You see, this is where you are wrong... Ashes is used as a benchmark, because it is built from the ground up for DX12. So there is the time spy thing... OK... I would like to point towards Hitman as well... Currently even Tomb Raider have decent DX12 support... Warhammer have DX12 support... Gears and Quantum Break are not really reliable, but only counting 1 game as DX12 game is not correct at all... CIV 6 is coming, Deus Ex is coming, BF1, Star fucking citizen... The DX12 support is important...
If you have not checked my video, here is a slide...
And roughly calculated... So picking up the one or two benchmarks, where the 1060 is doing well in DX12 is not really representative... And i am comparing pure stock to stock... Not aftermarket well cooled no throttling... Keep that in mind.
Look, i would accept 1060 as a viable choice only if it is cheaper. Right now, knowing the history of Nvidia and AMD, i would say that those 6-10% difference in DX11 will be gone this time next year. I said that in my video. If you are planning to get new GPU at that time, fine, then 1060 is faster and will do the job. But if you intend to keep your GPU for more than an year, i rather get the 480, because of all driver improvements and obviously untapped performance, that AMD can tap in with driver updates and future APIs...
My point was not 2 games vs 6 games, my point was that the entire sample size of DX12 and Vulcan games is minuscule in comparison to what the sample size is for DX11. I mean shit, look at this list:
The way Vulcan and DX12 is implemented isn't even really final, I heard something about Microsoft changing how multi-gpu setups are implemented on DX12. The way these things are programmed and will be programmed going forward is still somewhat in its infancy, you can't say AMD or Nvidia will definitely be the way to go when it comes to DX12 and Vulcan. I mean look at this timespy result:
Everyone has been running around talking about how Nvidia doesn't benefit from A-Sync, well look at this test. The 1080 gained nearly 7%, and the Gtx 1070 gained a sliver more than 5% when comparing A-Sync off and on. That is still not the damn near 13% the Fury X gained or the 12% the RX 480 gained, but it does show that Pascal does actually make some good gains using A-Sync. I mean shit this battle isn't even near over, the drivers for DX12 and Vulcan are no more mature than the way the API's are even being implemented. I think its safe to say that GCN lends itself to A-Sync compute, but there is also some evidence to suggest that Pascal can also make significant gains while utilizing A-Sync. 7% is nothing to scoff at, even if the Fury X did nearly gain 13% as well.
I understand that... And that is my point. Knowing the history, I rather have the thing, that will potentially make 10-15% performance increase in the next couple years, rather than 5-6-7... Oh, I could be completely off, and the next DX12 update changes things and bumps Nvidias performance like it does amd... But you see, those points are not made in the press. OK, may be if I dig in 3 years of DX12/Vulkan coverage, there will be mentioning of them... But there is no current review or whatever, that digs into that. It either sais 8% editor's choice or whatever... And the questionable benchmarks, and all the gameworks features... I'd like to see balls to the wall review... AMD with hairworx and whatever they have vs Nvidia with physx and hairworx and whatnot... The press just follows the guides and some of them obviously kiss some asses, there were even a review, that said 480 was faster... Did they test Hitman and Blops 3? Dafuq... It is just so inconsistent...
The testing suites are really skewed as it sits, I don't think any reviewer is perfect rn tbh. I mean even if a reviewer had the perfect selection of titles, the way they test every title could skew everything going forward. Maybe they test a section that favors A-Sync or maybe they test a section that doesn't, it could alone be the difference between the winner. Its going to be very weird going forward. Reviewers need to be adding asterisks out the ass right now frankly, as this period between old DX11 and new, better API's is going to be very fluid as to who is on top. GCN may well pan out and who know, maybe the Fury X will actually be faster than the 1080 due to AMD figuring out how to squeeze every single drop out of that thing with all the A-Sync special sauce. I mean on paper the Fury X still has more FP32 performance than everything short of Titan XP, so that's actually not even as far fetched as I would initially think. (8.2 Teraflops on 1080, 8.6 on Fury X, & 10 on Titan XP) Its going to be a very weird, fluid market for the next 2 years. Drivers will in the end determine everything I guess.
