Now that we’ve seen the speed of the GeForce GTX 580 1.5GB
, our attention naturally turns to the future. Unfortunately, while we know some interesting things about the forthcoming ATI Radeon HD 6900-series, we’re still not allowed to tell you anything about it, so I’ll focus instead on what I think the rest of the GeForce 500-series might have in store.
There are two worrying aspects to the GeForce 400-series if you’re an ATI employee of supporter – many of the the GPUs have had disabled units, and they’ve proved to be roughly 25 per cent overclockable. Add to this the new fp16 capabilities of the GF110 re-vamp that we should assume the whole range will use and you’re looking at some frightening numbers for the Radeon clan.
If we say that every GeForce 500-series GPU will have proportionally the same frequency increase as the GTX 580 1.5GB has over the GTX 480 1.5GB, and if each GeForce 500-series has one extra SM over its 400-series counterpart (as the GTX 580 1.5GB does), we could see the GeForce GTX 570 1.3GB being 29 per cent faster than the GTX 470 1.3GB. That should be enough to overhaul the Radeon HD 5870 1GB in most tests, but that’s not where things are really interesting – when we move to the mid-range, ATI could be truly scuppered.
If we stick to the same conjecture – one extra SM and proportionally the same extra frequency – we see a GeForce GTX 560 GPU that’s up to 38 per cent faster than the GeForce GTX 460 GPU. What’s more, as the chip would be the same size as the GF104 of the GTX 460, a GTX 560 would cost the same, giving us cards that cost £150 to £180. If this conjecture is even remotely true, the Radeon HD 6850 1GB and HD 6870 1GB will look laughably underpowered in comparison.
So, what can we expect from the GeForce 500-series? It could well be seen as one of Nvidia’s most brilliant GPU ranges since the GeForce 8-series. If so, ATI will have failed to take full advantage of the DX11 monopoly it was gifted late last year.
Click to see my (fairly rough) working for the above speculation.