More MS fudging from Win2k speed trials

A sharp-eyed techie working for a major OEM (he understandably wishes to remain anonymous) draws our attention to further fudging in the Windows 2000 benchmarks exposed here yesterday. He also directs us to the NT versus Win95 product comparisons Microsoft ran at the time of NT 4.0's launch when - tsk - the results were precisely the reverse of the more recent NT-knocking propaganda. Our informant notes that in the latest benchmarks, carried out by ZD, a little extra work was obviously necessary to get video performance up. "In one configuration, they replaced the video card; in another, they went for updated video drivers. All because the originals were not 'fully optimised' for Win2000?!" Given that many of the real Win2k video drivers are still staggering out of beta, he accepts there was a rational for that. "But why not go the same length with NT4 or others? For instance, the NT4 driver they used for ATI Rage IIC was default release .1 while the latest available version is .3247... Some difference, huh? Surely at least ATI and Matrox are known for constantly tweaking their drivers for optimal performance and functionality." Why not, indeed? While the Win2k hardware benefited from the replacement of a poorly-performing video card, the NT4 system was left with a dodgy sound card "even though they noticed and documented its adverse effect on system performance!" The performance hit NT took could therefore conceivably have been enough to give Win2k the edge at 32 megabytes RAM. But way back when Microsoft was saying "Running Windows 95 at work? Eight solid reasons to move to Windows NT Workstation 4.)" the company was saying, citing Business Winstone, that NT4 was on average 22 per cent faster than Win95 with 32 megs, and faster still with 64. This is not of course quite what the most recent benchmarks said. It also seems quite probable that, were the various problems with the NT4 system configuration ironed out, the headline would have had to be something like: "Win2000 20% slower than NT4! Go update!!" In addition to the DMA problem we noted in yesterday's story, our informant also points to the file system used: "OK, FAT16 is the best choice for producing comparable results. Still, they could have mentioned that using NTFS, the recommended Microsoft choice for NT4 and Win2000 in business environments, is detrimental to system performance, pure and simple. [so the file system makes the test results pretty academic] Also it would have been interesting to learn about the differences in using NTFS4 and NTFS5, or FAT32." ® See also: Microsoft fudges Win2k speed trials Great fudges in history: MS explains why NT4 is triffic