I know you are being sarcastic but that is a genuine surprise for me. I thought that devs would be able to outclass the driver easily considering the app knows what it needs and when unlike the driver which sometimes has to assume a lot of things about the game's needs.
So yeah, seriously I'm disappointed that DX12 is behind quality DX11 drivers. Considering top end devs have worked with such APIs before I was expecting a much flatter learning curve and a better experience right off the gate.
I was proven wrong, sadly. Although my experience with DX12 (ROTTR) is very positive, it does run better overall, slightly less in pure GPU bound scenarions (2-3 frames less on my 980).
Well, consider that DX11 drivers are 6.5 years old and were already mature 5 or so years ago.
D3D12 devs weren't really doing this type of work in their renderers previously - on a couple of fixed console platforms yes, probably, but not on PC where you have like 50 different GPU configurations at any given time which may easily explode into >500 variations if you add different RAM sizes, PCIE bandwidths and such stuff. Thus their typical present approach to PC as just another "console" doesn't really work and do produce a lot of unpredictable results even on configurations which are considered mainstream at the moment. They have to build that part of the DX11 driver in their applications and it has to be as good at handling the different PC h/w as DX11 drivers are. So it will take some years before D3D12 renderers will reach DX11 drivers here.
What's important though is that theoretically the dev should be able to build a better resource management system as he knows better what is actually needed by the application in what moment - DX11 drivers are just "guessing" this so they are fundamentally flawed here and can be beaten. But this will take a lot of time and effort on part of the development community.
Seeing such half-assed ports like QB is just plain sad as this is exactly the point where both Remedy and MS would've probably done much better if they'd just used DX11 for the PC version. D3D12 does require A LOT of effort from the developer. If all they wanted was a quick'n'dirty port from XBO then they should've used DX11 or at least given us the option to choose the API path.
I have not been talking about how performance optimization is or isn't on Remedy, or about if Nvidia is suggesting installing older drivers. Have they even ever done that? Even with utterly broken Division release drivers they didn't warn people after issues were found or suggest keep using old drivers, they just kept silent and released fix as beta drivers few days later.
I did rant about shit that NV has been shoveling as certed drivers and mentioned how I think it's cool that Remedy actually suggests using driver version that is quite widely atm considered to be best choice for NV users. Should Remedy be concerned about games general performance and maybe look into working together with NV & AMD? Sure, but not the point I was arguing.
That's completely offtopic though and that's something rather arguable as I personally have no issues with latest NV drivers - no more than usual anyway - and consider the R364 branch the best choice on my system as it does have several new features (Vulkan support for example).
What Remedy should do though is make sure that their renderer works fine with the latest driver branch from NV, not suggest some older version because of reasons. Other games seem to work fine on R364, RotTR DX12 got a boost on them but yeah, we'll suggest going back two months because some newer driver had installation problems on system with 2+ displays.
To me this just shows that they haven't really tested the release version of the game on NV h/w since 362.00 drivers were released back in February. That's how I read this recommendation of theirs.