Posted on Sunday, November 09 2008 @ 20:36 CET by Thomas De Maesschalck
Sylvie Barak from The Inquirer
writes AMD has submitted power efficiency benchmark scores of their upcoming Shanghai processor to SPEC. The power efficiency of Shanghai isn't as good as the one from Intel's latest Xeon chips but she noticed AMD also submitted results of a suspiciously poor performing dual Xeon L5420 system. We'll have to wait until Wednesday to see if it's true, but Barak speculates AMD may use these borked Xeon L5420s results to make their Shanghai chips look good on paper:
Shanghai at 2.7GHz manages to score 860, while the 2.5GHz manages a score of 731.
All well and good, except that with those kinds of scores, Shanghai doesn’t even come close to Intel's already existing Harpertown chips, let alone Nehalem.
Intel chips certainly seem to be the big winners as far as this particular benchmark is concerned, with Power Leader running Chipzilla’s L5430 Xeons and scoring 1135. IBM also pulls two very high scores from its hat with Intel’s Xeon X3360 2.83GHz (1064, 1054), while NEC also posts a high score of 1010 on a Xeon L5420 2.5GHz.
But there’s also something a bit suspect going on in these benchmarks, something which makes us think AMD might be up to some shenanigans.
AMD submitted its own results for Intel Xeon L5420s at 2.5GHz and somehow managed to come up with a score of only 561. How come? Well, by very badly configuring the system to ensure the Intel server got the lowest possible score, it would appear.
The system sports eight FBDimms at 2GBs each when, with only four FBDIMMS at 4GB each, a much better score can be achieved. The system would run faster and would also cost less.
But AMD seems to be stuffing its box to crank the power up via C, thus artificially hammering the power efficiency. What this amounts to is picking the worst case scenario and possibly using that as a base from which to post a particularly hostile submission.
AMD also appears to be using different memory configurations on its own machine, specifically 4x4 instead of 2x8. This, we assume is because 2x8 would slow AMD down a fair bit.
