor SRAM", but which is actually DRAM backed by a number of SRAM buffers acting as cache for the cache).

I haven't checked on Sun's offering for a while, but IIRC it was supposed to have shared L3 as well. Not certain about that, however.

Re:Speeeed

[daft_one]

Here's a clue: pluralization doesn't involve apostrophe's. ;-)

Itanium?

[StevenHenderson]

but AMD will have four of parts running inside an HP ProLiant DL585 server at its Austin plant later today.

Does this mean HP is offically ditching the Itanium2? If so, strange move, albeit likely a smart one...

Re:Itanium?

[OmniVector]

that would be a pretty cripling blow, seeing as how HP is pretty much the biggest supporter of itaniums

Re:Itanium?

[drinkypoo]

It means that HP is hedging their bets, like a smart little company. Itanic still has better floating point from what I understand, and if you are willing to spend absolute gobs of money to get it, itanic may yet be the right platform for you. Of course most of the problems that demand high quantities of floating point seem to be running on clusters these days but I'm no supercomputing expert.

Re:Itanium?

[Anonymous Coward]

sent. enjoy!

Re:Itanium?

[gadget junkie]

..remember, this may be a Server processor, but it is also the close cousin of the PC dual core offering, see http://www.anandtech.com/cpuchipsets/showdoc.aspx? i=2178&p=2 . So, we will be able to see if it is worthwhile to spend bucketloads of money on a 939 socket athlon motherboard today.

Re:Itanium?

[JayJay.br]

Not really. Opteron is part of HP's strategy for 64-bit computing everywhere.
The roadmap looks something like this:

- Tandem (NSK) will eventually turn to Itanium (as soon as lockstep is deployed and working fine);
- Alpha and PA-RISC will evolve into Integrity (Itanium2);
- Proliants (IA-32) will evolve into Opteron.

It's just that 32-bit computing is taking its last breath, and it's time to move on.

Now that looks like a smart move.

Re:Itanium?

[flaming-opus]

HP has several (6 actually) server product lines. They will probably use opterons in their high-volume/lower-profit proliant server line. However they have firmly commited to ditching pa-risc, mips, and alpha for their other 5 server platforms. The high-end/high-profit/low-volume systems are largely independant of the proliant group.

HP doesn't view itanium and opteron as an either/or proposition. Given their product porfolio, it's quite reasonable to use both. Itanium is fast and expensive, a good fit for a 128-way superdome. Opteron is pretty-fast and inexpensive, a good fit for a 4-way proliant.

Re:Itanium?

[StevenHenderson]

Okay, thanks. Very helpful. Now if only more companies were this open-minded about using AMD.

Re:Itanium?

[Jeff DeMaagd]

Itanium isn't so poorly priced once compared to the 8-way Opteron 8xx series. 8 way and up computers are the current target market of Itanium.

Re:Itanium?

[Anonymous Coward]

Then again, is there any physical reason why 8xx costs more than 1xx or is AMD making 1xx a tiny bit cheaper than it would be otherwise by keeping artificially the price of 8xx so high?

Re:Itanium?

[Lumpy]

A large number of people make the claims that Intel has better FPU's. yet over the past year I have seen AMD's FPU's meet or even beat Intel's offerings on a regular basis in real world use.

Granted the Itanium is still "alphaware" and who knows when it will have a full release, I find it hard to believe that the AMD64 doesn not have a superior FPU, or at very least a comperable one.

Re:Itanium?

[vDave420]

--2 Gmail invites for people that want them...

Feel free to toss me an GMail invite.

dave [AT] freepeers [DASH] com.

DASH = DOT

Thanks in advance!
-dave-

Re:Itanium?

[at_18]

Itanium is completely different from the x86 line, and its FPU unit absolutely crushes the P4 one.

Re:Itanium?

[Jeff DeMaagd]

It's just that 32-bit computing is taking its last breath, and it's time to move on.

It depends what you mean by computing. Network appliances and embedded systems will be 32 bit for a looong time.

I also don't expect the HP IA-32 systems to all shift to Opteron. I think that was a hedged bet in case Intel didn't ship its chips with AMD64 compatibility soon enough. It isn't as if the two brands of chips can't coexist in the same model line either.

Re:Itanium?

[hawkbug]

Yes, but we're not talking about if the Itanium has better FPU than the P4 - we're talking about the AMD opteron.

Re:Itanium?

[JDevers]

I would say that on real FP (not SSE-3 or whatever), AMD has dominated the x86 line for quite some time. The Athlon easily had a stronger FPU than the P4. However, he isn't comparing it to an x86 processor, the Itanium definitely has more FPU power than any x86 processor...which it should, given it's intended market.

Re:Itanium?

[Anonymous Coward]

but what about it's FPU unit unit?

Re:Itanium?

[StevenHenderson]

Sent. Enjoy!

Re:Itanium?

[Anonymous Coward]

The 8xx and especially the 850 are the lowest-yielding chips AMD sells. Every part of the chip works: the core, the cache, both channels of memory controllers, and all three hypertransport links. And it all runs at 2.4GHz. So there's no mystery why this is their most expensive product.

Re:Itanium?

[Anonymous Coward]

How many memory contollers/HT links/etc does 150 have then, BTW?
Besides, if 8-wayness was cheaper I'd probably rather pick a few more of the cheaper models instead of spending a huge buck on the last few hundred MHz.

Re:Itanium?

[drinkypoo]

Athlon and Opteron processors have superior FP to all but the latest P4 processors according to benchmarks. Itanic, however, has far better FP performance when measured either clock for clock or even core for core in spite of the fact that itanium's clock rates are far lower than those of Hammer-core processors.

Re:Itanium?

[rmayes100]

More than just a supporter HP is heavily invested in the Itanium. HP designed the Itanium 2 and at least co-designed the Itanium 1 with Intel. HP has no doubt learned to hedge it's bets like IBM has and will sell their customers whatever they want but at some level they have to push the Itanium stuff (as IBM does it's G5 servers, though at least IBM has Apple to sell some G5s as well).

Re:Itanium?

[Anonymous Coward]

I thought the Opterons used the 940 socket? If so, then spending bucketloads of money on the 939 Athlons hoping to be able to upgrade to a dual core Opteron is a waste.

Re:Itanium?

[Anonymous Coward]

Gee...ya think would think so since it's derived from their PA-RISC processors and they helped Intel develop it.

Re:Itanium?

[Anonymous Coward]

The 1xx has everything the 8xx has, except it lacks two hypertransport links. The 1xx series has only a single link, and this must be used for system I/O, which is why the 1xx is exclusively uniprocessor. The 2xx series have two HT links, so they can form multiprocessor systems, and the 8xx, with three, can form efficient systems up to 8-way.

Re:Itanium?

[mlyle]

Granted the Itanium is still "alphaware" and who knows when it will have a full release

What are you talking about? Itaniums have been shipping since May/June 2001, and the Itanium II came out July-August 2002. You can buy one today. So it's pretty well known what's in Itanium and that it's faster for floating point than AMD64 (though not more cost effective, except for certain very high-end computing loads).

Re:Itanium?

[AJWM]

HP has been more than hedging its bets on 64-bit processors for a while. They're one of the few to currently offer an AMD-64 based laptop, for example, and have been offering AMD processors in (Windows/Linux) servers for a while. (Their Itanium-based Integrity servers are mainly HP-UX, although you can put Linux on them.)

Re:Itanium?

[AJWM]

However they have firmly commited to ditching pa-risc, mips, and alpha for their other 5 server platforms.

When has HP ever done anything with MIPS? That's originally an SGI chip -- although the MIPS core shows up in a lot of embedded applications too.

Agree on your other points.

Re:Itanium?

[lweinmunson]

Tandem. They use MIPS chips. Like all things SGI it runs like this. Use something, buy the company that makes it, go "oh crap, we need money let's sell xcompany." So they used MIPS, bought the company, then spun it off and sold it when they needed money. The only SGI MIPS chips are the r5k and r10K lines. But even those go out to MIPS licensees to build off of. Crack open a Cisco router. There's a good chance you'll find an R5000 of some sort in there.

Re:Itanium?

[Anonymous Coward]

The first chart at this realworldtech article is a good summary of the relative performance of today's processors. The Itanium is way out front in floating point performance.

Re:Itanium?

[vincecate]

That graph is out of date. For more current info check out the SPECmine top 20. Opteron has seen clock speed increases since then and compiler improvements so now is faster than Itanium on SpecInt and not too far behind on floating point.

Re:Itanium?

[Bedouin X]

It crushes the Opteron too. This isn't a troll, it's just a hard fact. I'm too lazy to find a link but check out the SPEC_fp numbers. Itanium is a beast.

If I had a choice though, I would still go AMD.

Re:Itanium?

[AJWM]

Thanks for the info.

Re:Itanium?

[turgid]

HP doesn't view itanium and opteron as an either/or proposition. Given their product porfolio, it's quite reasonable to use both. Itanium is fast and expensive, a good fit for a 128-way superdome.

itanium does OK on one or two specific benchmarks, but most people view it as an expensive turkey judging by the lack of sales.

Now, what they should have done was to abandon itanic when they got Alpha from DEC via Compaq and continue to develop that. It's been sad over the years watching HP commit suicide.

Re:Itanium?

[hawkbug]

Oh, I know it crushes the opteron - it's just that the post said P4, and I have no idea why because nobody was talking about the P4.

Re:Itanium?

[Anonymous Coward]

I am enduser running manufacturing operation on HP Proliant planning on Opteron switch. Recently have discussion, HP official party line still Itanium. HP also indicated difficulthy because AMD does not give a clear-cut roadmap for subsequent deployment, (CPU, chipset, components and misc.) which I believe is not true.

Below is my additional observation:

1. HP best mid-range option is new Proliant Xeon DL 380 series. Redundant CPU, PowerSupply, Fans, nic, built-in SCSI-RAID, 6 hot-plug disks in 2U. In short, a machine that will go anywhere to meet most general business server needs.

2. HP offers Xeon DL320, DL360, DL140 series in case enduser tradeoff for different needs.

3. If you supplied a hypothetical DL385 with Dual Opteron, otherwise same spec as Xeon DL380, it will guranteed to be a highly successful item. Unfortunately, many DL380 buyer will switch to DL385 as well, This is a huge impact to both HP and Intel. According to HP, the DL380 is very popular. My view is if AMD gained a strong foothold on this critical segment, it will create tremendous ripple effects everywhere. For me, once switch, I will likely gradually deploy infrastructure upward and downward on Opteron

4. If you thought about HP's response on technical, chipset and misc items, note that the high performance shipping Opteron-based DL585 are HP-DESIGNED which offers almost all the benefits seen on other Xeon Porliant. So HP technical capability is not the question, HP intention is the question.

5. HP offers Opteron DL145, a dual-CPU box with no RAID and no dual power supply. thus indirectly redirecting general server users to Xeon Proliant. (stability and data/service protection is important to general business server users)

6. If Opteron DL145 is no good, HP suggests DL585. The machine is definately capable with all the usual benefit expected of Proliant. But then most volume business server sales are 2-CPU box, unless we are talking about database server.

7. So that's a gap, where's the Opteron DL385 that we need? HP is not going to give me that option, indirectly forcing me to consider alternate option. IBM intention on Opteron is uncertain. That leave me with Sun's Opteron machines and maybe Dell? in future.

HP issues are quite understandable.
1. Volume sales are 2-CPU box.
2. HP plan all high-end to Itanium. This will offset the running cost in maintaining Itanium team since they have obligation for HP-UX and other Alpha/Tandem/misc migration to keep Itanium.
3. Opteron severely disrupted that picture because a 2-CPU box becomes 2-CPU-4-cores, and 4-CPU becomes 4-CPU-8-core Opteron(dual-core) next year. For several years, that will be sufficient for most folks (aka volume sales). HP Itanium transition goes up in smoke.
5. For now, HP will only do half-hearted work since IBM is not pushing, Dell is not selling, smaller server vendors has no concrete grip on enterprise customer (seldom made major change) so there's no rush of any kind, so HP will not release the mid-range all purpose DL385. The only unknown is SUN. HP suspects SUN need to protect SPARC so will not be agressive. Hence everybody is holding out. AMD is aware but need HP co-operation to get into Enterprise so will settle for anything now.
6. Situation will immediately change if Dell enter the market, or SUN suddenly pushing multiple 100K units of Opteron per quarter. HP will activate the DL3x5-series Opteron servers
7. Some may ask why Intel not considered. Based on current scenario, I am not sure Intel has a quick and concrete answer to Opteron in near future. So Intel may not mean much for now in my perspective. Since dual-core Opteron has been shown, longer-term issues are also covered down the road.

Re:Itanium?

[drinkypoo]

Incidentally itanium is only a fp math beast. It's not any faster than anything else in any other category, but it IS truly badass with floating point. This is funny because AMD became the floating point leader with the release of the original Athlon processor (what we refer to today as Athlon classic, or at least AMD does) and retained that title utterly, in the face of all other competition, until the release of Itanium.

Re:Itanium?

[Hoser McMoose]

As the other poster mentioned, this is part of the Tandem Non-Stop line. Servers designed for extremely high reliability.

This is a sort of odd-ball situation that came about due to all the buying and merging that went on. Tandem was originally an independant company (or maybe split off from another company?) that was bought out by Digital many years back. Digital planned to switch the servers over to Alpha processors, but never go the chance before they were bought out by Compaq. Compaq didn't do much of anything with Tandem and then they merged with HP.

Through this all the Tandem line just stuck with their old MIPS chips as the Alpha design never really got off the ground. Now the plan is to switch it over to Itanium, and I understand that this has progressed fairly well along, though it's likely to be an extremely slow process. Due to the sort of market that they're dealing with here, HP will probably be offering MIPS based Tandem servers for at least another 5 years.

Re:Itanium?

[drsmithy]

This is funny because AMD became the floating point leader with the release of the original Athlon processor (what we refer to today as Athlon classic, or at least AMD does) and retained that title utterly, in the face of all other competition, until the release of Itanium.

Hmm.

550Mhz P3 Xeon: SPECfp95 = 17.3 (8/99)

550Mhz Athlon: SPECfp_base95 = 20.6 (8/99)

1Ghz P3: SPECfp_base2000 = 292 (8/2000)

1.5Ghz P4: SPECfp_base2000 = 543 (11/2000)

1Ghz Athlon: SPECfp_base2000 = 298 (11/2000)

1.2Ghz Athlon: SPECfp_base2000 = 328 (2/2001)

Seems to me intel lost the crown ca. mid-1999 when the original Athlon was introduced and (convincingly) regained it ca. late-2000 with the P4 (even the difference between the 1Ghz P3 and Athlon is practically nonexistant).

Re:Itanium?

[cujo_1111]

It has nothing to do with being open-minded, it is whether or not they feel comfortable doing business with a much smaller manufacturer with a total reliance on one or two fabrication plants.

Re:Itanium?

[gadget junkie]

..Yeah right,the point is that future 939 pin athlon 64 will be dual core as well, and having practically the same capabilities as the opteron

Ad on site

[Anonymous Coward]

Is it it me or does anybody else see the irony in the fact that there was an intel advert on the page.....

Re:Ad on site

[ms00skr]

It's you.

(We all use adblock)

Re:Ad on site

[Anonymous Coward]

Is it it me or does anybody else see the irony in the fact that there was an intel advert on the page....

The page looks more bronzy to me.

Re:Ad on site

[Anonymous Coward]

> It's you.

How are you gentlemen?

Re:Ad on site

[StarfishOne]

Somebody set Intel up the bomb ::rolleyes::

Re:Ad on site

[Anonymous Coward]

Awww. Hhahahaaha

Re:Ad on site

[boy_afraid]

ALL HAIL ADBLOCK!

Comparison

[Nos.]

It would be interesting to compare the price/performance of these AMD chips versus the 12 cpu transmeta workstations we heard about yesterday.

Re:Comparison

[alphan]

Just 12?

There is a 96-CPU Workstation .

Specs should be somewhere there.

At this to your to compare list.

Re:Comparison

[Anonymous Coward]

The CM-2 had 65,536 CPUs and achieved an incredible 2.5 GFLOPS. Oh, wait...

Re:Comparison

[Ziviyr]

In the speed arena, imagine the 8x Opteron box maniacally laughing at the 12x Transmeta box.

Well, thats what I'm doing, followed by imagining the owner thinking about creative ways to use the generated heat.

Re:Comparison

[Hoser McMoose]

The Opteron basically smacks the Transmeta cluster silly when it comes to performance. Best case scneario for that 12-CPU Transmeta workstation is that it might match the performance a 2P x 2 cores/processor setup, but the Transmeta cluster would be about 3 times as expensive.

Of course, in many workloads a single Opteron 140 (the cheapest/slowest Opteron out there) will easily out-gun that 12P Transmeta cluster. Even in cases that are multithreaded one or two Opterons running at a decent speed will beat out the Transmeta cluster.

The real advantage of the Transmeta workstation (beyond the fact that it's available today and dual-core Opterons are not) is that the power consumption would be somewhat lower. Compared to what is available today this is a fairly large advantage, though those dual-core Opterons should be able to get very similar performance/watt when they are released.

8-socket?

[Anonymous Coward]

No, 4-sockets, each with dual core CPU.

Re:8-socket?

[Lord Kano]

Maybe that's why it's called '8-Socket' at the top of the submission.

LK

Re:8-socket?

[ajlitt]

Umm... no. RTFA, the submitter screwed up. A socket is what the chip goes in, and there are four chips in this box.

Re:8-socket?

[corngrower]

8 Sockets may be correct. I'm not sure whether or not they have an actual dual processor chip manufactured, or have just implemented the equivalent logic that merges two existing single
core opteron chips. Different articles on the web
seem to indicate different things.

Re:8-socket?

[johne_ganz]

No, 4-sockets, each with dual core CPU.

Actually, 8 sockets would be correct. There's three flavors of opteron: single cpu (1XX), dual cpu (2XX), and eight cpu (8XX).

Of course, nearly all the motherboards you can buy today only use four of the eight way SMP capability. The slashdot title is a bit misleading, they're only demoing a 4 socket / 8 core version today but an 8 socket system is doable right now, today, with the 8XX series CPU's.

As the article says:

So what today might be an eight-way server will potentially become, mid-2005, an "eight socket" server with 16 processing cores.

Re:8-socket?

[Chris Burke]

No, 8 sockets is still wrong. What they are demoing is a 4 socket board with 4 dual-core Opterons in it. There aren't any 8-socket boards, and in fact the point is probably to demonstrate that they can make an 8-cpu system by putting their new dual core chip into the existing 4-socket board.

The possibility of making an 8-socket board doesn't make using "8 socket" correct in this context.

Are they made of...

[kayak334]

Unobtainium?

Oh wait, that's something else...

Re:Are they made of...

[southpolesammy]

Yeah, ask Oakley Sunglasses. They seem to have discovered it...

Re:Are they made of...

[pragma_x]

Either Unobtainium or Deaminite.

http://it.slashdot.org/comments.pl?sid=112052&ci d= 9509105

Re:Are they made of...

[damien_kane]

Ya, they're also the ones who, throughout the 90s, used the slogan "Thermonuclear Protection".
Now personally, if I at some point found myself in the middle of a nuclear explosion, I wouldn't be too worried about whether or not my eyes survived the blast.

Of course, were I Oakley, I wouldn't be too worried about claims of false advertising. Really, who would test their sunglasses in such a way?

Cheaper Processors

[COMON$]

While it will be a while before I will be able to justify one of these at home. I am happy for any technology that will further lower the price of processors. Maybe a nice AMD64 will be in the future of budget home users.

Re:Cheaper Processors

[essreenim]

Yes thankyou Athlon64 s939 for making my Athlon64 s754 cheaper!

Re:Cheaper Processors

[fitten]

Ditto * 2. I waited until the S939 was released and the S754s dropped as a result, then I bought two m/b, cpu, memory sets to upgrade two of my home machines.

Re:Cheaper Processors

[gid]

There are already budget AMD64 machines as long as your build your own. I don't know if any big name vendors build such a machine yet as I'm not into prebuilt machines, but I wouldn't doubt if there are some available.

You can get a Chaintech K8T800 socket754 mobo for $64, an Athlon 64 3200+ (newcastle) cpu for $218, a WD SATA hd for $68, maybe a 512 meg stick of DDR400 ram for $78, a case for $60. What else do you need? Most people probably have everything else they can canibalize from their old machine. All that comes up to $488. These prices are all from newegg.

I'm looking at a new setup myself, but using a nicer, probably nforce3 mobo with better sound (hopefully it won't pick up USB/HD noise as I hear some people are complaining about) either from MSI or Asus.

Re:Cheaper Processors

[Frequanaut]

Justify an amd64 for home? How about $1300 for an amd64 laptop?

Re:Cheaper Processors

[Anonymous Coward]

I have been building systems for some time, I have made many nForce3 systems and have never experienced any hdd or usb noise. Then again my company only sells performance cooled computers, could make a diffrence for sound seepage. I would recommend only using a 10K Sata Raptor drive, since they have been around it's all I use. The real world performance gain only relates to about 2K 3dmarks (3dmark2k1), however overall system speed and performance is greatly enhanced, especially if you are going to skimp and not put 1GB or more of RAM, that faster swap file is quite apparent. If your worried about losing HDD space get a DVD-RW drive at less then 70 bucks for a nice one, it only makes sense to store your stuff on a DVD anyway, for the performance boost.

On another note I should mention if you are into gaming the amd64 core does far more than a high end video card. Obviously the fusion of the two is stellar, however if you have to choose get the cheaper $100 video card and focus on the amd64 core (for all you skimpers out there) with at least 1GB of RAM, you will be very happy with the result.

Re:Cheaper Processors

[Anonymous Coward]

You don't even have to build your own for a good deal.

Upon reading through the Microcenter ad I get everyso often, I was astounded at the low prices of relatively decent AMD64 machines (and there were more than a few, and even a couple AMD64 laptops (though I can't imagine why that would be desireable, over other, lower power options)

Re:Cheaper Processors

[gid]

Well I was specing out a "budget system", for power users and myself I would get the full gig of ram.

Actually I bought new ram awhile back, 1 gig of corsair pc3200, so hopefully that should work in an amd64 board just fine. I already have an NV 6800GT, which is really what sparked this whole upgrading mess for myself. :) The video card is incredibly fast, so that I get the same framerate no matter what resolution I run in Doom3. I have a 4x agp Athlon XP 2600+ system now. (I have a feeling the 4x agp is what's really hurting me). Plus the faster cpu would really help with java deveopment.

Thanks for the advice on the HD, I was kind of fretting over that, how much faster the 10k rpm drives are. They are quite a bit more, and it looks like a 74 gig is the max you can get at $175 which is rather pricey. But looking back on it, doing SATA RAID level 1 mirroring with two 7200 rpm drives for increased speed might be better, since I can get two 7200 rpm drives for the price of a 10k rpm drive, plus I'd have some kind of redundancy in case a drive died. I know RAID isn't a backup solution, but it should hopefully keep my system up and running in case a drive dies.

Re:Cheaper Processors

[drinkypoo]

Do yourself a favor and move to digital audio. We're moving to digital video in the form of Digital DVI, you might as well come into the nineties and use S/PDIF audio while we're all heading down the road to the computer of the future.

Re:Cheaper Processors

[GodsMadClown]

Is that why Nvidia decided not to include the realtime digital encoding in their latest chipset? Because digital audio is the wave of the PC future?

Re:Cheaper Processors

[drinkypoo]

It's probably because they see nforce as a budget chipset, where people care more about saving money than anything else. What do you mean by realtime digital encoding anyway, 5.1?

Re:Cheaper Processors

[Ost99]

Thanks for the advice on the HD, I was kind of fretting over that, how much faster the 10k rpm drives are. They are quite a bit more, and it looks like a 74 gig is the max you can get at $175 which is rather pricey. But looking back on it, doing SATA RAID level 1 mirroring with two 7200 rpm drives for increased speed might be better, since I can get two 7200 rpm drives for the price of a 10k rpm drive, plus I'd have some kind of redundancy in case a drive died. I know RAID isn't a backup solution, but it should hopefully keep my system up and running in case a drive dies.

You'd need 4 drives if you want both speedup and redundancy. RAID 0+1 or something.

- Ost

Re:Cheaper Processors

[aka1nas]

It's probably not the 4x AGP, but the AthlonXP 2600+ and it's (relative to the P4 and A64) weak memory bandwith that holding back your performance. If it is a 2600+ you are running is it a 333FSb or 400FSB chip?

Re:Cheaper Processors

[gid]

It's an older thoroughbred cpu with a 133/266mhz FSB... so ya, that's probably it.

Available "Real Soon Now"

[Plural of Mongoose]

Running in 'a lab today' Available 'mid-2005' Uh-huh, like I'm gonna hold my breath - yet again!

64: Intel vs AMD

[minerat]

4 procs, dual cores? Kickass. A short read on implementation differences between AMD64 & Intel's 64. http://www.theinquirer.net/?article=17906

Re:64: Intel vs AMD

[Anonymous Coward]

Two cores? Yaaammmmmmy!

BTW, the article linked is wrong on one thing only. The AMD design offers better throughput than anything Intel is designing, or contemplating to design. The shared bus is still one bus and Intel's contemplated Dempsey well still have one bus per pair of cores. In a single socket system that gives theoreticaly as much throughpur as the Opteron. In practise the Opteron's integrated memory controller gives it the edge. For more than one socket on the board Intel's design is slower because now sveral dual cores fight for one bus.

Recently someone asked one of Sun's Opteron server designers if they are going to use Intel in their macho multiproc servers. He said that while they evaluate everything comming out of Intel, they don't see anything that will match the Opteron.

adverts for intel

[lkcl]

i particularly like the way that on the register, an advert for intel's xeon processor came up on a page describing how great the new dual-processor amd chip is.

Interesting, but realistic?

[Sheetrock]

Given the cothermic limitation on implementing 'cores' (or independent dies) on one surface, it seems a clever but limited hack to increase the performance by effectively implementing multiple CPUs on the same chip.

Clearly there is a performance benefit in both bandwidth and latency respects in multithreading/multioperating in this manner, but it's not difficult to see that the footprint limits the factor to which this technique can be exploited. Indeed even if they were able to fit three cores in the same chip the thermal energy would most likely outstrip the dissipation potential of conventional heatsinks -- unless of course the user is willing to invest in air conditioning or other mainframe-style cooling technologies (which may make sense for servers.)

Re:Interesting, but realistic?

[Fulcrum of Evil]

Given the cothermic limitation on implementing 'cores' (or independent dies) on one surface, it seems a clever but limited hack to increase the performance by effectively implementing multiple CPUs on the same chip.

Of course, in my experience, AMD64s are fairly cool compared to Intel's stuff. You could porbably do a dual core AMD64 at 2Ghz for way under 100W.

Re:Interesting, but realistic?

[Anonymous Coward]

Also, remember that these are 90nm while all shipping AMD64 chips are 130nm. AMD uses silicon on insulator which I believe means they will have much less problems with leakage at 90nm compared to intel...

Re:Interesting, but realistic?

[maynard]

Indeed even if they were able to fit three cores in the same chip the thermal energy would most likely outstrip the dissipation potential of conventional heatsinks[...]

Is that really true? Given the reduction in energy consumption and consequent reduction in heat dissipation with each lithographic step down in size, isn't going multi-core a reasonable way to use the additional footprint and reduced per-core energy consumption offered? I suppose one could argue that a larger on-chip cache is better use wafer real-estate and higher frequencies with the consequent higher energy consumption offers better performance than a slower dual core, but I doubt it would pan out. Larger caches and higher clock frequencies offer diminishing returns due to increased branch prediction failures and memory bus latency. Assuming a multithreaded OS environment, at a certain point a second on-chip core operating at ~80% efficiency is simply going to out-perform a faster single core+larger cache. A good questions is, where's the performance gain threshold between the two in the tradeoff? Would love to read a real chip guy explain this in detail.... --M

Re:Interesting, but realistic?

[flaming-opus]

Tell that to IBM. They package 2 cores on a die, and 4 dies in a multichip module to make up their high-end POWER4 & POWER5 based unix servers. Sun and HP are both putting 2 cores on a die, and hp even puts 2 itaniums on a daughter-board to approximate a dual core ia64 solution. Yes they do have BIG heat-sinks, but these are real servers, not little 1U sleds. Nobody is really worried about fan noise in this setting.

Opteron brings the price point for this down, but they will probably do it at a lower clock speed to fit it into the 1 and 2 U systems that dominate the opteron marketplace.

Re:Interesting, but realistic?

[Anonymous Coward]

AMD is now selling 2GHz 55W Opteron, thus dual core 2GHz Opteron would use 110W at 130nm process. However, dual core chips will use only 90 and thus they will consume less power.

Re:Interesting, but realistic?

[Lonewolf666]

Of course it is limited if you stick to Pentium4-style energy guzzlers (and I'm really curious how Intel will handle the heat from a dual-core Prescott). To a slightly lesser extent, the same applies to the Athlon64.
But if the individual cores are designed for low energy consumption, putting more of them on the die becomes feasible. For instance, four Dothan (the new Pentium M) CPUs would have a total power consumption of about 80 W. Not too much for conventional cooling.
Beyond four cores, I'd expect other problems to show up as well, like the shared bandwith to RAM being not enough for all cores. So a four core-CPU will probably remain the realistic maximum in the next years.

Re:Interesting, but realistic?

[Anonymous Coward]

Weren't there rumors AMD's 90nm Opteron was 105 Watts peak? Make no mistake about it. 90nm is NOT a "cool" process regardless of who makes it. If anything it will force AMD to put more power management in--something sorely lacking on the existing Opteron which has no P-states, for example, and barely any halt/stop grant.

Re:Interesting, but realistic?

[Anonymous Coward]

Since when is 89 Watts peak considered "pretty cool"? I have light bulbs less than that. Just picture how many amps are going through the CPU, and realize an amp is a coulomb/second, and that's a huge amount of current going through such a small device already. It will get worse with 90nm.

At least they gave Athlon64 P-states, but Opteron has none. You are either in halt/stop grant power saving (down to only a few watts!) or running full speed. No happy middle ground.

Re:Interesting, but realistic?

[Christopher Thomas]

Weren't there rumors AMD's 90nm Opteron was 105 Watts peak? Make no mistake about it. 90nm is NOT a "cool" process regardless of who makes it.

Most of the losses for chips like these are dynamic - i.e., caused by switching capacitive loads. A 90nm chip has features with half the area that a 130 nm chip has. Even with thinner layers for some features, this results in lower capacitance, and so less heat generation for the same clock rate.

The key words are "for the same clock rate". These chips are hot because they are run faster, not because of feature size.

Re:Interesting, but realistic?

[Anonymous Coward]

Except you don't have half the area. Because now we're talking dual core again! The area of a modern desktop CPU always hovers in about the same range between "cheap" and "size of your thumbnail".

We've passed the knee of the exponential curve on leakage current. 90nm leakage is worse by a few hundred percent compared to 0.13um. As for dynamic power, don't ever expect wire capacitance to decrease. Even if you have thinner metal layer rules, remember that you've shrunk your area again, so the wires are also closer together and--let's face it--they are as long as ever on average. And then there's the devices themselves, switching ever faster, and also getting more power hungry at 90nm and beyond.

It will take creative design to keep wattage in-check, but I don't expect it to level out. Clock for clock, 130 nm was worse than 180nm, and 90nm is worse than 130nm. That trend will not change.

Re:Interesting, but realistic?

[Anonymous Coward]

Obviously a b00b.

If anything, shrinking transistor size will mean *LESS* power consumed.

Re:Interesting, but realistic?

[Anonymous Coward]

But the opteron does have frequency scaling. ("Cool'n'Quiet" as they like to call it.)

Re:Interesting, but realistic?

[MikeBabcock]

Look up some "real" HSF (heatsink + fan) reviews. The stock stuff from AMD/Intel usually works but isn't that great.

Take a look for example at Zalman's Cu radial cooler, this thing cools high performance AMD or Intel CPUs while still running nearly silent at 1350 RPM.

Re:Interesting, but realistic?

[Anonymous Coward]

Ok I got to cothermic and I had to go to dictionary.com to find it. When it didn't find anything I checked google.com just to be safe. When you have to make up words 3 words into your paragraph you know you are into a bad start. If this was college your professor would give you an F and ask you to rewrite it. Here on slashdot you are Interesting.

Re:Interesting, but realistic?

[Christopher Thomas]

We've passed the knee of the exponential curve on leakage current. 90nm leakage is worse by a few hundred percent compared to 0.13um.

And, at twice the clock speed, dynamic power dissipation is 200% worse. I'm not trying to claim that leakage isn't a problem - just that it tends to not be the dominant problem for high-speed chips. And yes, IAACE :).

As for dynamic power, don't ever expect wire capacitance to decrease. Even if you have thinner metal layer rules, remember that you've shrunk your area again, so the wires are also closer together and--let's face it--they are as long as ever on average.

Only for global routing. Local scales with line width, because you *really* don't want to deal with line delays when it can be avoided. Even long bus lines are segmented, for this reason, though that doesn't alter their capacitance.

It will take creative design to keep wattage in-check, but I don't expect it to level out. Clock for clock, 130 nm was worse than 180nm, and 90nm is worse than 130nm.

What ends up happening is that chip performance become power-bound instead of area-bound. To some extent, this is true already. What I expect to happen is a device technology switch along the lines of the ECL/CMOS switch that happened when ECL ran into the same problem years ago.

In the meantime, there are a number of interesting techniques used to reduce power dissipation. Adiabatic circuits are my personal favourite, though that's more because I think they're nifty than because I think they're a workable solution (current-density limits imposed by electromigration end up limiting your clock speed to something far lower than heating limits it to).

Names?

[Skraut]

Would the Consumer model of these chips be called BiAthlons?

Re:Names?

[Anonymous Coward]

That name has already been used by the US olympic womens beach volleyball team.

Re:Names?

[Minwee]

No, those would be the processors that can interface with both Socket 939 and Socket 940 boards.

Re:Names?

[eli173]

Would the Consumer model of these chips be called BiAthlons?

And of course a small cluster would be a DecAthlon?

(Sorry, couldn't resist. :) )

Re:Names?

[Anonymous Coward]

no. Intel would turn that into ByeAthlons

Re:Names?

[SEE]

Nah, any consumer AMD system from HP is a DECAthlon.

Re:Names?

[scottgfx]

I see this as a case of "Don't ask, Don't tell". :)

Upgrading servers

[Archangel Michael]

So what today might be an eight-way server will potentially become, mid-2005, an "eight socket" server with 16 processing cores.

And just think, it was only last week when it was shown that most servers are never upgraded (Core Components), and that most people already buy their servers with growth in mind.

This kind of stupid comments are not helpful.

My question is this, how is this going to affect M$ licencing of OSes? I buy a dual socketed board and put in a couple of these babies is M$ going to complain that I have 4 CPUs and XP won't load because I have the 2 CPU version?

The idea of licencing software by HW is stupid, don't you think?

Re:Upgrading servers

[Astadar]

I would think that if Windows is going to recognize the 4 x 2-way processors as 8, then yes, your dual proc motherboard with 2 x 2-ways will probably run afoul of any checks that Windows does.

Re:Upgrading servers

[Anonymous Coward]

But also remember that a P4 with Hyperthreading registers as 2 CPU's to Windows as well, and Windows XP Home which is only licensed for a single CPU still recognizes and utilizes hyperthreading properly.

The point of this is: Does the BIOS report one or two physical processors? In the case of hyperthreading it reports a single physical processor (Socket 1) with two virtual processors. Logically in a dual-core setup, the BIOS will report a single socketed chip with two processors and I'm hoping and assuming hypothetically that Windows would treat this similarly to hyperthreading.

Additionally, my assumption is that Microsoft is going to either accomodate the advancement by recognizing it as a single chip, or else will simply do away with a per processor licensing limitation, at least with server versions of thier OS (and have the Pro version allow up to 4 cores and two physical devices, and Home 2 cores on 1 physical device--which with hyperthreading they currently do)

I honestly don't see this as a problem, and I'm quite sure MS isn't going to gouge people by saying NO NO NO you have to buy Windows Server 2003 to run on your Dual Athlon 64^2 (which is what im guessing this will be called...math people leave me alone). If its a workstation class machine with 2 sockets and 4 cores, i can guarantee you that MS will let this go through. They know the workstation market isn't going to run a server OS, and they know that they need to handle it right because both AMD and Intel are going down this road--and they really don't want to tick off the companies and customers of those companies who well...make up the entire market thier products operate on. It doesn't make business sense at all, so I can say they won't be dumb with this.

Re:Upgrading servers

[joib]

I read somewhere a few weeks ago that Oracle, MS and some other big software companies with per-cpu licensing intend to count a hyperthreaded cpu as 1 cpu since it's basically virtual processors. Multi-core cpu:s, OTOH, will be regarded as multiple cpu:s, as there are, duh, multiple cores. They just happen to sit on the same die.

Re:Upgrading servers

[drinkypoo]

Most servers' core components are never upgraded because the upgrade path is usually costly and inefficient. This should provide a way for people to upgrade one processor at a time, and get a substantial performance boost, without spending a metric assload of money on it.

Re:Upgrading servers

[hxnwix]

"And just think, it was only last week when it was shown that most servers are never upgraded (Core Components), and that most people already buy their servers with growth in mind.

This kind of stupid comments are not helpful."

Sun has made headway with their single to dual-core ultrasparc upgrade path. Are they stupid?

Re:Upgrading servers

[Archangel Michael]

The only place where I can see CPU upgrades making sense is in High Density Blade servers.

The Only Speed that Counts: Rate of Market Growth

[reporter]

The only speed that counts is how fast you can grow the market for your product. In that category, AMD wins. AMD appears to be on a roll these days. In the latest quarter, the Opteron (AMD) outshipped the Itanium (Intel) by a ratio of 10 to 1. AMD shipped 60,000 units, and Intel shipped 5665 units. Apparently, the survivors of the microprocessor wars in the 1990s are the PowerPC architecture and the IA32-X64 architecture. The Itanium architecture will survive, but it will be relegated to a high-performance graphics engine.

Of course it is realistic...

[Kjella]

In general, power dissapation scales in frequency with n^2, in multiple cores with n. So for the power of a processor 2x as fast, you could probably deliver 2^2=4x with 4 cores.

Granted, this is only true if the task is parallellizable, but with todays multi-tasking computers I could at least use two cores. (If main task is blocked, there's probably a dozen other background processes who'd like a few cycles).

Kjella

Re:Of course it is realistic...

[Rufus211]

Everything I have ever read (in relation to overclocking) has said that power increases linearly (n) with frequency but quadratically (n*n) with voltage (which implies current).

The real reason to go dual core is that manufacturing processes have been reaching a frequency barrier but arä

Re:Of course it is realistic...

[Christopher Thomas]

In general, power dissapation scales in frequency with n^2, in multiple cores with n.

Power scales linearly with both frequency and the number of cores (or more accurately, with the amount of capacitance being switched per clock). It scales quadratically with _voltage_ (as capacitively stored energy is (1/2)CV^2).

Multi-core chips are used because we have more transistors available on a die, and both increasing cache size and increasing issue width on processors have reached diminishing returns for performance (though SMT-style chips help with the instruction issue problem). If you have twice as many transistors, the best way to improve performance nowadays is to build two cores on the die.

Re:Of course it is realistic...

[non]

at work i've used a p4 2.6, and i'm currently using a single cpu g5. the experience is nowhere near as pleasant as using my old dual p3 at home.

Four of parts?

[Canthros]

I know, don't pick on the lack of grammar on Slashdot, lest ye be struck down by Great Powers On High. I just can't help but wonder if that's a minor arcanum/suit for some sort of geek tarot or playing card deck.

Benefits of dual core?

[timeOday]

I wonder if somebody could explain why dual-core CPUs are a good idea. If it's a pair of cores on a single piece of silicon, it seems it would take the same silicon as two separate cpus, so where's the benefit? You'd save the cost of an extra socket on the motherboard, but then again yield decreases roughly exponentially with die size, which argues for 2 separate cpus.

Re:Benefits of dual core?

[Anonymous Coward]

They are probably two independant cores manufactured seperately, they are just combined in the die. So if any given core fails along the way, you aren't discarding a second perfectly good core. Unless they fail when manufacturing adds them to the die.

Re:Benefits of dual core?

[Tranzig]

If you'd read the article you would know that a stock 4-socket Opteron mainboard used, turned into a 8 CPU system. Such simple (cost effective?) upgradability is a very good reason in my optinion.
And if you take into account that both Intel and AMD decided to go for dual-core, it might be the most logical way to improve the performance of the chips.

Re:Benefits of dual core?

[mr_walrus]

and quiet(er) too!
two sockets == two noisy fans
1 socket (dual core) = one noisy fan.

smp scaling without the noisy fan scaling is a good thing!

Re:Benefits of dual core?

[Dan Ost]

I'm sure there's more to it than just this, but part of the appeal of dual-core
CPUs is that I can double the processing power of an existing machine without
having to upgrade the motherboard if the motherboard already supports the
correct socket.

Also, it means that smaller form factor machines can have more processing
power.

Re:Benefits of dual core?

[Anonymous Coward]

As said, if one of the cores is messed up, it's simply disabled and the thing is sold as a plain single core processor. There is no reduction in yields. It's done all the time with cache. The price of production is only very slightly increased. Don't worry. Enjoy the speed. It's good.

Re:Benefits of dual core?

[TimSee]

This article explains it rather well: http://arstechnica.com/cpu/004/prescott-future/pre scott-2.html

Re:Benefits of dual core?

[alphorn]

Dual core makes economic sense. There is a sweet spot for die size (around 100mm squared), below which the production costs start becoming negligible. But with technology improvements, 100mm allow you to fit more and more transistors, and dual core is what gets you the maximum processing power (and therefore money) out of those extra transistors.

Re:Benefits of dual core?

[joib]

They are probably two independant cores manufactured seperately, they are just combined in the die. So if any given core fails along the way, you aren't discarding a second perfectly good core. Unless they fail when manufacturing adds them to the die.


No, they are manufactured together. If one core happens to be defective, it is disabled and the package is sold as a 1 core processor. Same as is commonly done with cache today.

IBM also a few years ago sold a "HPC Regatta" p690 system with 1-core power4 processors instead of the normal 2-core ones (i.e. a maximum of 16 cores instead of 32 as on the normal p690). I guess they had defective second cores, instead of IBM designing a special 1-core version of the power4 just for the HPC Regatta.

Re:Benefits of dual core?

[Christopher Thomas]

I wonder if somebody could explain why dual-core CPUs are a good idea. If it's a pair of cores on a single piece of silicon, it seems it would take the same silicon as two separate cpus, so where's the benefit?

Less packaging overhead, and faster communication between cores (on-die bandwidth and latency are far, far better than any motherboard's crossbar's bandwidth and latency).

You also have less contention over memory, for single-chip systems with multiple cores vs. multi-chip systems. Instead of having to muck about with cache coherence across a bus, the chip looks like a single processor as far as the memory subsystem is concerned, with coherence operations only involving the first one or two cache levels on-die.

yield decreases roughly exponentially with die size, which argues for 2 separate cpus.

Processes are optimized so that you can build a chip with 1-2 square centimetres of area with reasonable yield (as this is what chip manufacturers demand). This has been pretty constant (or if anything, has been increasing). However, with each design generation, the number of transistors available in this area has doubled. We're now at the point where we can get high yields on chips with enough transistors that multi-core designs make sense.

A chip with N cores also doesn't take N times as much area as a single-core chip, as the lowest levels of cache aren't duplicated (just L1 and usually now L2). So overhead is reasonable, and the real estate is there. It makes a lot of sense to use it.

Re:Benefits of dual core?

[Anonymous Coward]

Because it's a reality check.

Everyone is talking dual core now because they can't talk 5-6 GHz for one CPU now. And the reality is AMD chips are very VERY corespeed limited, not memory bandwidth limited. That's why a dual channel benchmark (Opteron) is only a few percent faster than a single channel benchmark (Athlon64 754 socket). I'm sure if it was easy for AMD to crank up the megahertz it wouldn't have taken years to reach 2.4 GHz. I'd still rather have a 5 GHz Opteron for MAME, though. Becaue the other reality check is not everything can benefit from multithreading!

Re:Benefits of dual core?

[graveyhead]

If it's a pair of cores on a single piece of silicon, it seems it would take the same silicon as two separate cpus, so where's the benefit?

That's just it - it doesn't. Most of the physical space on a chip is taken up by the local cache. If 2 cores can share the same cache, but keep their state separated, voila, price/performance sweet spot... aka $avings.

I bet we even see processors with more than 2 cores in the near future.

Re:Benefits of dual core?

[Anonymous Coward]

A Dual Core is a good idea. Its really nothing more than splitting pipelines and making the halfs run independantly. It dosent benefit workstations, but does great for multi-tasking.
If one pipeline stalls, it only flushes half the pipeline, instead of flushing the whole shebang.

They need to start calling them split-pipelines, instead of dual-core.

Re:Benefits of dual core?

[vincecate]

The chip only has one crossbar, one memory controller, one set of 3 hypertransports, and one set of pins/package. Only the caches and the heart of the CPU are duplicated. So it is much less than double the size.

The other big win is now one motherboard with 4 sockets handles 8 cores. So you buy half the motherboards, half the racks, and pay half the co-location fee.

Re:Benefits of dual core?

[Chris Burke]

That's not correct. What you're describing may be Intel's Hyperthreading, a type of SMT where multiple threads can execute on the same functional units, which as you say requires only flushing or stalling one thread on a mispredict or long latency op.

What AMD has is really dual-core. There are two complete K8 cores on the die. Everything up to and including the L2 cache is duplicated. The cores communicate through the crossbar in the on-chip northbridge. It's an SMP system on a single chip, or CMP as the TLB (three letter buzzword) goes. It isn't a split pipeline, as each core has the same complement of resources as a single-core K8.

Re:Benefits of dual core?

[ottffssent]

It's not double the hardware. That's benefit 1. Also, it's massively more closely coupled than a DP system.

AMD's dual-core chips have two separate L1 caches, two separate L2 caches, etc. However, they share the HT hardware and memory controller, etc.

The caches are ~50% of die area, the cores maybe 30% and the other 20% is memory controller, HT controller(s), and other miscellaneous hardware. So, a dual-core processor performs better than a DP system and uses 180% of the die area of a UP system rather than 200%. The savings is slight on the CPU, but it also saves a CPU socket, motherboard traces, power circuitry, and on and on and on.

In fact, the only disadvantage is that you now have two cores to feed with 1 memory controller rather than 2 cores with 2 memory controllers, and I'd be willing to bet that AMD's memory controller has some headroom.

So, while a dual-core CPU will probably cost about the same as 2 separate cores (more expensive die offset by lower packaging and testing costs), the systems they go into can be a lot cheaper for the same performance. Also, this lets you pack more performance into a given area, and do it cheaply too, which is something AMD have always been good at.

Dammit, AMD -- quit inventing shit so fast!

[IronChefMorimoto]

Christ in handbag covered in ketchup, AMD! WHY THE HELL CAN YOU NOT JUST TAKE A BREAK AND STOP INVENTING SHIT FOR A WEEK OR MONTH?!

It's not like AMD fanboys like me are going to let you go out of business. We'll still be buying your underpriced processors in lieu of Intel chips for a while to come. And we'll show up in droves to events that really tout your existing product line. We swear it!

Plus, Intel isn't moving that fast these days. I've read more about trouble for Intel in the past 2 months than I have in 5 years. "We can't frabricate this processor, or we're not responsible if that processor burns your house down when you overclock it." Come on! Let 'em catch up for the sheer thrill of beating 'em again with the Athlon128 a few years down the road.

Why? Why my insistence on your taking a g'damned break from inventing shit? You wanna know why?

I can't f---ing afford another upgrade for awhile. So stop it. Now. Dammit. Give us poor home built computer bastards a break.

And pass the f---ing message off to those asshats at ATI and Nvidia.

John Carmack too -- the "we're gonna change the world of gaming hardware every time we release a new game" motherf---er.

IronChefMorimoto

Re:Dammit, AMD -- quit inventing shit so fast!

[ciroknight]

That, good sir, was the best /. post ever.

Re:Dammit, AMD -- quit inventing shit so fast!

[biz0r]

Am I the only one who can almost hear Cartman screaming this at the top of his lungs?

Re:Dammit, AMD -- quit inventing shit so fast!

[Anonymous Coward]

F---ing hilarious, IronChefMorimoto!

BTW, why do you use fucking hyphens?

Re:Dammit, AMD -- quit inventing shit so fast!

[Hassman]

This is really funny. :)

But just a bit of information. Intel is coming out with a similar chip with either 4 or 8 (I can't remember now) processors on the same chip. Then when compaines order them they can designate how many processors are turned on.

Pretty cool stuff.

Re:Dammit, AMD -- quit inventing shit so fast!

[Methuseus]

What do you mean almost......?

Re:Dammit, AMD -- quit inventing shit so fast!

[Cornelius the Great]

Damn, IronChefMorimoto... that's two +5, Funny posts in a day!

Mind if I befriend you?

Re:Interesting, but realistic? - Yes...

[CodeMaster]

Clearly there is a performance benefit in both bandwidth and latency respects in multithreading/multioperating in this manner, but it's not difficult to see that the footprint limits the factor to which this technique can be exploited

Actually - if you have two cores on the same die you can minimize the needed bus transport path and use processor scale path => less heat... you still need the same components to provide the bus external to the two processors, but the speed gains from having a dual core should not have an impact on the heat dissipation other than just having two cores to cool down (and with modern HSF technology that is not a problem - If I can cool down a P4 3Ghz with a quiet HSF combo - AMD can do it too...)

Get your free iPod![it really works! - my buddy got his after I signed up, I have just 2 more referrals to go...]

Re:The Only Speed that Counts: Rate of Market Grow

[Dan Ost]

How does Opteron shipments compare to Xeon shipments?
Or, more importantly, how fast is the Opteron market growing compared to
the Xeon's?

Backwards compatible, too!

[hirschma]

This is the hottest part. It means that I can take my current Operton dual CPU machine and make it into a 4-way, likely with just a BIOS upgrade.

I think that a lot of folks are going to go for this type of upgrade, just because the upside is so huge.

Re:Backwards compatible, too!

[JollyFinn]

No no, you got it totally wrong it takes bios upgrade and 2000$ , or perhaps just 1000$ for lower end models...

Re:Backwards compatible, too!

[Anonymous Coward]

Support for more processors is done in the OS not in BIOS.

Re:Fantastic

[boristdog]

The server is sitting in the room next to me right now. (Guess where I work?) Hardly what I would call vaporware.

8-way is nice...

[Kjella]

...but I'm hoping for a more reasonably priced socket 939 CPU to go with a Shutle SN95. 3500+ carries a nasty premium, no new processors on that platform since June 1st, where's Winchester (90mm Athlons) and some more mainstream procs, 3000+ would do just fine for me...

Or maybe I'm just antsy because my main PC is unstable (mysteriously so for the last 2-3 months, can't find anything wrong in hardware, nor software) and I'm itching for something new... sigh. Computer addiction should be qualified as an illness, always wanting what's in three months can't be right.

Kjella

Re:8-way is nice...

[WuphonsReach]

Or maybe I'm just antsy because my main PC is unstable (mysteriously so for the last 2-3 months, can't find anything wrong in hardware, nor software) and I'm itching for something new... sigh.

Fire up the Prime95 client from Mersenne.org - it's not a hardware test program, but it may as well be for as often as it will stumble over a hardware issue. Even hardware issues that Memtest86 / Memtest86+ miss. Overclockers have been using it for years to figure out if their systems are stable.

A close runner-up, in terms of programs that are extremely sensitive to mis-timings or other glitches is QuickPar. Tougher to find errors with that one, unless you try to create a recovery set for 20GB worth of data.

(My story ended by throttling back the CL value on my memory... the CL 2.0 memory that I had wasn't stable until I dialed back to CL 2.5.)

And in other news...

[ardustry]

...Intel has released it's newest chip, the Cho-Ken Chi-Ken III, a 16 Core, 22 Hyperthread, 324mb L1 and L2 Cache, with Artifical Intelligence Pipelining and Branch Prediction, which prevents premature ejaculation of code segments from data registers. This chip will revolutionize gaming (again?) and push Intel's Customer Relations further into the Twilight Zone.

Bad terminology

[Anonymous Coward]

Socket implies the physical chip. An 8-socket system using a dual core chip implies 16 processors. The poster really meant a 4-socket dual core system.

Why don't they.....

[Anonymous Coward]

research ways of

1) Speeding up hard drives
2) Getting data between the HD, memory and processor faster

Surely this would improve the overall computers speed rather than just the processor.

Re:Why don't they.....

[cybrthng]

Look up HyperTransport and the reason why the memory controller is onboard on opterons and you will see that they're working on this.

On the flipside, HD/Video card companies make money off SLOWING DOWN technology so people get hooked and play the upgrade game.

Companies don't make money off selling something that lasts 5 years anymore... they make money off selling something you buy every year. Thanks to consumer demand. We don't want to pay 500 bucks for a video card that lasts 5 years, we want to upgrade every year with a video card that costs 100 bucks.

Longrun price is the same, but in the short term consumers think they're getting a better deal and for ultra-consumers it feeds the addiction.

Re:Why don't they.....

[Hassman]

It is more complicated than that.

We're talking about servers here as opposed to desktops. In a desktop, absolutely, lets get those hard drives faster, and get rid of some of the latency. A personal computer is constantly reading / writing depending on what the user is doing.

But for servers? It is more about handling thousands of requests at the same time, as opposed to accessing the hard drive. Generally servers have gigs of memory, so it becomes less important on how fast they access the memory as it does on making sure the needed information IS in memory.

If the CPU can pre-empt or predict what to put in memory before it is needed, there is a HUGE bonus on speed. Plus, why would AMD research hard Drives? They are a CPU maker.

Re:Why don't they.....

[drinkypoo]

You can't make a video card for $500 that will do today what a $100 video card will do in five years. It makes more sense for everyone to buy a $100 video card every year, whether you personally like it or not - especially since you can skip a year if you don't need more power just then. Consumers ARE getting a better deal, because they can upgrade more frequently if they choose to do so.

Re:Why don't they.....

[cybrthng]

Video cards were just one aspect of "consumeristic" computing that i felt like citing. Geforce 3 cards are an excellent proof to my point that well designed and technologically advanced ahrdware can outlast the latest "consumer" versions that only boast minor improvements.

If you want another example - figrue this. I bought an Ultra160 SCSI rig with 10k rpms for 700 bucks 3 years ago. Say you have bought an ATA 66, ATA 133 and now an SATA rig over those last 3 years. The "consumer" technology is still slower than the SCSI technology as well as over those years of upgrades the "long run" figurative cost is cheaper than the short term "consumeristic" costs.

Not too mention its taken years for people to figure in scalable IDE solutions and even now firewire and USB technologies are gathering up steam to compete.

However - you can by a shared bandwidth USB device and upgrade and upgrade and upgrade and move to usb 2.0 and only see a franction of the performance and value that a firewire setup would have cost you. Again - while firewire was a larger initial cost its technology has outlasted the "consumer usb device" simple because it was engineered over time for being a serial interface to high speed devices.

ofcourse firewire "b" is out and you can uprgade as with any technology advancement - my point is that if you chose firewire vs usb 1.0 when they both were announced years ago the firewire technology at "a" release is comparible to the many revisions of the USB "consumer" technology and the cost is moot over time.

Just choose a technology company and not a consumer company to be your vendor - while you may not have the latest and greatest specint scores your overall capacity, performance and useabaility over the time will be better, cheaper and more reliable.

Re:Why don't they.....

[drinkypoo]

If you want another example - figrue this. I bought an Ultra160 SCSI rig with 10k rpms for 700 bucks 3 years ago. Say you have bought an ATA 66, ATA 133 and now an SATA rig over those last 3 years. The "consumer" technology is still slower than the SCSI technology

The interesting thing about that is that I paid about $500 for a couple of 18.2GB Ultra160 7200RPM/2MB Cache drives and a used (and possibly "hot") 29160N and then later I traded all of that shit for an 80GB EIDE, and bought another 80GB EIDE, and put them together on a hardware RAID controller on my motherboard. My performance is about the same (although I'm striping so my reliability is half or less of what it was) and I have 150GiB of usable space instead of about 32GiB.

The moral of the story is that what you buy should suit your needs and be as inexpensive as possible. Video cards lend themselves to frequent replacement and purchasing a last-generation card or even one two generations (or more!) old for your casual gamer or non-gamer. The reality is that this is true of all computer hardware at the current rate of advancement.

Why Austin?

[Epi-man]

What I don't understand is why are they doing this in Austin? Last I heard Fab25 was down to just producing flash on a 0.18 um process. Why are they not doing this in Dresden or Sunnyvale?

Re:Why Austin?

[boristdog]

CEO lives in Austin.

Re:Why Austin?

[Anonymous Coward]

HP/Compaq OWNS the server market.

Well maybe some folks like me run Dell, but only because they're much cheaper used.
Besides, how much computer do you need to stuff a 90MBPS circuit with php and bittorrent?

In any case, the big bux is in the higher priced servers. As long as the corporate folks insist on running Mr Softie single user OS, requiring minions of button pushers and sustaining the corporate to IT-wannabe welfare system (albeit a volunteer through ignorance based system) there will be a market for loads of these HP/Compaq 'server class' machines. I'm GLAD to see Compaq/HP running with AMD. Why the hell then is the AMD stock sitting where it is? Does someone know the test results will prove disappointing? I don't think so. Compaq/HP would not let this happen with their name glued on the box! Suggest you guys jump on some AMD stock before it doubles! Missed google's offering, big deal, bet they will be rnning the new AMD's!

Re:Why Austin?

[mwg_stpaul]

HP is in Austin, not AMD. It's the HP factory he's talking about.

Re:Why Austin?

[Anonymous Coward]

Cause Austin R0xx0rz j00r 4$$!
Don't forget where the IC was invented.

Re:Why Austin?

[Anonymous Coward]

More importantly though, The bulk of AMD Marketing is located in Austin. A lot easier to ship 4 chips from Dresden to Austin than to ship a bunch of suits the other way.

Re:Why Austin?

[Anonymous Coward]

Because I live here...Duh!

Re:Why Austin?

[Epi-man]

From the article:

but AMD will be have four of the parts running inside a usually four-way HP ProLiant DL585 server at its Austin plant later today.

That certainly sounds like it is AMD's plant since they are the subject for the sentence.

Re:Why Austin?

[Anonymous Coward]

Why Austin? Because the people who DESIGN the chips work in Austin. The big fabs are elsewhere, but the big fabs are for mass-producing stuff once the design is finished.

Oh, by the way, a lot of other chip design is done in Austin as well. The now-defunct Somerset Design Center (the PowerPC consortium thing between Motorola, IBM, and Apple) was here in Austin. Intel does a bunch of chip design here too, and IBM has a big center that I believe does some chip design as well.

Re:Why Austin?

[Anonymous Coward]

HP is in Austin, not AMD. It's the HP factory he's talking about.

Nope, there is definitely an AMD presence in Austin. I live in Austin, and my Linux machine is running an Athlon that I got for free from a friend who used to work at AMD, because they had spare Athlons just lying around. Granted, it's a 1.3 Ghz, but it was free...

Itanium - high-performance graphics engine

[Tanktalus]

Hmmm .. a graphics card with an Itanium core? That might be pretty sweet ...

Re:Itanium - high-performance graphics engine

[norton_I]

It has been done before. HP's visualize workstations used to ship with graphics cards that were build around the FPU core of their PA-RISC chips.

Re:Itanium - high-performance graphics engine

[bhtooefr]

Also, IIRC, the i960 was a VERY popular GPU long after it died as a CPU. (It was also a pretty popular embedded chip)

Re:Itanium - high-performance graphics engine

[HuguesT]

Correct, the NeXTDimension expansion card had one of them to run the 32-bit Display Postscript stuff.

Wait, 8 Sockets?

[Myriad]

"AMD will be have four of the parts running inside a usually four-way HP ProLiant DL585 server at its Austin plant later today."

Wait... four-way with dual core processors... so what they are saying is

THERE ARE FOUR SOCKETS!

Blockwars: free, multiplayer Tetris like game

Re:Dammit, AMD -- quit inventing so fast!

[SenseiLeNoir]

ROFL!

Seriously, its soo great that we have an "Underdog" performing so well! I used to be an all intel person, until the AMD Thunderbird came out, since then, i cannot be convinced to buy a Intel, even if my life depends on it..

Re:Fantastic

[Tanktalus]

If you work at AMD, then it's still vaporware. Anything pre-announced is vaporware until it escapes the lab and into consumers' hands.

Re:Imagine ...

[RsG]

Imagine... a /. story where beowulf clusters are on topic! Imagine a cluser of cluster computing stories! Imagine... wait, what do you mean "recursive loo[NO CARRIER]

and it could be silent too!

[mikieboy]

those nice people at zalmantech.com have created a completely silent case for a 4way opteron so you can have power and silence for the correct amount of $$ http://itpro.nikkeibp.co.jp/free/NC/NEWS/20040716/ 147359/

MOD PARENT UP

[Anonymous Coward]

MOD PARENT UP
He is not a fucking moron like the rest of you who cannot read the FUCKING ARTICLE before complaining about shit.

Re:Fantastic

[boristdog]

Nope, I don't work for AMD.

hot because they are run faster, not because of fe

[dpilot]

Bzzzt. 90nM appears to be at the turning point where leakage becomes a significant part of total power. In prior generations leakage was only significant with clock reduced or stopped, but negligible compared to active power. It's still much lower than active power, but no longer negligible. Try a simple search on the terms "prescott", "leakage", and "power" for a little flavor.

Intel Demoed Dead Wafer of Dual-Core Itanics

[vincecate]

Note that Intel demoed a wafer of dead Itanics. So Intel did not get working ones on the first try, which AMD seems to have.

Somehow a Slashdot thread on Itanium and Opteron did not get into the Intel section.

where are the 64 bit drivers, HP?

[Anonymous Coward]

HP having Opteron Servers?
Who the fuck cares?
Certainly not HP support.
We had a demp unit on loan.
No 64 bit drivers available for the integrated RAID controller.

all we got to test was 32 bit software.

fuck em.

Cache coherency implications

[Geiger581]

From what has been published prior, the maximum number of coherent HyperTransport links in one Socket 940 interface is 3 and the number of logical processors has been limited to 8 to keep cache snooping traffic managable. Because each dual core chip will have 2 independent caches, the coherency traffic will increase regardless of whether external dual cores are addressed as single HT units. Will this result in either: a) reduction of sockets for general-purpose servers to 4 or b) entirely new ccNUMA protocols being developed from previous generation Opterons?

OS loaders and schedulers can help keep chatty processes allocated to the right mem/processor, but something more has to be said about hardware-level coherency standards. The X-box was fast and efficient largely because its CPU used the video RAM natively, but PCs still have to slog data over the slow and non-coherent PCI, AGP, or PCI-Express busses between the CPUs and GPUs. An inter-vendor standard could bring PC CPU-GPU interaction efficiencies much higher. ccPCI-Express or HyperTransportx16 slots anyone?

Re:Cache coherency implications

[Shadow99_1]

I've seen 16 socket Opteron boards, though no one seems to make one... So 8 is not the limit... In fact I've heard of a single board 32 cpu solution...

Also Hypertransport is not designed for connecting devices... It has no slot orc socket interface, it's used to conenct chips... Also PCI-E is already out and on soem boards, but with less than stellar effects so far...

Re:Cache coherency implications

[Geiger581]

True, there is not a hard limit of 8 cores for an Opteron system, but that limit does seem to apply to -glueless- systems. I have not yet seen an Opteron system with greater than 8 cores without extra logic processors for inter-chip communication. The recently announced Newisis systems (16 and 32 core) use clusters of 4 CPUs. Furthermore, the AMD presentation for multi-core Opterons only mentions 2- and 4- way systems. This is not good news for AMD.

And yes, I do admit the HT is not a physical spec nor does the committee plan to develop one. I just wish that they would. It makes no sense to make a socket interface spec since CPUs have different power pin needs (and number of HT connections for that matter) and chipsets have an arbitrary number of pins to peripherals. However, a slot interface would make perfect sense if you wanted to treat select devices like GPUs as peers to the CPU, cutting latency and increasing effective bandwidth to them. PCI-Express was designed explicitly to cut the number of pins/traces needed for a device, and in that regard, HT is not quite as efficient.

Re:Fantastic

[Tanktalus]

My next guess would be Intel. :-)

I'll be at the dual core demo

[ruiner5000]

Look for a report this afternoon on AMDZone.com.

Re:I'll be at the dual core demo

[ruiner5000]

Well you can see pictures of all three Opteron cores up from the demo. single core 130nm, dual core 90nm, and single core 90nm. I have some info including the two demo boxes. One with Linux and one with Windows. I'll have more extensive coverage up later with system pics and additional information.

Intel has nothing to throw at this?

[Tibor the Hun]

Whatever do you mean?
Let me count:
-They have speedy Celerons,
-They have shiny 32bit HT P4s
-They have shiny blue and orange stickers that say "You just paid too much for a CPU... err, I mean Intel Inside!"

IBM . . .

[cgh4be]

IBM has been doing this for several years, now, starting with their Power4 and now Power5 chips. Granted, it's not exactly the same market (x86 vs PPC), but they will run Linux and *BSD, and Windows pretty soon as the rumor goes.

You're a buffoon and karma whore.

[Ayanami Rei]

Hur hur, it doesn't scale Look ma I can makeup long fancy words!

Please.

Re:hot because they are run faster, not because of

[drinkypoo]

Do a google search on "prescott SOI". Know why none of the top hits seem to be relevant? Because prescott is using strained silicon and avoiding SOI. SOI reduces capacitance, reducing switching time, and improving efficiency. Apples, Oranges.

They won't

[Anonymous Coward]

The dual core Intel chips will be based on the Pentium Ms. Pentium4s are dead end tech.

English Please

[Zebbers]

I don't tend to be a grammar Nazi because as long as I get the idea of the post I tend to ignore it, especially on slashdot.

This shit, however, needs to stop. What the fuck do all these 'employees' do all day? How hard is it to read the submission and realize "FOUR OF PARTS" doesn't sound right?

I would have subscribed awhile ago, and continued contributing but not with this kind of crap. Slashdot is on top the same way MS is, mindshare and sheer numbers. They don't do anything better than anyone else these days.

Re:Gmail

[CoyoteGuy]

e d i t o r _ g a @ h o t m a i l. c o m and id appreciate it greatly!!!

Re:hot because they are run faster, not because of

[dpilot]

I know Prescott isn't in SOI. But that's a capacitive issue, and we were talking about leakage. Peaches, Pears.

Within the realm of SOI there may be some things that can also be done to reduce leakage. Since as you say, the capacitance is lower, you elect to lengthen the channel slightly and give up a little performance, just to reduce leakage. Also in SOI you can modulate the body under the channel, and play some leakage games that way.

In either technology you can sprinkle your shortest channels into the critical path, and back off elsewhere, again reducing leakage. I know AMD has done this in the past, sprinkling 'next generation' channel lengths into later revisions of the current generation. On Usenet they get criticized for 'not really being XXnM technology,' but for anyone with real experience, you know enough to control your design aggressiveness, and sprinkling can be a Good Thing.

Re:Fantastic

[Anonymous Coward]

The server is sitting in the room next to me right now. (Guess where I work?)
McDonalds?
Burger King?
Since you're not the server, I'll guess that you're the one flipping the burgers.

Re:hot because they are run faster, not because of

[Christopher Thomas]

Bzzzt. 90nM appears to be at the turning point where leakage becomes a significant part of total power.

There is a large difference between "significant" and "dominant". I am not disputing that it is a serious concern. However, it doesn't alter the validity of my response to the grandparent post.

Re:Cheaper Processors (Opteron vs AMD64)

[WuphonsReach]

How do the Opteron CPUs stack up against the AMD64 CPUs? (I want ECC memory!)

Is a 2.0Ghz Opteron roughly equivalent to a 2.0Ghz AMD64?

(I'm probably going to go with the Asus SK8V, VIA K8T800 chipset, because I've heard that PunkBuster doesn't work with the NForce3 motherboards at the moment.)

The 2.0Ghz Opteron 146 chip seems to have vanished from the channel. Now you either have to go with the 144 chip for $230 or the 148 chip for $445. Pity, because the 146 chip was a decent trade-off between price/speed.

The Registers Grammar

[Anonymous Coward]

"AMD will be have four"

We be havin fun, believing a story that does not follow simple grammar rules.

This sentence will be have two verbs.

BUT, AMD Rockon with your bad self. I cant wait to see them top the 64-bit 4 way.

Re:Cheaper Processors (Opteron vs AMD64)

[Anonymous Coward]

The Opteron (940, and FX 939) stack up to between a 6% and 10% performance gain over a normal AMD64 (socket 754). This is based on the dual channel memory controller and the extra on die cache. As far as ECC memory I see some benifit but for the price, I shy away from it. cas2.0 mushkin! ddr ram sticks are quite pricey (good luck finding non-mushkin cas2.0 ecc ddr) and you lose performance when you go to a cas2.5 or cas3.0 stick, this diffrence is often apparent to the user.

You can see much higher speeds for far less price with normal unbuffered DDR. As far as errors and stability the AMD64 design is so good I have yet to see the major stability benifit from ECC ram, unless you are planning on serving many users I'd save the cash and get a socket 754, 2800+ and a MSI kt8 Neo Platinum which totals about 300 bucks right now. An on die memory controller like the AMD64 really seems to solve a lot of stability issues as long as the CPU is well cooled. I generally use a nice screw mounted swiftech cpu cooler, with a quiet 120MM fan mounted, or a 92mm "silent cat" thermaltake. The only memory error issues I have seen are related to CPU overheating regardless of Opteron/AMD64 or ecc/non-ecc, and those issues are related to the on die memory controller not clocking correctly or overheating due to overclock or poor cooling solutions.

Re:The Only Speed that Counts: Rate of Market Grow

[Loki_1929]

"The Itanium architecture will survive, but it will be relegated to a high-performance graphics engine."

Funny, I said that would be the case some time last year. I was laughed at for saying that.

Re:hot because they are run faster, not because of

[dpilot]

Well, I guess Intel's recent habit of clocking at Ludicrous Speed does have some effect on the issue.

It's just that for decades you had active power and standby power, and CV**2 and leakage. The formers fit together, and so did the latters, and they DIDN'T interfere with each other. Maybe leakage doesn't dominate active power, but to even have it become significant was a rude change.

I know all about you.

[Anonymous Coward]

You have almost no idea what you're talking about. It might be a good idea for you to simply accept your fate as a mediocre sysadmin, a profession any number of teenagers and grease monkeys could have just as easily. I will repeat: you do not know what you are talking about. Shut the fuck up, Gelinas, your stupid pulp novels about economic history don't even relate to the shit you (try to) explain. I almost want to fucking puke every time I see you post this stupid crap, stroking your goatee and twiddling your pony tail, because deep down inside you desire to have sex with the men you pretend to intellectualise with. Unfortunately for yourself, I'm addicted to hating you, J. Maynard Gelinas of MIT's Nuclear Laboratory system administration department (lol, what a phoney title, ape).

Re:The Only Speed that Counts: Rate of Market Grow

[Veridium]

Speaking only of the server segment, IDC reported that Opteron based server growth was 2183 percent over last year. Xeons was 10 percent.

But growth rate, however, really isn't a fair comparison. Xeon has dominated the market for several years in the x86 server space, so when a successful newcomer comes on board, of course the newcomer will have more rapid growth AT FIRST. Unfortunately, I don't have the specific numbers of units shipped, but of course Xeon will have insanely more shipped.

Here's the Itanium/Opteron SPECfp numbers..

[Veridium]

http://www.digit-life.com/articles2/amd-hammer-fam ily/

Basically, 2.0 ghz Opteron SPECfp peak 1170
1 ghz Itanium 2 SPECfp peak 1356.

Re:The Only Speed that Counts: Rate of Market Grow

[Hoser McMoose]

There are somewhere in the order of 1.4M Xeon units shipped in the same time-frame that AMD shipped 60,000 Opteron servers. Strong growth rate from previous servers, but still less than 5% of the market.

AMD Now Wins Floating Point Race

[Nom du Keyboard]

AMD now wins the floating point race against Itantium...

...on a per chip level. A dual core Opteron beats an Itantium Chip.

And you thought I couldn't make my case, didn't you?

Re:AMD Now Wins Floating Point Race

[Anonymous Coward]

Except Intel have dual core Itaniums at about the same stage as well.

Re:AMD Now Wins Floating Point Race

[vincecate]

At least some people expect Intel will demo a dual-core Itanium or Xeon at the Intel Developer Forum next week, though so far they have only shown dead dual-cores on a wafer.

But a dual-core Itanium with 24 MB of on chip cache is much harder to make than an Opteron with 2 MB of cache. AMD will pass through 10,000 chips/month well ahead of Intel.

Re:The Only Speed that Counts: Rate of Market Grow

[Miguelito]

Yeah, but you're talking 32bit chips vs 64bit chips.

The next couple of quarters, comparing shipments of the Intel EM64T chips vs opterons will be very interesting, and very telling.

I finally got my hands on a nacona a couple weeks ago and put SLES9 on it, and compared it to some opterons. At first I thought the nacona just edged out the opterons when running the binaries of openssl (I like to use openssl for some speed tests) but earlier today realized I did the opteron tests on hosts with a 2.4 kernel. So I redid the tests and found that the opterons still beat the nacona.

What's really sad is that the nacona is a 3.4GHz chip, and the 2.2Ghz opteron is able to beat it. If I recompile openssl myself, opterons get another boost.. much larger then the tiny boost the nacona's get (which just shows that AMD working with the gcc maintainers was a good thing).

I can't wait to see comparisons of large memory programs too.. to compare the AMD design on the memory controller on the CPU. Especially when in larger systems with multi-cpus where the Hypertransport will help even more.

Sun Opteron Server

[Anonymous Coward]

Best part it Sun is working on an 8-way Opteron server. Imagine if they ran a dual core Opteron in one of those!

Goodbye Itanium