NT outperforms self on same hardware

by Leonard Richardson

Published on segfault.org 04/15/1999

Microsoft Windows NT Server 4.0 is 2.5 times faster than Microsoft Windows NT Server 4.0 as a file server and 3.7 times faster as a Web server, according to a recently published Mindcraft study. A third test setup, which was never turned on, performed even worse.

NT 4.0 loses badly to NT 4.0 when it comes to file serving

In the past, Mindcraft has used Microsoft money to do benchmarks of Windows NT against Novell Netware, Red Hat Linux, Sun's Java Development Kit, Netscape Navigator, Corel Wordperfect Suite, cold cabbage soup, and anything else seen as a competitor to Microsoft. But this is the first time NT has been benchmarked against itself.

"Microsoft Windows NT Server 4.0 certainly took a heavy hammering from itself in our tests," said Mindcraft techie Jack Ptarmigan."On the other hand, this benchmark shows conclusively that Microsoft Windows NT Server 4.0 is the environment of choice for the spin-conscious enterprise. The Microsoft Windows NT Server 4.0 on the left, I mean. I pity the poor sap who recommends the Microsoft Windows NT Server 4.0 on the right."

Asked how to tell which version of NT was running on a particular computer, Ptarmigan took us into the wild world of benchmarking.

"The only real way to find out if you're got the 'good' Windows NT Server 4.0 or the 'bad' Windows NT Server 4.0 is to do the benchmarking yourself. There are many tuning parameters avaliable to you in the convenient Registry, so what you do is set up your system and keep running the test and fiddling with the parameters until you get one of the two graphs shown in our report. If it's the one on the left, you're in business. If it's the one on the right, you need to upgrade your system to get the full power of Windows NT Server 4.0. And today only, Windows NT Server 4.0 is just $699 for a ten-user license!"

I asked Microsoft vice-president Jim Allchin why Microsoft would dare to pit its flagship product head-to-head against itself. After showing me an informative video which laid to rest any doubts I might have had, he sat down to discuss the larger issue of credibility in benchmarking.

"We used to do all our benchmarking in-house, but found that the results lacked credibility outside of Microsoft. So we have outsourced our benchmarks to outside companies. We keep things quiet in our own channels and let the mainstream press pick it up right from the ones who actually performed the benchmarks."

Allchin cracked a smile. "You might even say our benchmarks are 'Open Source'."