Summary: With NVIDIA's recent spate of beta Detonator 40 driver releases, we figured it was time to take a more in-depth look at what they have to offer. We've examined their performance as well as image quality and compare them to the 30.82's as well as the third-party Omega drivers that are floating around on the web. Check out our finding's in today's driver report!
At a pace faster than a drunk stumbling to his porcelain abode, once again we’ve been caught in a barrage of NVIDIA drivers. This tends to happen every time a certain someone starts getting antsy about their position in the world – that and the NV30 is waiting to crack out of its little NDA covered shell.
Differences
The main difference between the 30 and 40 series of drivers is the inclusion of a refresh rate fix. Lord knows how long this has been needed, but it is through no fault of NVIDIA that this issue even exists in the first place. Microsoft generally caters to the lowest common denominator, and in doing so figured that everybody likes to run monitors at a migraine-inducing 60Hz. There are those amongst us who couldn’t tell the difference, but the vast majority in this community could tell the difference between 60Hz and 100Hz - even with their Coke bottle bottom glasses behind their backs.
Omega
For kicks we decided to include the third-party Omega drivers. Mind you these drivers are not tuned for straight performance, the creator had only image quality in mind. Without causing too much of a loss in performance, he’s managed to surpass NVIDIA’s reference drivers in that end. Mind you, he isn’t just turning on anisotropic filtering and FSAA. Through some hocus pocus on the backend settings, he has managed to achieve some pretty good results. These drivers can be found at: Omega Corner.
His method of improving on the NVIDIA drivers is not done through actual recoding, as he does not have the source to enable him to do so. Instead, he tweaks the settings a bit. The drivers also seem to have some more added options. If you take a peek at the pictures, you can see that the Coolbits hack is enabled out of the box, as well as some extra D3D settings. The drivers also come with their own refresh rate fix as well as some extra toys. [image]
At the moment he is basing his latest Omega driver off of v30.82. Apparently the author has qualms with v40.xx, as he has found that they tend to cause more problems, despite the performance increases. Image Comparison
We took some pictures in Giants: Citizen Kabuto to show the differences between the various drivers. The results were that the Omega drivers were considerably more detailed than any of the NVIDIA driver revisions. In fact, most of the NVIDIA drivers looked pretty much the same, regardless of revision. The main differences arose in texture quality and how far back the high quality textures were being implemented. Provided are the main pictures, compressed for obvious size constraints, and then following them are zoomed and cropped versions of the scenes on both sets of drivers. What’s really amazing is that even though we converted the bitmaps to jpegs, the differences are still readily noticeable even in the main pictures. Omega man certainly did his homework when he made these drivers.
Mecc Zoomed [image]
Kabuto [image]
[image]
SIDEBAR: An ostrich's eye is bigger that it's brain.
Test System
Intel Pentium 4 2.4GHz (533MHz FSB) Benchmarks
3DMark 2001SE - 32-bit, 32-bit textures
Quake III v1.32 - High Quality
Notes
The Detonator 40 drivers open up with a slim performance improvement over 30.81 at 800x600, but as the screen resolution is increased, the margin between the Detonator 30’s and Detonator 40’s increases. In particular we witnessed a roughly 2% performance improvement at low resolutions, but by the time the resolution is cranked up to 1600x1200, the gap opens up to as much as 7%.
3DMark 2001 - DirectX 8.0
Notes
Once again we see a nice performance improvement in the Detonator 40.xx drivers. We see an improvement of approximately 8 or 9% percent depending on the driver in particular as well as screen resolution.
Sean Connery, George Lazenby, Roger Moore, Timothy Dalton, and Pierce Brosnan.
Serious Sam II v1.07 - OpenGL
Notes
The margins close considerably in Serious Sam II. While our test results with Quake III and 3DMark 2001 demonstrated a performance improvement of greater than 7% at some resolutions, we’re looking at a much slimmer 2% margin here.
Jedi Knight II v1.03 - OpenGL
Notes
Our results with Jedi Knight II pretty much mirror what we saw with Quake 3. At low resolutions, we’re looking at a very slim performance improvement, but by the time the screen resolution is cranked up to 1600x1200 the gap grows to 10%.
Unreal Tournament Demo patched – flyby at 32-bit
Notes
At low resolutions, we encountered a slight performance decrease with the 40.xx drivers, but at higher resolutions things changed drastically! For instance, at 1600x1200 we’re looking at a 14% performance improvement – the widest margin in any test we’ve used in this article!
Budget: $120 million
Gross: $352 million to date
Goldfinger
Budget: $3.5 million
Gross: $124 million to date
In the process of running through these numbers we were surprised to find such a dramatic increase in performance when switching from the 30.XX to the 40.XX set of drivers. Unreal Tournament 2003 in particular managed to really pull something out from down deep to lay down the proverbial smack on the older NVIDIA drivers. Performance jumped by over ten percent across the board at high resolutions!
It certainly is impressive that NVIDIA managed to find this much latent performance still in these chipsets. Moving on up to the 40 series of drivers probably wouldn’t be a bad idea, although we can’t completely vouch for their stability or application compatibility, as the author of the Omega drivers states.
| |||||||||||||||||||||||||||||||||||||||||||||||||
| © Copyright 2003 FS Media, Inc. |