Presbytier
[H]ard|Gawd
- Joined
- Jun 21, 2016
- Messages
- 1,058
Since when did AMD dominate the sub $300 range? Also does that mater, and is it that large of a market? Because if it that large how is Nvidia share so much higher than AMD? Lastly where in the world did we get the idea a card that can barely outperform a 970 will outperform a 980?Like I've said before, I don't think amd gcn will ever be as efficient as nvidia gpus, even after accounting for different fabs. Because AMD packs more "stuff" onto their cards. polaris has two hardware schedulers, is there an equivalent on pascal cards? That alone take additional power, and who knows what else is on the board that draws more resources from nvidias more stripped down gpus.
That's not the worst thing ever, we saw that the initial strength of a stripped down design with maxwell had real costs in dx12 performance and concurrent graphics/compute workloads. I still don't think pascal can do the latter, but from what I read the penalties are far less severe due to better preemption of the gpu workload (or something, not exactly sure what is going on that is so different with pascal, maybe it's just high clocks). And that is the rub, it's OK to have a gpu design that is inherently more power hungry if you GET something for that power. AMD did in the gcn vs maxwell and earlier matchups, even with the nvidia defenders lying about 970s all the way to the end, but with pascal? We are not seeing the advantage there aside from select games like hitman.
None of this matters at the moment, amd still completely dominates the sub 300 dollar segment, and the custom AIB cards will likely push polaris past 980/390x levels, but where will the 1060 lie?
Going off the guesswork in raw cuda cores and rumored die size, will it be around 2/3 the performance of a 1070? (that would put it ~ between a 390/390x) or higher? Who knows.
If we go by the shady/blundering die size math people like ledra threw up, then it might be as high as 75-80% of a 1070. We'll see soon enough.