GeForce 6800 GS, GT go head to head

I explained at some length in my GeForce 6800 GS review how the new GS cards were essentially performance clones of the GeForce 6800 GT at a lower price point with fewer transistors. The math is pretty straightforward: the 6800 GS has fewer shader units and the like, but they run at a proportionally higher clock speed to yield very similar theoretical performance peaks. Unfortunately, our benchmark results in the article didn't include any scores for the old GeForce 6800 GT for direct comparison. I had intended to include those numbers, but my 6800 GT card was having some stability problems. I did manage to revive the thing, though, by feeding it a potent cocktail of Paxil and valium, and I've now added 3DMark05 scores for the 6800 GT to the review. I've got other things on my plate, so there's no time to subject the GT to a full range of tests, but the 3DMark results are still illuminating.

In terms of pixel throughput, the 6800 GS trails the 6800 GT in theory, but not really in practice. 3DMark05's synthetic tests show the 6800 GS outdoing the 6800 GT in single-textured fill rate. That's something of a shock given the fact that the 6800 GS has half the ROPs of the 6800 GT (eight versus sixteen) and consequently only 3400 Mpixel/s theoretical peak pixel fill rate to the 6800 GT's 5600 Mpixels/s. ROPs don't appear to be the bottleneck for these NV4x chips in this test. Perhaps somehow the GS's higher GPU clock speed gives it the edge here. (Could it be the higher clock speed in the memory controller?) The 6800 GT makes good on its slight (500 Mtexel/s) multitextured fill rate advantage in the multitexturing test, though.

As for the other synthetic tests, well, the 6800 GS ties the GT in pixel shader throughput and outdoes it in vertex shader throughput. The pixel shader results are more or less consistent with what we saw in the fill rate tests, and the vertex shader results aren't unexpected; the 6800 GS does have a bit of an advantage in terms of vertex shader cycles per second (2100M for the GT versus 2125M for the GS).

3DMark05's overall score and its component game tests are a virtual dead heat between the GeForce 6800 GS and GT. The GS edges out the GT by the slimmest of margins at lower resolutions, but older brother makes a comeback at 1600x1200. The especially tight numbers in 3DMark's game tests suggest that the real-world, functional difference between a 6800 GS and GT would be pretty much nil.

You may have noticed that I did manage to sneak the 6800 GT into our power consumption and noise level testing before the review first went up. In both categories, the 6800 GS showed up its elder sibling, consuming less power and producing less noise.

Incidentally, when I wrote my conclusions about the 6800 GS, I mentioned the possibility that the GS would be discounted from its suggested retail price, but I have to admit that I didn't expect to see the things selling for $209 at Newegg that very day. At that price, the GS is a quite a deal.

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.