Dhellman, thanks.
I had an enormous laugh after reading Dean Kent's impression, and then reading Van Smith's. That idiot can wail and hollar all he wants, and his band of cult zealots will hang on his every word. He can go ahead and join the lawsuit, too, since he seems to think it appropriate. Hey, whatever gets Intel down, eh? <ggg>
As for reality, it's good that someone was able to bring these benchmark issues into a better light. It's certainly not fair to bias a benchmark to favor one architecture over another. However, I think hypocrite Van Smith does exactly that with his COSBI benchmark, but since no one cares what Van Smith has to say, I doubt anyone will care to explore it in depth.
In the future, having AMD as a member of BAPCO (something long overdue, IMHO) will enable the committee to make better provisions in the benchmark. I don't think it means anything negative for BAPCO. After all, as Dean puts it, "I don't think it means SYSmark should be 'canned', since with this information the individual data can be extracted, so it can be a very useful tool - if one analyzes the data properly. It just means that SYSmark may truly be a niche benchmark (along the lines of STREAM, perhaps). Also, since AMD is now a member of BAPCo, there may be an 'adjustment' made for the next version that will address this problem."
Now that you realize that Intel is eager to "fudge" numbers to put themselves in a better light, you should also take a look back to DEC, Compaq, IBM, Commodore, Amiga, Apple, AMD, and countless others who have done the exact same in the past. When Apple shows off a couple supra-optimized Photoshop filters to prove that the Power MAC has the strength of two Pentium 4 CPUs, I don't see any law suits or criminal charges. Apple, like Intel, can claim any usage model they want, and technically be correct, at least from certain points of view. If you are an avid Photoshop user, and make use of those particular filters, then a Power MAC might be for you. Likewise, I am an avid Photoshop user myself, and I can tell you already that I make use of quite a few filters that Van Smith found as part of the SysMark Pentium 4 favored suite. It isn't *wrong* to use these filters, though it might be misleading to use *exclusively* those filters. The usage model for SysMark 2002 therefore is not *bad*, just *uncommon* - like many other benchmarks I can think of! Try Sandra or Viewperf or STREAM, or better yet, COSBI.
Of course, I fully expect the zealots to make a big deal over this, since they have little else to fault Intel at this point. Meanwhile, they continue to miss the fact that AMD's model numbers are similarly apt to come into scrutiny, especially since AMD markets them as the primary performance differentiator between their CPUs.
wbmw |
