Tein yhdelle forumille postauksen enkä sitä nyt jaksa Suomentaa, mutta tälläisen testin tein ja tämä on copy/paste postauksestani maailmalla eli kokemukseni uudesta DLSS AI-mallista.
I just made some tests. EVO must be forced to use newer (DLSS4 based) super resolution AI-model on newest nVidia APP with 572.16 drivers.
Original DLSS C-model (old CNN aka Convolutional Neural Network -model):
And my forced DLSS K-model (new Transform -model):
Picture quality is SUPERB on K-model DLSS with same settings! Absolutely *****/*****.
You can see it already on for those my little pics, asphalt textures are much sharper, on track/live everyhing is so sharp textures now. It takes about minus 5-10% performance, BUT it looks like an without Super Sampling > Native, or even better.
You can already found on some youtube videos that on some games, it actually looks better than native gaming, and i believe it now.
1 solo lap on Mt.Panorama, on same driving lines, apex to apex, multiple times and Steam+ACE restart after changing graphic settings or DLSS model on nVidia APP.
Ultra settings = GPU usage 99% ALLTIME = bottlenecking...
DLSS Quality NCC C-model : Avg. 151fps, 1% 104fps
DLSS Quality Transformer K-model : Avg. 141fps, 1% 99fps
Low settings = GPU usage 96-99%...a little bottlenecking
DLSS Quality NCC C-model : Avg. 257fps, 1% 204fps
DLSS Quality Transformer K-model : Avg. 227fps, 1% 183fps
I also tried on K-model Balanced settings, this also swipes NCC Quality settings, its also looking so much better than NCC Quality and almost same fps (Ultra settings 146/101fps).
Edit. I see on K-model 2-3 minor graphic errors, dunno its a these 572.16 drivers or a DLSS model? But just an blink of eye, doesnt bother to me at all.
And we must remember, there are currently
no nVidia drivers with EVO support, maybe next drivers.