backtop


Print E-mail del.icio.us 55 comment(s) - last by josmala.. on May 28 at 4:57 PM

A few preliminary benchmarks are in, but the picture is still cloudy over "Woodcrest" performance

The Tech Report's Scott Wasson had an incredible opportunity to sit down and benchmark Intel's upcoming Woodcrest processor in a dual server configuration.  Intel's Woodcrest processor, as we've previously covered, is the upcoming Socket 771, 65nm dual core server processor for Intel servers.  Intel gave the media a preview of Woodcrest outperforming AMD Opteron by 33% in synethetic benchmarks during IDF, but today The Tech Report has a full scale analysis using its own benchmarks.  The review was still done under the watchful eye of Intel employees during a reviewer's workshop, but the publication was able to conduct its own benchmarks.

Absent from the benchmarking suite was the application that Intel used to demonstrate a 33% victory at IDF, dual Woodcrest 3.0GHz processors manage to squeeze by with 15% victory over dual Opteron 285 however though.  SPECapc on 3DStudio MAX put the Woodcrest processor at a much higher advantage performance, but most other benchmarks that Wasson ran only put Woodcrest in the 5 to 10% favored range.  The only benchmarks that gave Woodcrest considerable advantages were SiSoft Sandra artificial benchmarks.

When the power consumption between Opteron and Woodcrest were compared, Woodcrest had slightly lower power consumption, particularly when idle.  Intel's "33%" estimates from IDF obviously extrapolated the 10-15% performance victories to a performance-per-watt scenario.

We were a little surprised that Intel only provided Opteron 285 CPUs for comparison, since Opteron 290 processors have been available for several weeks already through the correct channels.  To be fair, neither Woodcrest 3.0GHz nor Opteron 290 have been officially announced yet. 


Update 05/23/2006: Scott Wasson tells me that Intel did not supply the Opteron 285 processors, and that the CPUs were actually tested independently in The Tech Report labs. 


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Dissapointing
By akugami on 5/23/2006 5:37:20 AM , Rating: 4
From a 30+ percent advantage to a mere 5-10% and 15% in a few select applications is a huge drop. While this would give Intel the lead again, it's not the Athlon killer we were lead to believe initially. Still, we'll need to wait for more sites like Anandtech, [H], Xbitlabs, etc to benchmark these before coming to a more solid conclusion.




RE: Dissapointing
By nrb on 5/23/2006 6:01:42 AM , Rating: 2
We're talking about Woodcrest vs Opteron, here - server CPUs. The speed advantage Conroe will have over Athlon 64 has been independently verified and really is the 20%+ originally announced.


RE: Dissapointing
By Griswold on 5/23/2006 6:10:43 AM , Rating: 2
Ah really, what is the difference between opteron and athlon64 again? :P You could argue the test suite is different, but then again, they didnt even test real world server applications.

Nope sorry, with these figures and a woodcrest running 400MHz faster than the 285, I really want to wait for real benchmarks conducted by reputable sites.

I have little doubt that core 2 will be faster than what AMD has been offering for months, but with previews like this one, things appear in a new light.

Let's wait and see.


RE: Dissapointing
By GoatMonkey on 5/23/2006 8:47:42 AM , Rating: 2
I would bet that Hypertransport would account for the difference considering that they are comparing dual processor machines.


RE: Dissapointing
By Griswold on 5/23/2006 11:03:56 AM , Rating: 2
Yea more HTT links. But that doesnt translate to higher IPC. Which was kinda my point.


RE: Dissapointing
By josmala on 5/28/2006 4:57:11 PM , Rating: 2
No the difference is not only that in those comparisons.

Woodcrest vs opteron

Server apps.
Both have TWO processor chips which to syncronise between.
Woodcrest syncronises through chipset while opteron goes directly to other cpu core.

Consider this, path from one cache to other cache in woodcrest, is
prosessor1->chipset->processor2->chipset ->processor1
Opteron:
processor1->processor2->processor1

conroe vs A64

Desktop apps.
No multiple dies to syncronize with
Advanced prefetcher on conroe has dedicated the memory controller bandwith to its self instead of having larger risk of conflict with other die in woodcrest.

So there it is, different benchmarks and syncronisation between multiple dies, makes the difference from conroe comparison to woodcrest comparison.


RE: Dissapointing
By proamerica on 5/23/2006 8:46:38 AM , Rating: 2
"We're talking about Woodcrest vs Opteron, here - server CPUs. The speed advantage Conroe will have over Athlon 64 has been independently verified and really is the 20%+ originally announced."

These are Dual processor systems, Intel's FSB bottlenecks shoudn't become a big issue until 4p and up.

AMD should be able to take back the performance crown just by moving their architecture to 65nm and ramping clock speed as they have said they will do.


RE: Dissapointing
By IntelUser2000 on 5/23/2006 9:20:42 AM , Rating: 2
quote:
From a 30+ percent advantage to a mere 5-10% and 15% in a few select applications is a huge drop. While this would give Intel the lead again, it's not the Athlon killer we were lead to believe initially. Still, we'll need to wait for more sites like Anandtech, [H], Xbitlabs, etc to benchmark these before coming to a more solid conclusion.


No... go read the article.


not dissapointing - it beat the AMD!
By hstewarth on 5/23/2006 9:48:07 AM , Rating: 1
I still can't believe AMD fans, even though it beat the AMD in ever test except for memory ( not even using 667Mhz ram )

This is also just a preview - so the real stuff is not release yet.


By bpurkapi on 5/24/2006 3:24:45 AM , Rating: 2
so a new car will beat an old car, how suprising! The comparisons between the chips is getting absurd. The problem is that the chips are not really comparable, the amd chip used is not exactly the correct chip to compare to a product that has yet to launch, Lets all hold our bias back until real world situations and then we can talk which one performs better. sometimes you have to patient before crowning a king of performance(%)


This explains Dell's move quite well...
By Viditor on 5/23/2006 7:12:04 AM , Rating: 2
1. This comparison of a 2.6 GHz Opteron with DDR and a 3GHz Woodcrest with FBDimms was much closer than I think anyone expected...

2. With the new lower power process and then the move to 65nm expected this year for AMD, the Opteron will quite probably maintain it's domination over 2P servers (and of course 4P will remain AMD unchallenged). However, the Woodcrest looks to be very good for the lowend 1P server space, and I expect Intel will make some nice inroads there.





By DigitalFreak on 5/23/2006 7:54:45 AM , Rating: 1
LOL


RE: This explains Dell's move quite well...
By Phynaz on 5/23/2006 9:48:55 AM , Rating: 2
quote:
2. With the new lower power process and then the move to 65nm expected this year for AMD, the Opteron will quite probably maintain it's domination over 2P servers (and of course 4P will remain AMD unchallenged). However, the Woodcrest looks to be very good for the lowend 1P server space, and I expect Intel will make some nice inroads there.


Forever the fanboy.

Did you actually READ the article?

Is AMD going to get an instant ability to ramp clockspeed 30% by going to 65nm?


RE: This explains Dell's move quite well...
By Viditor on 5/23/2006 10:16:32 AM , Rating: 2
quote:
Is AMD going to get an instant ability to ramp clockspeed 30% by going to 65nm?


You should really just say no to drugs...

1. If YOU had read, you'd notice that they used the 2.6 GHz instead of the more current 2.8 GHz Opteron DC
2. Ramping clockspeed is not linear to ramping performance, so they don't need to ramp the clockspeed by 30%. They need to improve performance by 5-20% (you really SHOULD read the article you know...).
3. For every 200MHz increase, the Opterons tend to increase performance by ~7-10%. This means that...gee...at equal clockspeeds they would be performing at ~the same performance. Since the 2.8 GHz is shipping already and about to be announced (with an already tried and qualified platform I might add...), AMD needs only a single speed bump to equal and 2 to surpass the Woodcrest at 3GHz. Something tells me that it MIGHT be possible at 65nm...what do you think?
4. My main point was the power useage...at 65nm, don't you think it's POSSIBLE that the Opteron will be a little lower than their already superior power profile?


By Phynaz on 5/23/2006 10:53:06 AM , Rating: 2
??

Please explain how an identicle cpu with an increase in clock speed of say 5%, can increase its processing speed by greater than that.

So any way let's say clockspeed did ramp to 3Ghz on the AMD cpu. That's (3.0 - 2.6)/2.6. A 15% increase. That would make it match the Woodcrest on ONE benchmark. It would be behind on all the others.

Not to mention that Woodcrest will not be standing still while Opteron ramps in clock speed.

Really Dude, I don't understand where you come from. Why so much grasping at straws? Are you hurt somehow that Intel has a superiour cpu? Do you need some kind of validation that you bought the right computer?



RE: This explains Dell's move quite well...
By defter on 5/23/2006 11:32:59 AM , Rating: 2
quote:
1. If YOU had read, you'd notice that they used the 2.6 GHz instead of the more current 2.8 GHz Opteron DC


Uhm 2.8GHz DC Opteron doesn't exist. In fact, 2.6GHz DC Opteron was just launched last month.

Second, it's quite silly to compare mid-2006 Intel CPU to possible December-2006/Q1-2007 65nm CPUs from AMD. Considering the low power consumption of Woodcrest, Intel could easily increase clock speed of the Woodcrest in the latter half of 2006 if necessary.


RE: This explains Dell's move quite well...
By Viditor on 5/23/2006 1:03:06 PM , Rating: 2
quote:
Uhm 2.8GHz DC Opteron doesn't exist. In fact, 2.6GHz DC Opteron was just launched last month


At the end of this article...

"We were a little surprised that Intel only provided Opteron 285 CPUs for comparison, since Opteron 290 processors have been available for several weeks already through the correct channels"



By ShapeGSX on 5/23/2006 1:37:56 PM , Rating: 2
The Opteron 285 is $1051.
The Xeon 5160 will retail for $851, according to a number of sources.

They were already in the wrong price range.


By ianwhthse on 5/23/2006 2:58:10 PM , Rating: 2
And from the ORIGINAL article,
quote:
"After the workshop, we returned to our labs and attempted to configure an Opteron-based test system as closely to the Woodcrest system as possible for comparison. We were able to get pretty close, but we wound up using different hard drives and a few things like that, as you'll see on the next page. I don't believe these differences are likely to have affected the results of the few benchmarks we ran, fortunately."


And at least newegg, froogle don't sell the 290.


RE: This explains Dell's move quite well...
By hstewarth on 5/23/2006 3:24:14 PM , Rating: 2
This has absolutely nothing to do with Dell with AMD. Dell thing is likely becausse of Law suit. Dell will have these processors also. Lets just wait to see when 667Mhz ram chips are out and real tests are done.