VOGONS


First post, by Baoran

User metadata
Rank l33t
Rank
l33t

I have been trying to figure out what would be best nvidia driver version for my Gigabyte FX 5950 Ultra GT video card when it comes to both performance and compatibility. So far I have only been running 3dmark2000 and 3dmark01. First I had driver version 82.69 and everything seemed to work and I got about 17100 in 3dmark01 benchmark. After that I downloaded driver 71.84 from gigabyte support web page for my card and I got a really good benchmark result of about 19100 in 3dmark01, but problem was that 3dmark2000 didn't work at all and just crashed when I tried to start running the benchmark which didn't happen with 82.69. I also read some recommendations online that 45.23 would be a good driver version and everything seemed to work, but I got the worst result of 15872 in 3dmark01 so it made me wonder about compatibility and huge difference between benchmark results of different driver versions.

Does anyone has any experience of a good driver version that would have both good compatibility and performance?
Is using 3dmark good way to test these things or is there some other software you know that is better suited for testing driver versions?
Is it normal to have such big differences between driver versions or is there something else going on?

Reply 1 of 4, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Back in the day, Nvidia had some driver sets that did some maybe cheaty things to get higher 3DMark scores. Then later on the 3DMark people came out and said that NVidia wasn't cheating although the change in quality with certain driver sets was able to be seen and those sets that lowered the quality got better scores.

I would trust real game benchmarks for driver version comparisons way more than I would trust 3DMark.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 2 of 4, by Baoran

User metadata
Rank l33t
Rank
l33t
cyclone3d wrote on 2020-03-02, 15:51:

Back in the day, Nvidia had some driver sets that did some maybe cheaty things to get higher 3DMark scores. Then later on the 3DMark people came out and said that NVidia wasn't cheating although the change in quality with certain driver sets was able to be seen and those sets that lowered the quality got better scores.

I would trust real game benchmarks for driver version comparisons way more than I would trust 3DMark.

Which game would you recommend for testing compatibility/performance of different driver versions?
I do consider that there is a chance that the drivers from gigabyte web site overclock the video card from 475Mhz to 520Mhz, but I have found no proof of that. Some old reviews say that but I have not found confirmation of this in rivatuner or nvidia control panel.

Reply 4 of 4, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Back in the day, Nvidia had some driver sets that did some maybe cheaty things to get higher 3DMark scores.

You need Forceware 5x.xx for maximum 3DMark 2001 SE results. Ironically, it's one of the drivers which was shipped with later Win XP Service packs.

I must be some kind of standard: the anonymous gangbanger of the 21st century.