don't understand how many of these whql drivers released by nvidia. quantity=quality. cant manage to update them. also why nvidia don't use drivers number with a sequence like 388>389>390.

entr0cks:

Nvidia should do something about their drivers - this one doesn't interest me the slightest - G-sync fix, support for a VR game and support for the latest 3k priced card. Getting notifications from Windows and Geforce experience to update. Each one around half a gig of data. Same time old drivers remain installed and have junk in like 3-5 locations, adding up to several dozen GB of JUNK. Not excellent for solid state storage.

You can delete the old ones, genius.

ocsystem:

why nvidia don't use drivers number with a sequence like 388>389>390.

Because we would end up with 5-6+ digit driver name hell otherwise. Branches and sub-branches make more sense for developers and consumers especially if new feature is introduced.

Hitman's D3D11 renderer is also slower on this one, D3D12 is same'ish.

ocsystem:

also why nvidia don't use drivers number with a sequence like 388>389>390.

Because they do? Or is it a joke post?

Vidik:

Because we would end up with 5-6+ digit driver name hell otherwise. Branches and sub-branches make more sense for developers and consumers especially if new feature is introduced.

Most probably NVIDIA follows common scheme "major.minor" here (not "branch.subbranch") https://en.wikipedia.org/wiki/Software_versioning

dr_rus:

Another game with a 100% reproducible performance loss - but again hard to say if it's this driver's fault as both the game and Windows were updated since the previous benchmark: https://i.picpar.com/jdBc.png

Maybe reset ingame cfg then back to your old settings, or from idk ultra to high (apply go back in-game), then change back to ultra. This fixed it by FC4, it looked like there was some cache glitch or the way new driver accessed it.

Judging by this 388.53 vs 388.43 https://www.3dmark.com/compare/aot/239255/aot/238594/aot/235756/aot/172950 R388 branch has lower DX12 calls per frame . I added some random 4th driver R375 and it looks like it had the same higher DX12 call per frame

-Tj-:

Maybe reset ingame cfg then back to your old settings, or from idk ultra to high (apply go back in-game), then change back to ultra.

Why? Never had to do this previously and see no reason why I should do this now. Games settings didn't change in any way and I do reset them when benchmarking anyway. ROTTR is another one affected, although in its case it seems to be about 50% on the new 811 patch which managed to tank performance pretty universally while adding... nothing. So basically I'd say that this driver has some issue in CPU utilization which leads to lower results in D3D11 games which are mostly CPU limited. @ManuelG, you guys should take a look at this, too many games are showing loss of performance compared to .31 release when running in low resolution in D3D11 (examples: Shadow of War, Far Cry Primal, Hitman, Rise of the Tomb Raider, and I'm sure there are more). And to end this, here's a beautiful frametimes graph from ROTTR built-in benchmark in b811.3_64 running on 388.59: https://i.picpar.com/BdBc.png How these obvious things get past QA is beyond me.

Why, why.. Just try it, well I didnt have to do it either until I installed certain driver by FC4. Edit: Will check ROTR @ default v.high dx12 exclusive, fxaa and monitor frametimes But overall I didnt see any regression yet, except small boost by 3dmark FS that somehow got lost with 388.43 & .31.

-Tj-:

Why, why.. Just try it

No, I won't since it's obviously snake oil. When many games show the same issue no amount of changing their settings will help to solve it. It's either the driver or the last 1709 build update.

ok whatever, anyway ROTR looks ok here. 2 runs https://s6.postimg.org/a1grzhdk1/2_runs.png

dr_rus:

The trend is x80 card being released before Titan so this definitely break the trend somewhat. I do expect 2080 to be faster than this Titan V in gaming too.

You expect a GV104(Performance chipset) to be faster than the High-end GV100 chipset ?

Memorian:

You expect a GV104(Performance chipset) to be faster than the High-end GV100 chipset ?

Its possible. The GV100 wastes a lot of space and therefor power on features that don't benefit gaming (Tensor Cores, DP FLOPS, etc), which results in a gaming GPU to be clocked significantly higher. At the very least a GV102 "Ti" card will be faster in games, a GV104 is hard to predict how many shaders they give it.

@entr0cks even most "normal" users will uninstall previous software/drivers, you didnt. most experienced ppl use DDU to get rid of any leftovers anyway, so yeah. besides that, most files overwrite others, so ending up with this much "leftovers" shows its NOT a driver issue, but the way you update the rig, which almost everyone would have reimage or clean installed by now anyway (if only to gain performance back and install with newest updates/drivers from the start (vs just updating existing ones).

I cant race them, nvidia is way to fast for me vs my old fiat panda 500 and there bugatti chiron, they pooping drivers by the day lol. I just updated to 388.43 this is mental.

I'm going to wait for the next drivers in about 2 weeks which will hopefully come with PUBG optimization now that has a release date. I have to say the previous 2 Nvidia drivers have been spot on for me, but only when uninstalling using DDU for some reason, otherwise I get G-Sync problems.

entr0cks:

Nvidia should do something about their drivers - this one doesn't interest me the slightest - G-sync fix, support for a VR game and support for the latest 3k priced card. Getting notifications from Windows and Geforce experience to update. Each one around half a gig of data. Same time old drivers remain installed and have junk in like 3-5 locations, adding up to several dozen GB of JUNK. Not excellent for solid state storage.

Run driver store explorer and delete old drivers: https://driverstoreexplorer.codeplex.com/ Run DDU and delete the driver installed, then extract the exe of the installer for the driver you want to install, make a .txt file with this in it rd /s /q Display.Optimus rd /s /q Display.NView rd /s /q Display.Update rd /s /q GFExperience rd /s /q GFExperience.NvStreamSrv rd /s /q GfExperienceService rd /s /q HDAudio rd /s /q LEDVisualizer rd /s /q Miracast.VirtualAudio rd /s /q MS.NET rd /s /q MSVCRT rd /s /q Network.Service rd /s /q nodejs rd /s /q NV3DVision rd /s /q NV3DVisionUSB.Driver rd /s /q NvBackend rd /s /q NvCamera rd /s /q NvContainer rd /s /q NvTelemetry rd /s /q NVWMI rd /s /q NvVAD rd /s /q NvvHCI rd /s /q PhysX rd /s /q ShadowPlay rd /s /q ShieldWirelessController rd /s /q Update.Core change the extension to .bat, stick it in the folder where you extracted the driver, run it, - it will delete 800MB of garbage from the installer then install the driver with the setup.exe. (Essentially just delete all folders except NV12 and Display.Driver) Bringing it down from something like 1.2GB extracted to around 400MB, if compressed with 7zip it's 300MB exactly. You can also replace the old 1.0.61 Vulkan runtime in the Display.Driver folder with 1.0.65.1 released today which you can find in the Vulkan API here: LunarXchange Now you're left with pure Nvidia drivers and only the Control panel. If you want to save a couple gigs more space download Image health https://toolslib.net/downloads/viewdownload/248-image-health/ and run the extended garbage cleanup it will take a long time but your WinSxS folder will go back down below 7GB from whatever it's at. (Can grow over 20Gb...)

bad drivers for ac origins,stuttering mess. back to 388.43 and fine again

dr_rus:

And to end this, here's a beautiful frametimes graph from ROTTR built-in benchmark in b811.3_64 running on 388.59: https://i.picpar.com/BdBc.png How these obvious things get past QA is beyond me.

I tested RotTR and everything played fine. There were a few skips with the DX12 version, but that's to be expected as the DX12 renderer isn't very high quality. The DX11 renderer on the other hand was smooth as butter and had higher framerates as well.

kilyan:

bad drivers for ac origins,stuttering mess. back to 388.43 and fine again

Tested AC Origins and had no problems at all. Very smooth with these drivers.