Official IBM Barcelona benches: INT and FP approx -10% to Xeons

Wombat2

Distinguished
Jul 17, 2006
518
0
18,980
From pdf:

SPEC CPU2006 Benchmark:
x3455 %u2013 Quad-Core AMD Opteron Model 2347 Processor (1.9GHz, 512KB L2 Cache per Core)

SPECint2006 11.3
SPECint_rate2006 83.2
SPECint_rate_base2006 72.8
SPECfp2006 11.2
SPECfp_rate2006 73.0
SPECfp_rate_base2006 68.5

Edited ...
 

the_vorlon

Distinguished
May 3, 2006
365
0
18,780
Nice find...

I guess it comes down to scalling...

Clock for clock at 1.86 vs 1.9 ghz these parts look pretty darn close.

Intel currently has a huge frequency lead, as AMD closes this gap will hypertransport allow AMD to pull ahead as the ghz gap closes?

On the other hand, can AMD ramp Barcelona speeds faster than Intel can ramp Penryn speeds..?
 


I agree. If the chips are close clock to clock it will be goodnight AMD. Intel's manufacturing is superior and has a lot more dollars at its disposal. AMD is trying to get 65nm to work while Intel is getting 45nm going.
 

nruo

Distinguished
Apr 28, 2007
9
0
18,510

anyone how reads the reviews and believ it is being objective
 

weskurtz81

Distinguished
Apr 13, 2006
1,557
0
19,780
BTW,

They said the press didn't sign the first NDA, they said they had to sign
a second NDA telling them they could not post confidential info without the approval of AMD.
"On Day 2 though, they were presented with another NDA to sign before a factory visit. This one stipulated that "any confidential information from this visit would need written approval from corporate communications before it could be used"". It seems that this website is being a little biased in the reporting because they end the article with

"It is highly probable that the same NDA will be used to force everyone to toe the line and publish only suitably-positive, pre-approved articles. It's either that or face the threat of lawsuits. So, keep that in mind when you wade through the inevitable deluge of articles and reports on the AMD Quad-Core Opteron (Barcelona) processor from tomorrow onwards" after they said the first crappy nda wasn't ever signed.

"Finally, AMD agreed to let Don and the other journalists attend the event without signing that particular NDA."

Am I reading this properly? It seems to me the first one wasn't used but the second one was for confidential information about the FACTORY visit. And it talked about CONFIDENTIAL information. Of course AMD might have some stuff in the factory they don't want being talked about. I would expect no less when it comes to the factory visit. So, where is the story in this article.... seems much to do about nothing.
 

bixplus

Distinguished
Jun 2, 2006
398
0
18,780


Whoa, this is WAAAY off what AMD has advertised recently. Then again, they likely aren't doing a clock-for-clock comparison...most likely TDP(Intel) vs. ACP(AMD) values. That would be like comparing a ~1.6Ghz Intel with a 2.0 Barcy...hmmm, not sure what to think of that just yet, but that would account for the AMD inflated numbers a bit.
 

weskurtz81

Distinguished
Apr 13, 2006
1,557
0
19,780
The real test will be when the hardware review sites get to do some hands on. The Spec benchmarks are interesting and do offer some insight but I am more interested in other benchmarks.

wes
 

weskurtz81

Distinguished
Apr 13, 2006
1,557
0
19,780
bixplus,

how was this 1.6ghz and 2.0ghz when it is clearly pretty close clock for clock? It's basically a 1.9ghz to a 1.86ghz.... not 1.6-2.0.
 

bixplus

Distinguished
Jun 2, 2006
398
0
18,780


I'm just comparing this to what AMD has advertised recently, that they'll be 20%-35%(or something like that...correct me if I'm way off) faster than Intel in some of these benches, but don't specify exactly what the Intel processor was/is. They seem to avoid the "clock-for-clock" verbiage, or I haven't seen it anyway...so I am only guessing that they are using some other metric to compare...in this case TDP vs. ACP. /speculation

Who knows, I could be way off on this...but do agree with you that the real test will be when third parties run their own benches.
 

turpit

Splendid
Feb 12, 2006
6,373
0
25,780



If these numbers are remotely accurate, then It looks like AMD is using the wishful advertising vs. reality comparison.
 

weskurtz81

Distinguished
Apr 13, 2006
1,557
0
19,780
bixplus,

further up the page wombat posted:
"For comparison:
Hewlett-Packard Company ProLiant DL360 G5 (1.86 GHz, Intel Xeon processor E5320):


http://www.spec.org/cpu2006/result [...] 00517.html
SPECint2006 = 11.9


http://www.spec.org/cpu2006/result [...] 00585.html
SPECfp2006 = 10.9


So -7% on INT and +0.6% on FP clock for clock compared to Conroe class arches. Will be +- -20% and +- -10% compared to Penryn arch."

I am assuming those numbers are accurate, and that is what i was basing my statement on. Regardless of the above data, we both seem to agree that the most important test will be the third party benchies from trusted sites. It doesn't look all that great atm, but who knows how it will turn out.

wes
 

bixplus

Distinguished
Jun 2, 2006
398
0
18,780
Wes, agreed, the IBM behches that Wombat posted do seem to indicate similarity on a clock-for-clock basis. I'm just trying to wrap my mind around what AMD was thinking with their inflated performance numbers when a clock-for-clock would indicate differently.

That leads to a further discussion of whether clock-for-clock is a valid comparison, which both manufacturers will use or ignore depending on how theirs' performs against compeition at a point in time. In this case, I'm postulating that AMD will avoid the clock-for-clock comparison and use it's own home cooked TDP vs. ACP instead. I'm also wondering if this is what is implied by the title "AMD's Quadcore: defending new found territory" found in this thread: http://www.tomshardware.com/forum/245039-28-anandtech-barcelona-review

 

bixplus

Distinguished
Jun 2, 2006
398
0
18,780


lol...yeah that would be bad!!
 

Verge

Distinguished
Jul 30, 2007
20
0
18,510


No (IMHO).

Firstly, I don't see AMD really closing the frequency gap. Intel is using a process which will go further in clock frequency than the process which AMD is currently using (Hi-K and smaller geometries versus SOI. Intel's process expertise versus AMD's. Intel's budget versus AMD's. I'm going to call that a clear win to Intel, although maybe it will be closer than I expect and it may end up being close to a draw. But I don't see AMD winning that battle.)

If the hypertransport bus gives AMD an edge, that will show up in the multi-socket arena. That is, 4 socket and 8 socket servers (each socket populated by a quad cpu... just don't think about the power consumption for home use) will hit the wall first. When this truly happens, Intel will have their new architecture which will circumvent this problem (and, judging by the things Intel has been patenting, it looks superficially like HT, but with some interesting refinements).

Then what happens is that some of the tech that intel has introduced into the multi-socket server space makes its way down into the desktop space.

So, if Intel were to leave their arch alone for ~5 years and just turn up the clock speed, this would be a problem, but they aren't that dumb, can see this coming and won't. Sorry. (Mind you, it does look as if they won't introduce it a minute before it is needed, either.)

So (in my opinion) AMD must have a better arch in clock for clock terms than Intel if they are to win, or even fight effectively, the performance war. They could go back to the socket 7 days when they didn't fight the performance war but offered better value in mid-range computing, but its not what I'm hoping for and I don't think its what you are hoping for either.
 


Pre-frag.

NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA NDA
 

bixplus

Distinguished
Jun 2, 2006
398
0
18,780


lol...awe, man, come on. Is that all you can give us? :eek:
 

weskurtz81

Distinguished
Apr 13, 2006
1,557
0
19,780
I think we have come to all valid conclusions on this argument. I agree with you bixplus. I am not sure how (if the above figures are accurate) in the world AMD came up with the numbers that they released in the past months. I know an employee of said company that keeps reassuring me that a release on the 10th will be a big one(won't say if ATI is releasing something too). This person seems sure AMD will come out with some good stuff tomorrow.

It seems now though that this good stuff cannot be the Barcy launch. I will reserve my end judgment of the CPU though. I want to see some 3rd party stuff.

wes
 

weskurtz81

Distinguished
Apr 13, 2006
1,557
0
19,780
what do you guys think about the NDA article posted above? It really doesn't seem like a big deal because none of the journalists signed the first NDA agreement.
 

bixplus

Distinguished
Jun 2, 2006
398
0
18,780


Yeah, I've seen you mention this in other threads...got me real curious too.
 

turpit

Splendid
Feb 12, 2006
6,373
0
25,780


With all the hype from AMD this year, one has to question what they consider "big", or "good". Remember, the "Ulitmate Enthusiast System" turned out to be ultimate only in AMDs overestimation of its performance, and that AMDs idea of "Something Big" this summer turned out to be a tiny logo sticker, for a product which didnt exist, placed on the side of an F1 drivers helmet