NVIDIA 3-Way SLI Performance Preview

Our Test Machines and 3DMark06

HOW WE CONFIGURED THE TEST SYSTEMS: We tested all of the graphics cards used in this article on either an EVGA nForce 680i SLI motherboard (NVIDIA GPUs) or an Asus P5E3 Deluxe (ATI GPUs) powered by a Core 2 Extreme QX6850 quad-core processor and 2GB of low-latency Corsair RAM. The first thing we did when configuring the test systems was enter their respective BIOSes and set all values to their "optimized" or "performance" default settings. Then we manually configured the memory timings and disabled any integrated peripherals that wouldn't be put to use. The hard drive was then formatted, and Windows Vista Ultimate was installed. When the installation was complete we fully updated the OS, and installed the latest DX10 redist and various hotfixes along with the necessary drivers and applications.


The HotHardware Test System

Core 2 Extreme Powered


Motherboard -

Video Cards -

Memory -

Audio -

Hard Drive

Hardware Used:
Core 2 Extreme QX6850 (3GHz) 

EVGA nForce 680i SLI
nForce 680i SLI chipset

Asus P5E3 Deluxe
X38 Express 

GeForce 8800 Ultra (2)
GeForce 8800 GTX (2)
GeForce 8800 GT (2)

GeForce 8800 GTS (2)
Radeon HD 3870 (2)
Radeon HD 2900 XT (2)

2048MB Corsair PC2-6400C3
2 X 1GB
2048MB Corsair DDR3-1333 C7
2 X 1GB

Integrated on board

Western Digital "Raptor"

74GB - 10,000RPM - SATA

OS - 

DirectX -

Video Drivers

Synthetic (DX) -
DirectX -
DirectX -
DirectX -
DirectX -
OpenGL -

Relevant Software:

Windows Vista Ultimate

DirectX 10

NVIDIA Forceware v169.18
ATI Catalyst BETA v8.43

Benchmarks Used:
3DMark06 v1.0.2
Company of Heros - DX10
Crysis - DX10
Half Life 2: Episode 2*
Enemy Territory: Quake Wars*

* - Custom Test
(HH Exclusive demo)

NVIDIA Monitor - Full System Load Burn-In

Performance Comparisons with 3DMark06
Details: www.futuremark.com/products/3dmark06

3DMark06 is the most recent addition to the 3DMark franchise. This version differs from 3Dmark05 in a number of ways, and includes not only Shader Model 2.0 tests, but Shader Model 3.0 and HDR tests as well. Some of the assets from 3DMark05 have been re-used, but the scenes are now rendered with much more geometric detail and the shader complexity is vastly increased as well. Max shader length in 3DMark05 was 96 instructions, while 3DMark06 ups that number to 512. 3DMark06 also employs much more lighting and there is extensive use of soft shadows. With 3DMark06, Futuremark has also updated how the final score is tabulated. In this latest version of the benchmark, SM 2.0 and HDR / SM3.0 tests are weighted and the CPU score is factored into the final tally as well.



Our very first benchmark score out of the gate should tell you exactly where 3-Way SLI is useful and where it is not.  We've provided you a baseline here of more standard, "mainstream" resolutions and AA settings to see the performance profile of 3-Way SLI at these settings and then at extremely high resolutions and image quality settings, where 3-Way really shines.  We'll continue to show you all the datapoints like this along the way in this article, to give you the full perspective.

At the default 3DMark06 setting of 1280X1024 with no AA on, there is hardly any gain with 3-Way SLI.  However, if we specifically brake out the Shader Model 3 test, when turned up to a full 2560X1600 resolution as you would run on a 30" LCD monitor, with 4X AA, we see sizable gains for 3-Way SLI.  Standard dual-GPU SLI is about 90% faster than a single card in this test and with three 8800 Ultra GPUs slicing up the workload, 3-Way SLI is 165% faster than a single GeForce 8800 Ultra card.

Related content