How so?Coming from you, that's high praise indeed!
I tend to get less stutters and jutters on my R9 290 at 1080p but both perform somewhat similarly.
GTX970 is unplayable at max settings on 1440p, R9 290 is leaps and bounds ahead at that res.
How so?Coming from you, that's high praise indeed!
Not many people are in your position of being able to compare the two side-by-side, and there are lots of GTX970 owners who didn't experience what you're seeing, but I take you at your word.GTX970 is unplayable at max settings on 1440p, R9 290 is leaps and bounds ahead at that res.
GM204 hast 5.2 bln transistors, GM206 has 2.94 bln. That's 8.14 combined, both combined also have 6 64 Bit memory controllers, 3 MiB L2-cache, 6 GPCs and 24 SMMs. Subtract the fixed costs (PCIe, Display, Video) and it's close enough for my taste, given that the 8-bln-number from Huang probably was round and not quite exact. I do not think that there are major surprises on GM200 at this point.GK110 was 7B transistors. So at 8B transistors, I'm expecting about the same number of ALUs as the GK110 but with more SMMs than GK110 had SMXs.
$1500? It was $3000 in the beginning, it was quite late when they pushed it down to $1500 and even that was (at first) only for OEMsThe market didn't tolerate $1500 for the dual titan, so I doubt it would go over too well for a single GPU board. $1k is already a ludicrously high price point.
I think a lot of it depends on whether or not Nvidia will still market the Titan X as a card for semi-pros (people that supposedly wanted the then massive vidmem or higher DP-rate of Titan) or if they indeed keep DP at 1/32 (which would be my guess) and market it as a gamer card. If the latter, then I hope they keep their senses, price-wise.
It has the TITAN name so no it will not be a gimped card with 1/32 DP. It is expected to have a 1/4 ratio.
The Gamer version of Titan will be called 980 Ti and that card will probable have a lot of the DP fused off.
Most likely they are not planning version for HPC.There are rumors floating around that GM200 itself has a low DP rate. I'm not quite sure why NVIDIA would do this, but that's what the rumors say.
It may be imposed on them simply because of process realities. On one hands, their next flagship gaming chip has to be 28nm. On the other, it needed to be at least a decent margin faster than the tier below it. With gm204 400mm2, and a practical limit of around 600mm2, that doesn't give much options.There are rumors floating around that GM200 itself has a low DP rate. I'm not quite sure why NVIDIA would do this, but that's what the rumors say.
Not much of an option if your compute investment is CUDA based.Depending on price, they could get a FirePro W91x0 with 16 GB...
That's what you get for digging yourself down with a proprietary, closed solution.Not much of an option if your compute investment is CUDA based.
That's a valid point if the solutions are otherwise equal in terms of needed features, supported libraries etc. But we all know that Nvidia spends way more time on CUDA than AMD spends on OpenCL. You can't expect that to be of no consequence.
It may be imposed on them simply because of process realities. On one hands, their next flagship gaming chip has to be 28nm. On the other, it needed to be at least a decent margin faster than the tier below it. With gm204 400mm2, and a practical limit of around 600mm2, that doesn't give much options.