Project cars has sold 1/2mil on PC (steamspy/vgcharts), while Doom has sold 1mil for PC. Project cars has far outsold Assetto Corsa (~350K) and that game has been out a year longer, so if you're including a race game, it would seem Pcars is valid.
Dirt Rally 2016 has sold 330K. So again, you would go with the higher selling version by far right? Or you don't like cars because NV wins? Get over it. I say bench the highest selling games in each genre and more of particular genre's if they just awesome sellers vs other genre. Then again, I'd add that if a game doesn't sell 300K what is the point in wasting time benching it at all? There are far too many games above this. Should we concentrate on games nobody plays, or the most popular stuff? The answer should be simple. Should be concentrate on DX12/Vulkan (5-10 games...ROFL) or the other 99% of the market? Again, with DX12 only being on 19% of the machines out there (and that is by MSFT partner netmarkeshare numbers...LOL fudge factor says 10-15% tops for win10 IMHO and ZERO for enterprise), and a large portion of those being $199 laptop cheap pc junk (under $200 the OS is free) or people who accidentally had it forced on them (autoupdate etc...LOL) that can't run games, I'd say DX12 should be an afterthought at best. Vulkan too until I see at least a dozen games running it.
Even if you say 10% of the market has win10 and GAMES on that OS, what portion of those have DX12 hardware? Even if it is 50%, that would be 5% of the market. Benchmarking much stuff for that small a slice would be like wasting time on 4K in every review when it is <2.5% of the market and many games won't run 4K without turning details off which is against my religion. LOL. If you're benchmarking 10 games, it should be DX11-9, and DX12-1 because even that is being generous regarding market share of games available, OR market share of the OS with dx12 gpus. I care about what we DO today, not what we HOPE we do in a year or two. I could see running a DX12 ARTICLE testing all the games (since there is so few) and a vulkan game or two on 3-5 gpus or something (top card/mid from both sides or something). But not wasting time on it in every gpu review. That is overstating it's importance by orders of magnitude.
One more point on Vulkan Doom. There isn't really a point to benching it until it works properly for pascal IMHO.
http://www.hardocp.com/article/2016/07/19/nvidia_geforce_gtx_1060_founders_edition_review/4
While AMD is winning here, note NV isn't using the latest API yet. It remains to be seen how much improvement you get from 1.0.8 vs. 1.0.11.1 vs. 1.0.17.0. Since the libraries are loaded with NV's driver, this may just be a glitch that will be fixed shortly. Until they're on the same playing field I'm skeptical here.
http://www.pcgameshardware.de/Doom-2016-Spiel-56369/Specials/Benchmark-Test-1195242/
One more point showing something fishy with doom/pascal. 980ti should not be eating 1070 for breakfast! Clearly pascal is hindered by something. I'm thinking the glitch is massively hurting NV's new cards perhaps because Id obviously has had more time with the older gen which sold pretty massively by now.
https://community.bethesda.net/thread/54585?tstart=0
Bethesda faq, showing NV not working right yet, updates coming, and many cards with issues:
“Does DOOM support asynchronous compute when running on the Vulkan API?
Asynchronous compute is a feature that provides additional performance gains on top of the baseline id Tech 6 Vulkan feature set.
Currently asynchronous compute is only supported on AMD GPUs and requires DOOM Vulkan supported drivers to run. We are working with NVIDIA to enable asynchronous compute in Vulkan on NVIDIA GPUs. We hope to have an update soon.”
Kind of important to have the right stuff working or what is the point in benchmarking it and claiming a winner? The other product is CLEARLY not working right yet as shown by hardocp and the german site.
Also note, Hardocp shows doom easily runs out of 4GB in their review. So out goes the 4GB 480 IMHO and this will start happening more and more with new games. So no point in comparing 4GB 480 to 1060 obviously as some kind of bargain. I'm far more worried about hitting 4GB limits than dx12 perf. I would not buy a card under 6GB today unless I'm broke (and I'm not...LOL). It was wise of NV to just put out a 6GB 1060. No point in a 3GB that would likely be running 1080 most of the time and running out a lot since most devs are aiming 4GB or higher.
"This game will not run "Nightmare" settings with any video card less of 5GB of VRAM, it displays an error message and won't allow it." said hardocp above. As noted 980 was excluded due to this. IF buying a 480, you'd best get the 8GB model! Dynamic memory might save you some, but you're counting on the game dev, or AMD supporting it (or both?). Not sure yet about how that works, and I don't care until Vega hits and I look more closely at a buying decision and I may end up going the new titan for xmas if it's great at some app stuff I'm interested in now. I hope toms tests titan against some quadros for 3dsmax/maya/blender and maybe some AE cuda stuff too. Cuda NV vs. AMD OpenCL if you include an AMD card in there or whatever is fastest for each side in the chosen apps.
Most Titan's are not sold to gamers, but rather pro people on a budget who game as an afterthought also. IE, M6000 is $4K! Which makes the coming $1200 Titan a no-brainer if you're into content creation on top of gaming. I doubt its aimed at fp64 and hope not, as those people can buy a tesla! Content creation on the cheap is what I'd want from it and that makes $1200 very reasonable if it's anywhere near m6000 (or better in some stuff probably). M6000 would only make sense then if you were doing stuff like 4K vid with large data sets needing the 24GB, otherwise save some cash or heck, buy two and still save $1600...LOL. Can't wait to see Vega/Titan.