Introduction
It's been over four months since NVIDIA introducted their G80 graphics processor, powering first two, and now three, GeForce 8-series products. Aimed squarely at the upper end of the performance and price segment of the consumer graphics card market, our initial G80 analysis took in the architecture before we followed up with image quality not long after.
We promised you the last article in the trio -- this one, where we look at G80 performance -- would come in short order. We misjudged what short order actually means by some margin, referencing again those expired four months mentioned in the opening salvo. Whoopsy, shouldn't happen again.
Regardless, that does give us some time to reflect that in that quarter and a bit of time gone by, nothing's sat up to challenge GeForce 8800's positioning at the top of the tree, be it realised features via the architecture, image quality, or performance. There's also been no mid- and low-end introduction of the same base architecture either. Thus our performance evaluation is timely, sliding into view just before AMD drop the D3D10 bomb and NVIDIA fill out their D3D10 range.
We mentioned in the architecture analysis piece, where we touched on performance analysis briefly, that, "If there's a major performance flaw in what it's capable of, we're still looking for it, and we've been able to realise theoretical rates or thereabouts in each of the main processing stages the hardware provides, under D3D9 at least.". This article seeks to see if that statement still holds true.
To that end, we sat down with a gaggle of G80s, major competition from NVIDIA's own pre-G80 high-end stable and a SKU from AMD (quiet at the back!) to see how the two G80 launch SKUs compared to their immediate peers. We'll cover GeForce 8800 GTS 320MiB separately.
We take a look at AA and AF scaling on pages 14 to 16, general shading on page 17, and game performance from page 3 onwards towards the tail end of the article, should you wish to skip ahead while we implement correct page tite indexing.