Official Opteron CPU2000 scores

[Random Variable]

Einux A4800 (Opteron 144)

Integer - 1095/1170 (base/peak)
Floating-Point - 1122/1219

Einux A4800 (Opteron 142)

Integer - 991/1053
Floating-Point - 1029/1120

Einux A4800 (Opteron 140)

Integer - 880/933
Floating-Point - 934/1012

ASUS A7N8X Deluxe Motherboard, AMD Athlon (TM) XP 3000+

Integer - 960/995
Floating-Point - 776/869

Intel D875PBZ motherboard (3.0 GHz, Pentium 4 processor with HT Technology)

Integer - 1164/1200
Floating-Point - 1201/1207


http://www.spec.org/cpu2000/results/cpu2000.html

 

[Sunner]

Quite impressive, I think that's the first time I've ever seen a top AMD CPU score higher than a top Intel CPU, especially in SpecFP, if only Peak.
Of course Im not counting Itanium, but they won't be competing, so I think that's fair enough.

 

[Random Variable]

Of course Im not counting Itanium, but they won't be competing, so I think that's fair enough.

I read that preliminary tests have indicated that the 1.5GHz Itanium 2 will score over 2000 on the CFP2000 benchmark.

 

[Eug]

Hmmmm... Sounds like the Opteron 142 is ballpark IBM PPC970 2.0 GHz territory, but that won't be out for a few months yet.

(PPC970 @ 1.8 GHz is: SPECint2000 - 937, SPECfp2000 - 1051, so extrapolating, the 2.0 should be around 1041 and 1168)

 

[Random Variable]

The 1.7GHz Power4+ (in a IBM p690) posted floating-point scores of 1598/1699. But the scores aren't official yet.

http://www-1.ibm.com/servers/eserver/pseries/hardware/system_perf.pdf

 

[Eug]

Originally posted by: Vespasian
The 1.7GHz Power4+ (in a IBM p690) posted floating-point scores of 1598/1699. But the scores aren't official yet.

http://www-1.ibm.com/servers/eserver/pseries/hardware/system_perf.pdf

Wow, that's pretty impressive, and the Itanium 2 scores are even more impressive.

I'm more interested in the PPC970 scores simply because I want to buy a laptop with it in 2 years. Unfortunately, there is not much chance for a Power4+ laptop...

 

[dullard]

Anyone know when the Opteron 140 and 142 will be available? I'm assuming the 144 will be delayed like the 244 is.

 

[thorin]

Originally posted by: Sunner
Quite impressive, I think that's the first time I've ever seen a top AMD CPU score higher than a top Intel CPU, especially in SpecFP, if only Peak.
Of course Im not counting Itanium, but they won't be competing, so I think that's fair enough.

Ok the Opteron has a 97 point spread (peak-base .... ~7.96% variance in results) while the P4 has a 6 point spread (peak-base .... ~ 0.5% variance in results) and you're impressed that the 'peak' Opteron score is 12 points higher then the P4 ?? (Which is less then 1% at the level they are performing).

BTW Why is everyone saying that Opteron isn't competing with Itanium???? Itanium = 64bit CPU and Opteron = 64bit CPU therefore it would stand to reason that they're in competition.

Now don't get me wrong I do really like AMD and I'm glad they can beat Intel sometimes but I don't see anything impressive here. They can beat Intel ... um ... what like 2% (maybe 3%) of the time by ~1% ... BFD Even assuming Intel always scored it's base score that'd only make the Opteron beat it ~10% of the time by ~1% still not a BFD ...... even if the Opteron beat them 20% of the time by ~1% I don't think that'd be impressive in my books

Just my 2 cents.

Thorin

 

[Sunner]

Originally posted by: thorin

Originally posted by: Sunner
Quite impressive, I think that's the first time I've ever seen a top AMD CPU score higher than a top Intel CPU, especially in SpecFP, if only Peak.
Of course Im not counting Itanium, but they won't be competing, so I think that's fair enough.

Ok the Opteron has a 97 point spread (peak-base .... ~7.96% variance in results) while the P4 has a 6 point spread (peak-base .... ~ 0.5% variance in results) and you're impressed that the 'peak' Opteron score is 12 points higher then the P4 ?? (Which is less then 1% at the level they are performing).

BTW Why is everyone saying that Opteron isn't competing with Itanium???? Itanium = 64bit CPU and Opteron = 64bit CPU therefore it would stand to reason that they're in competition.

Now don't get me wrong I do really like AMD and I'm glad they can beat Intel sometimes but I don't see anything impressive here. They can beat Intel ... um ... what like 2% (maybe 3%) of the time by ~1% ... BFD Even assuming Intel always scored it's base score that'd only make the Opteron beat it ~10% of the time by ~1% still not a BFD ...... even if the Opteron beat them 20% of the time by ~1% I don't think that'd be impressive in my books

Just my 2 cents.

Thorin

Yes it's impressive, cause Intel has traditionally been far ahead of AMD in SPEC, see for example the lead the P4 holds over the AXP.

As for Opteron and Itanium competing only cause they're both 64 bit CPU's, that's got to be one of the most flawed arguments I've ever seen.
By that logic any 64 bit MIPS would be competing with Itanium.

 

[thorin]

Ok if you're impressed by a 1% lead sometimes then that's cool. I just don't see it If it was a 1% lead 100% of the time I wouldn't question it's impressiveness though I still wouldn't be impressed. I'd simply think "ho hum it's 1% but at least it's consistant". Mind you I also don't see the point in paying for high octane gas for my Corolla

As for Opteron and Itanium competing only cause they're both 64 bit CPU's, that's got to be one of the most flawed arguments I've ever seen.

I was completely serious with that question, why does everyone insist they aren't in competition?

1) P4 and Athlon64 (currently Barton) are mid/workstation chips.
2) Celery is still the low end. The Duron is discontinued so I guess Barton is in here too (there's a rumoured 1/2 cache version of barton .... yes that's basically just a Thoroughbred ... don't ask me it's a rumour). There's also a low end Athlon64 on the road map for next spring with small cache.
3) Itanium and Opteron are both 64bit chips aimed at the server market, yes there are other 64bit chips and yes they all compete I've seen numerous articles about Intel (and AMD) competition with HP, Sun, etc.....

I don't see how they aren't in competition ..... if they somehow aren't that's fine, I just don't see why ppl are saying that. Fill me in

Thorin

 

[Elcs]

Originally posted by: thorin

2) Celery is still the low end. The Duron is discontinued so I guess Barton is in here too (there's a rumoured 1/2 cache version of barton .... yes that's basically just a Thoroughbred ... don't ask me it's a rumour). There's also a low end Athlon64 on the road map for next spring with small cache.

Thorin

The lower cache version has been called Thorton. Dunno what bus speeds but Id put my money on it being released at the new 400mhz alongside the 400 Bartons. Just to fill you in there.

I dont follow much of the action above Desktop level but considering my limited knowledge. Id be impressed if at the first attempt (I think) that AMD can get anything half decently performing against Intel. Like playing a skilled sport against a professional or a more experienced player. The skills of the experienced player may be far greater than yours but showing competancy at your early attempts shows promise.

EDIT: It is only 1 benchmark program. We have seen some programs can be favourable to one side so to get a good view of the situation, Id need more proof.

Just my view, no matter how narrow minded it may be to others.

 

[Sohcan]

Originally posted by: thorin

BTW Why is everyone saying that Opteron isn't competing with Itanium???? Itanium = 64bit CPU and Opteron = 64bit CPU therefore it would stand to reason that they're in competition.

This fact alone is hardly an argument...the Cray X1 is also a 64-bit microprocessor. But I know of at least one of its architects that would scoff at the mention of these other "toy" scalar ISAs (vector guys tend to do that ).

There certainly is some overlap of their respective markets as Itanium continues to drive down (there are, after all, workstation models from HP). But Itanium 2 shared-address space systems go far higher, up to 32-way (Unisyst ES7000) to 64-way (HP superdome and SGI Altix 3000). From their product literature, it's quite evident that AMD is going after the Xeon and Xeon MP with their 2-way and 4-way systems.

* not speaking for Intel Corp. *

 

[thorin]

Originally posted by: Sohcan

Originally posted by: thorin

BTW Why is everyone saying that Opteron isn't competing with Itanium???? Itanium = 64bit CPU and Opteron = 64bit CPU therefore it would stand to reason that they're in competition.

This fact alone is hardly an argument...the Cray X1 is also a 64-bit microprocessor. But I know of at least one of its architects that would scoff at the mention of these other "toy" scalar ISAs (vector guys tend to do that ).

There certainly is some overlap of their respective markets as Itanium continues to drive down (there are, after all, workstation models from HP). But Itanium 2 shared-address space systems go far higher, up to 32-way (Unisyst ES7000) to 64-way (HP superdome and SGI Altix 3000). From their product literature, it's quite evident that AMD is going after the Xeon and Xeon MP with their 2-way and 4-way systems.

* not speaking for Intel Corp. *

Oh ok now I understand what I was missing. I DID NOT KNOW that Intel still made a Xeon processor .... I thought it was all P4 or Itanium.

Thorin

 

[dullard]

Originally posted by: Sunner
Of course Im not counting Itanium, but they won't be competing, so I think that's fair enough.

Look at what AMD says here. They give 18 benchmarks, and in that they compare the Opteron to the Itanium in 14. Thus it looks to me like AMD is puting Opteron and Itanium in direct competition. As a consumer looking for a good server (for number crunching) I will compare the Opteron to the Itanium. So AMD is comparing them, and the consumer is comparing them. That leaves one question: why on Earth do you think they aren't competing?

They both can be run in low number CPU configurations, and they both can scale up without limits. I'd love to see a 64 processor Itanium to 64 processor Opteron benchmark. Well get to see those possibly when the 800 series Opteron is released... What market do you think AMD is going after with their native 8-way processor support? What about their unlimited processor capablities with external chipset?

A platform can support an unlimited number of AMD Opteron processors in a multi-way configuration with external/additional chipset support.

 

[Markfw]

dullard, I think that the Opteron 1xx series are for workstations only, and the 2xx and 4xx and 8xx are for 2 way/4 way/8 way servers that DO compete with the itanium.

 

[dullard]

Originally posted by: Markfw900
dullard, I think that the Opteron 1xx series are for workstations only, and the 2xx and 4xx and 8xx are for 2 way/4 way/8 way servers that DO compete with the itanium.

Exactly. AMD has multiple target markets. Some of which will compete with Xeon, others with Xeon MP, and still others with Itanium. It is foolish to say "Opteron won't compete with Xeon" and it is foolish to say "Opteron won't compete with Itanium". Honestly Opteron initially won't put up much of a fight in the Itanium battle, but it will be a battle.

 

[thorin]

Originally posted by: Markfw900
dullard, I think that the Opteron 1xx series are for workstations only, and the 2xx and 4xx and 8xx are for 2 way/4 way/8 way servers that DO compete with the itanium.

That also makes sense.

Thorin

 

[dexvx]

Many people say that the biggest loser will be Intel in the Itanium v Opteron v XeonMP war. However, IMO, the biggest loser will be Sun Micro. The only thing that is keeping them afloat is their legacy support and high sustained IO backplanes. Their modern procs cannot even hold a candle to anything that AMD or Intel launches, and Intel has steadily improved their server IO (I cant comment on Opteron).

The only semi-decent reason to use a Sun workstation was to develop code that you could use on larger platforms (64bit), but now that is diminished. Intel has been slowly chomping away at the workstation market, and this will only accelerate. With newer 16/32/64 way Itanium/Opteron servers, there is very little reason to go Sun, unless you want to keep your legacy. I know many companies past that had nothing but Sun workstations completely transformed to Intel/AMD based windows/Linux machines. IMO, Sun will die out in ~10 years given their current rate of decline.

 

[Random Variable]

However, IMO, the biggest loser will be Sun Micro.

Maybe. But their processor roadmap is very ambitious.

http://forums.anandtech.com/messageview.cfm?catid=27&threadid=986116&FTVAR_MSGDBTABLE=arc

And the performance of the 1.2GHz UltraSPARC III isn't really that bad.

 

[Sunner]

Well, of course Opteron will face Itanium once in a while, but Itanium isn't the main competitor.
Xeon DP/MP are.

I DID NOT KNOW that Intel still made a Xeon processor .... I thought it was all P4 or Itanium.

Intel won't stop making Xeons for a LOOOONG time, they're making big money of them.
And besides, they don't have anything to fill that spot, look at Itanium prices, IIRC they start at $1.500+ and go up to over $5.000, depending on cache and speed.

 

[EdipisReks]

Originally posted by: Eug
I'm more interested in the PPC970 scores simply because I want to buy a laptop with it in 2 years. Unfortunately, there is not much chance for a Power4+ laptop...

IBM has stated the that .09 nanometer 970's are going to be available in q4. these are the chips that are suitable for laptops, so i hope to be able to sell my current TiBook and get a 970 based PowerBook sometime within in the next year.

 

[Eug]

Originally posted by: EdipisReks

Originally posted by: Eug
I'm more interested in the PPC970 scores simply because I want to buy a laptop with it in 2 years. Unfortunately, there is not much chance for a Power4+ laptop...

IBM has stated the that .09 nanometer 970's are going to be available in q4. these are the chips that are suitable for laptops, so i hope to be able to sell my current TiBook and get a 970 based PowerBook sometime within in the next year.

0.9 in Q4?????? Do you mean 2003? Holy crap! Somehow I don't buy that though. No 0.13 blades or Macs are even available yet.

But given that 0.9 might scale up to 2.5 GHz (or even higher), my prospects for a 2005 laptop are looking promising. Mind you I dunno what Intel/AMD will be up to by then.

QUESTION:

In a single processor config, is the performance of a 144 (assuming the same GHz) identical to a 244?

 

[BDSM]

I think that what is going to make the Opteron a hit isn't it's performance but it's price.

Opteron based systems willl be by far the cheapest entires to the 64 bit market.

Why is that important? Because the opteron can handle more than 4 GB of ram. Something that the Xeons can't do without a huge performance hit.

So.. I think the Opteron will do well in low to mid end servers for applications that need more than 4 gigs or ram. And this is a rapidly gorwing market.

Also.. we don't know the performance of the opteron in 8 processor systems just yet. Maybe it will be great?
We'll just have to wait and see.

 

[EdipisReks]

Originally posted by: Eug

Originally posted by: EdipisReks

Originally posted by: Eug
I'm more interested in the PPC970 scores simply because I want to buy a laptop with it in 2 years. Unfortunately, there is not much chance for a Power4+ laptop...

IBM has stated the that .09 nanometer 970's are going to be available in q4. these are the chips that are suitable for laptops, so i hope to be able to sell my current TiBook and get a 970 based PowerBook sometime within in the next year.

0.9 in Q4?????? Do you mean 2003? Holy crap! Somehow I don't buy that though. No 0.13 blades or Macs are even available yet.

But given that 0.9 might scale up to 2.5 GHz (or even higher), my prospects for a 2005 laptop are looking promising. Mind you I dunno what Intel/AMD will be up to by then.

QUESTION:

In a single processor config, is the performance of a 144 (assuming the same GHz) identical to a 244?

yep. 2003. supposedly the .13 has been shipped to Apple's manufacturers for some time for desktop machines. when i said "IBM stated" that might have been too strong. what i should have said is that "it has been stated that IBM has stated" . if true, i imagine that the .09 is some kind of expidited portable only chip. In my opinion, Apple is living by their laptops right now, and they would need IBM to push out portable capable chips as soon as possible after the unveiling of the 970 desktops. i think it'll be like when the G4 was introduced: desktops had it fairly early, but laptops had to wait until a mobile version came out later in the yeat. i certainly hope that this info is true, however, as the whole Mac line desperately needs a kickstart of CPU speed. i have no idea of whether the 144 and the 244 perform the same in single processor configs. i do know this: if 970 based laptops come out soon and live up to their potential, i'll probably sell that rig in my signature and use a PowerBook exclusively.

 

[Eug]

i do know this: if 970 based laptops come out soon and live up to their potential, i'll probably sell that rig in my signature and use a PowerBook exclusively.

I"ll keep my PC, because my PC desktop and my Mac laptop complement each other. Some things I can't do on the Mac, I do on the PC, and some things I can't do on the PC, I do on the Mac.

If there is no big architecture upgrade soon, I'll likely keep my 1 GHz PowerBook for some time. But if a super fast PPC970 PowerBook were to show up mid-2004, I suppose I could be convinced to upgrade a year earlier than I had originally planned.

By the way, the Centrino chips seem pretty good, but so far I haven't seen as nice a laptop as the PowerBook.