Hier Klicken

v5.xx drivers comparsion (1/2)

Whenever NVIDIA releases a new product, the company usually marks the occasion by releasing a new generation of drivers, with a new version number to reflect that change. These are always released as a set of unified drivers, meaning that they support older chipsets as well as the newly released ones. The drivers themselves consist of two parts, one of them responsible for Direct3D and the other in charge of OpenGL, with each driver core developed by a separate team.

NVIDIA frequently posts so-called reference drivers on its website. Card makers use these reference drivers to build their own set of drivers on them, adding their own logos and often additional tweaks or features. Today it is literally unheard of that a videocard manufacturer develops its own drivers - which is just fine with the chip companies, as they have no interest in unwittingly giving away potential secrets or design details of their products.
From time to time, a set of "unofficial" reference drivers will pop up on the web. So what does unofficial mean? Basically that they weren't posted by NVIDIA, but have instead been leaked to websites by way of one of the NVIDIA developers' websites or one of the retail companies. Since this "leakage" means that the drivers pass through many hands and servers, there is the possibility that they may have been manipulated by a third party along the way. So whoever uses these unofficial drivers needs to be aware of the risk. Besides, these drivers are neither supported by NVIDIA nor by any of the videocard makers. Usually you'll know very quickly whether a given set of drivers is "good" or not. If you're one of the more cautious types, you should wait a while and have a look around forums and websites to check out what people have to say about them.

This obviously brings up the question which driver is best. Does newer necessarily mean better? Most of the new GeForce2 cards use drivers built around the 5.16 release. A few manufacturers are already offering their own versions of releases 5.22 and above. The most recent unoffical version is v5.30. Although there were rumours about a version supposedly called v5.31, the consensus seems to be that these are not genuine. But getting back to the question at hand… To find out which drivers are the best ones, we (my trusty mascot and I) ran a slew of benchmarks for your pleasure.

We tested the following releases: v5.16, v5.22, and v5.30. The testbed system consisted of a Tekram P6BX-A mainboard equipped with a PIII 800 E and 192 MB RAM PC-100 running a fresh install of Windows98 SE (Ghost rules!)

The Benchmarks

Since the drivers literally consist of two subcomponents (OpenGL and Direct3D), it seemed logical to test both. To do so we used Quake3 (v1.17, with disabled sound) and 3DMark2000. Testing with or without sound is basically a philosophical decision. Testing without sound lets you push the limits of your the graphics card. On the other hand, testing with sound yields much more realistic results, or do you know anyone who plays without sound? (Unless your playing an absolute newbie and just don't want to hear him dying over and over again, begging for mercy!).

Quake 3 Arena v1.17 - Demo001
driverversion v5.16 v5.22 v5.30
Normal 1024 110,9 110,4 110,4
Normal 1280 88,1 88,0 88,0

High Q 1024 79,6 79,6 79,4
High Q 1280 44,8 44,3 44,3

The results show us that not much has changed where OpenGL performance is concerned. The differences could just as well be attributed to the normal inaccuracies inherent in testing. To see if T&L performance has seen any improvement I decided to put the drivers through their paces using NVIDIA's CPU-hungry NV15Demo.

Quake 3 Arena v1.17 - NV15Demo
driverversion v5.16 v5.22 v5.30
Normal 1024 26,0 25,9 26,0
High Q 1024 25,5 25,6 25,8

Again, par for the course, with no differences to speak of. Lets take a peek at Direct3D performance. 3DMark2000 was the test of choice in this department.

3D Mark 2000
driverversion v5.16 v5.22 v5.30
1024x768-16 6005 6022 6016
1280x1024-16 4894 4899 4892

1024x768-32 4613 4615 4608
1280x1024-32 3107 3107 3100

No real differences across the board, with negligible deviations. You can get the detailed results from the 3DMark2000 graph below. All tests were run at 1024x786-16:

3D Mark 2000 - 1024x768-16

The only real differences are visible in the High Polygon Count tests. Interestingly, though, it's the oldest version that comes away with the best results. This is a moot point, however, since the "real-world" sections of the suite (i.e. the game tests) once again show a level playing field.

NVIDIA
Detonator 2

Driver comparsion

 

 

Copyright: 29.06.2000 -   RIVA Station 2000 - Lars Weinand
No Copy without Permission!

URL of this Article: www.rivastation.com/v5xx_drivers_e.htm - If you want to link to it, please use this URL! :-)

Next page