[Quick Test] NVIDIA ForceWare 186.18 vs ForceWare 190.38

Here is a quick performance test of ForceWare 186.18 WHQL vs ForceWare 190.38 WHQL.

Testbed:

• CPU: Intel Core 2 Duo E8400 default clocks
• Mainboard: EVGA 790i Ultra SLI
• Memory: 2048Mb DDR3
• Graphics card: EVGA GeForce GTX 295 CO OP Edition – default clocks (GPU core:576MHz / Shaders: 1242MHz and memory: 1008MHz)
• WinXP SP2 32-bit
• FurMark 1.7.0 (OpenGL)
• FluidMark 1.1.1 (OpenGL / PhysX)
• Lightsmark 2008 (OpenGL)
• Unigine Tropics 1.2 (OpenGL / Direct3D)

Settings:

• FurMark and FluidMark: fullscreen 1280×1024 60 seconds
• Lightsmark 2008: default
• Unigine Tropics 1.2: fullscreen 1280×1024 anisotropy 4X

	ForceWare 186.18	ForceWare 190.38	Gain
FurMark	4239	4275	+0.85%
FurMark SLI	8507	8505	0%
FurMark MSAA 16X	1122	1120	0%
FurMark MSAA 16X + SLI	2232	2230	0%
FluidMark 1.1.1	10452	10546	+1.0%
Lightsmark2008 SLI	716	733	+2.4%
Lightsmark2008	652	732	+12.3%
Unigine Tropics D3D9	1879	1920	+2.2%
Unigine Tropics D3D9 SLI	3372	3473	+3%
Unigine Tropics D3D9 SLI + MSAA 8X	2678	2703	+1.0%
Unigine Tropics OpenGL	1631	1625	-0.3%
Unigine Tropics OpenGL SLI	936	942	+0.6

Conclusion: under Windows XP, there’s a little performance gain but if you are fine with ForceWare 186.18, it’s not necessary to update them unless you want to enjoy the new features like power management or CUDA 2.3 (you need the 190.38 to play with CUDA 2.3 SDK samples…).