facebook rss twitter

Review: NVIDIA's SLI - 6800 GT Performance

by Ryszard Sommefeldt on 24 November 2004, 00:00

Tags: NVIDIA (NASDAQ:NVDA)

Quick Link: HEXUS.net/qa42

Add to My Vault: x

Fillrate and geometry performance, GL_EXT_reme

We already know that AFR is the only way to double geometry performance with NVIDIA's SLI and that it provides similar theoretical increases for fillrate and pixel power. Let's examine those improvements for the 6800 GT, using AFR and a few theoretical tests. The test system is the same as you'll find in the first article, here. The GTs are clocked at 350/1000 and use the same 66.93 driver.

Fillrate

Fillrate

An obvious graph, but nice to see that GTs have no performance issues with (theoretical) fillrate.

Geometry performance

Geometry performance

The geometry performance scales well with AFR on both card types, hidden surface removal less so but performance still sees a fine boost from SLI.

GL_EXT_reme

Humus' tool works really well with SFR as a theoretical performance measure for that SLI mode.

GL_EXT_reme

Fillrate tests are accelerated well, the overdraw tests are similar. Transform and lighting performance is a near 100% increasing using Ultras, slightly less acclerated with the GTs.

So depending on the theoretical test used, you can see acceleration for both basic SLI modes. Scaling is evident between Ultra and GT, reflecting the performance differences in the two SKUs.

And the end result of it all is to make your games go faster. Let's peek at Far Cry, Half-Life 2 and Doom3 performance.