RSSIntel: Harpertown Xeon Performance Will Beat QC Opteron
By Mykel Nahorniak | Published September 19, 2007, 12:01 PM
In the only raw performance test figures revealed yesterday by Intel Senior Vice President Pat Gelsinger at Fall IDF in San Francisco, preliminary tests of the company's forthcoming quad-core "Harpertown" architecture Xeon 5400 show it beating a quad-core Opteron 2300 system in the critical SPEC floating point throughput rate test by just over 4%.
"We have leadership on a broad range of benchmarks," Gelsinger proclaimed yesterday, "literally every benchmark. But the area where competition is closest is in bandwidth-intensive and floating-point areas. And what we see here is [that], with the improvement of the 1600 [MHz] front-side bus, [and] the larger cache of 3.2 GHz, going from [Intel] Clovertown to Harpertown, a 34% improvement."
Intel's tests showed a current generation (Clovertown) Xeon X5365 scoring a 66.9 in the SPECfp_rate2006 test. BetaNews confirmed that number this morning with the SPEC database, which revealed it to be a peak performance number, not a base (mean or average) performance number; for the dual-processor Supermicro X7DB8+ system used by Intel in that test, the base performance score was 63.1.
Assuming Intel is comparing peak scores only, Gelsinger's 32% improvement claim does check out, although the company has yet to submit a formal score for the Harpertown generation X5400 to the Standard Performance Evaluation Corporation.
Gelsinger cited an 86.3 score for an AMD Opteron 2360 processor-based system in the same test. But that may be an estimated score, since the 2360 is the anticipated 2.5 GHz edition "Barcelona" model that is still slated for release later this year; the 2.0 GHz 2350 is the highest performance edition available now. SPEC posted AMD's floating point rate score for that model at 77.3 peak performance, 72.4 base.
Setting aside the possible adjustments between estimates and final values for a moment, is Intel's comparison truly - to avoid the other connotation - oranges to oranges? The X5400s used in Gelsinger's comparison are 3.2 GHz processors that are resting on Intel's newly accelerated 1600 MHz front-side bus (1333 MHz being Intel's norm). Maybe the X5400 does beat the 2360 by 4% or so, but even Gelsinger himself didn't obscure the fact that his chip was revving at 22% more "tick-tocks" per second.
Still, he went on to proclaim, "Harpertown wins on [SPECfp_rate]...the fastest machines on the planet."
The emphasis on 'how many cores' is superfluous.
The real issue is inter-core communication.
And in this respect AMD DOES have an advantage over Intel.
Heat and power consumptions aside, there is more to actual performance than simply the number of cores. The abilkity to effectively leverage the multi-core architectures must take precidence.
And with AMD's close ties to IBM, and IBM owning the advanced tecnologies in this respect (witness THE SWITCH from the RS6000SP!), I would hope that AMD can leverage this relationship to kick tje inter-processor communication to a much greater scale.
Score: 0
|AMD should come up with a 3.14 core processor and that will beat Intel's pants off!!!1
Score: 0
|Maybe they should just skip the 3GHz mark and go straight on for the 4. That'll show Intel who's boss.
Score: 0
|No thanks, i don't want to change every board and possibly memories just to get those "performance boosts".
I prefer to keep my AMD systems and change the to Ncores cpu from time to time.
This is how companies see extra money in the end of the year, not changing the hardware every time intel decides to change chipset, ..., and all this to get % of boost measured in a OS that was optimized for it and that gives penalties to other companies cpus.
Maybe i change my mind if you show me the boots in a linux system :P
but this is just my 5 cents opinion...
Score: 0
|they should start measuring the performance in a "premium tax" so you can see what you'll be spending "extra" for what margin of performance gain. sure, great, it's .5% faster than an AMD chip of the same caliber, but is it 200-500 more per chip/board/ram? it that be the case, keep your .5%
Score: 0
|On the subject of "premium tax," does your electric bill not count? AMD chips suck up more juice than equivalent Intel chips do, and they do less with it. Why do you think AMD is trying to sue Intel for antitrust laws they haven't even broken? And why are they filing the case in Europe, rather than here at home? Because it's baseless; Intel is hurting them, and they're desperate. And American courts are too by-the-book to hear their "case."
In a time and place where natural resources are quickly being depleted (hopefully you realize the reason American troops have spent the last six years in a country that never attacked us is oil), and industries are still balking at solar power despite impressive advances, driving energy costs ever higher...that means something. If you keep your computer for more than just a year or two, and use it enough to justify its initial purchase, an Intel processor will return to you every penny of the difference, and then some.
Score: 0
|