Does the new Nforce 2 driver decrease performance?
Lost 30 FPS after installing
By Fuad Abazovic: Wednesday 21 May 2003, 15:16
NVIDIA RECENTLY introduced 2.41, the first set of drivers that can figure out whether you have Nforce 2 Ultra 400 or just the 64 bit version of this chipset called Nforce 2 400 and there's a fair list of enhancements.
Nforce 2 in the past had quite a nice pack of drivers that were able to make these boards work quite smoothly. The IDE driver was not WHQLd but this time Nvidia got this certificate, circumventing those annoying little warnings that otherwise pop up.
The first thing we wanted to check was that whether there was any performance increase in these drivers.
But our tests appear to show that with the Epox 8RDA3+ board that features the Nforce 2 400 Ultra chipset you actually experience a loss in the application suite we used for testing.
In the tests that we ran, we saw a performance decrease in Aquanox. Where I used to get 103.6 FPS with the previous driver on 10x7, with 2.41 I scored 97.7 FPS. On 640x480, the performance drops from 90.7 to 85.0 FPS. In some other game based tests, you can expect similar results.
Quake 3 dramatically fell from 313 FPS to 279.7 FPS at 640x480 while at 1024x768 performance dropped from 299.5 FPS to 271.3 only.
In Sandra 2003 we are also getting similar results but the CPU and Multimedia results haven't suffered as they are directly related to the microprocessor speed.
However, memory performance in the first test fell from a previous 2664 to 2523Mb/s, while in the second test performance decreased from 2491 to 2373MB/s.
Could it be that in order to get WHQL certificate Nvidia had to adjust its IDE performance?
Since Nvidia does not wish to talk to the INQUIRER, I'd quite like feedback from other hardware editors and reviewers on this subject.
Meanwhile if you don't have any problems with your previous Nforce 2 driver, it might be wise to wait and see what verdict the hardware jury returns.
It seems to me that WHQL always comes with something of a performance penalty. µ
http://www.theinquirer.net/?article=9606