Driver comparison Forceware 196.21 vs 186.18 (GTX 260 Benchmarks)

Introduction
----------------
After figuring out that low 3DMark Vantage GPU score is linked to the drivers (Forceware 195.62), I decided to continue using the latest drivers as this may be a one off case and usually it is recommended to use the latest drivers. When Forceware 196.21 was released I updated my drivers. I have used these for quiet some time and did not find any major problems.

This week I installed Batman Arkham Asylum and ran the inbuilt benchmark at various settings with 196.21. I found that the results are consistently lower by upto 10% compared to a similar configuration. A quick search revealed that a lot people are complaining about low performance and stuttering in various games. After some research 186.18 are the best Forceware drivers so far for gaming performance. I felt that going back to 186.18 will improve the performance. So I decided to run some tests using 186.18.

I installed 186.18 (Used driver sweeper and ccleaner in safe mode to clean driver and registry traces) and ran some tests. So did it improve the performance. Lets look at some results.

The Tests
------------

Test Machine
------------
Intel i5 750 (@ stock with Turbo enabled)
2x2GB G.Skill RipJaws @ DDR3-1333, 9-9-9-24
500GB WD Black WD5001AALS
XFX GTX 260 xxx edition (core: 640/ mem: 1000/ Shader: 1380)

OS
--
Windows 7 64-bit

GPU Drivers
-----------
Forceware 186.18
Forceware 196.21

Software Used
-------------
3DMark Vantage v1.01
Far Cry 2 v1.3
Batman Arkham Asylum v1.1

3DMark Vantage
--------------------

I ran this test with the performance preset and noted down the GPU scores.

196.21
--------
GPU SCORE: 8051

186.18
--------
GPU SCORE: 9857

As I pointed out in the other article, Vantage score is considerably more with
186.18. There is noticeably less stuttering with 186.18.

Far Cry 2
-----------

For this test 16x AF was forced in nvidia control panel and ultra high preset is selected with Fire, Physics and Real Trees set to very high. Vsync is disabled.

for the gaming results following format is used.
Min, Avg, Max FPS

196.21
-------
No AA, 16x AF, DX10
40, 50, 73

4x AA, 16x AF, DX10
33, 43, 62

186.18
-------
No AA, 16x AF, DX10
42, 53, 75

4x AA, 16x AF, DX10
36, 46, 65

There is an improvement in the FPS.

Batman Arkham Asylum
-----------------------------
In game settings are set to very high and PhysX is set to Normal. D3DOverrider is used to force vsync and triple buffering.

196.21
-------
1680x1050, 4x AA, vsync, triple buffering, Physx Normal, very high (
25, 43, 57

1680x1050, 4x AA, no vsync, Physx Normal, very high
28, 42, 56

186.18
-------
1680x1050, 4x AA, vsync, triple buffering, Physx Normal, very high
30, 50, 61

1680x1050, 4x AA, no vsync, Physx Normal, very high
30, 50, 68

There is a nice boot in avg. FPS. I was surprised, as I was expecting the newer drivers to perform better in newer titles. The gameplay is smoother and image quality has improved.

Conclusion
--------------
I got good performance boost (at least in the apps I use), smoother gamplay and better image qulity by going back to 186.18. For now I will stick with these, till I face some problem or newer ones perform better. I loose openCL and DirectCompute support but I don't need it for the time being.

Extra Batman Arkham Asylum performance data I have with 196.21
--------------------------------------------------------------------------------
1680x1050, no AA, no vsync, no physx, very high
69, 109, 150

1680x1050, no AA, no vsync, Physx Normal, high
32, 52, 78

1680x1050, no AA, no vsync, Physx Normal, very high
26, 50, 71

1680x1050,4xAA, no vsync, no physx, very high
49, 81, 105

1680x1050, 4x AA, no vsync, Physx Normal, very high
28, 42, 56

1680x1050, 4x AA, no vsync, Physx High, very high
20, 36, 55

1680x1050, no AA, vsync, triple buffering, Physx Normal, very high
28, 49, 61

1680x1050, no AA, vsync, triple buffering, Physx Normal, High
27, 51, 62

1680x1050, 4x AA, vsync, Physx Normal, very high
24, 43, 60