The AMD Driver Thread

aceytrixx · 9 Jul 2015 at 18:36

So the issues with witcher 3 wont record. As soon as I hit record they seem to stop and my framerate drops to 52/53fps and is really stuttery. Theres also black artifacts ocasionally on screen

Gregster · 9 Jul 2015 at 18:55

3 games tested between 15.15 and 15.20 on the Fury X.

Superpatsy · 9 Jul 2015 at 18:58

What's the forum that Matt does support on now?

Found an issue with Far Cry 4 and don't ever seem to see Joe here...

humbug · 9 Jul 2015 at 18:59

^^^^ https://community.amd.com/community/support-forums

Gregster said:
3 games tested between 15.15 and 15.20 on the Fury X.

There are 3 different versions of 15.20, which are those? also can you test 15.7?

Gregster · 9 Jul 2015 at 19:06

flopper · 9 Jul 2015 at 19:10

MyBrains said:
I had to use the UNi-Xonar-1822-v1.75a-r2 driver for my Asus Xonar card to work.

same works flawlessly tho.
essence stx for me.
build 166 downloading.

Lokken86 · 9 Jul 2015 at 19:14

Gregster said:
3 games tested between 15.15 and 15.20 on the Fury X.

Is that the right way round or are these new drivers giving less performance then before?

Gregster · 9 Jul 2015 at 19:15

Lokken86 said:
Is that the right way round or are these new drivers giving less performance then before?

Yer, I checked and double checked to make sure, as I was questioning it as well but it is the correct way round and I have learnt what happens when I make a mistake on the internetz

Lokken86 · 9 Jul 2015 at 19:16

Colour me disappointed then.

Gregster · 9 Jul 2015 at 19:27

Lokken86 said:
Colour me disappointed then.

For other cards, there could well be some nice improvements but for me on the Fury X, it is actually a downgrade.

uksoldierboy · 9 Jul 2015 at 19:35

performance doesn't seem any better or worse for me using 15.7 drivers on my Fury X.

MyBrainz · 9 Jul 2015 at 19:35

flopper said:
same works flawlessly tho.
essence stx for me.
build 166 downloading.

Yea Asus drivers dont even pick up my soundcard in Win 10 :rolleyes:

Done a quick API test on the new 10166 build for Win 10, just margin of error in the results.

And anybody on Win 10 that installs a new build, watch out for MSI AB if being used, it will reset the Official Overclocking Limits and uncheck that option, well at least with me it did.

UTmaniac · 9 Jul 2015 at 19:35

Gregster said:
3 games tested between 15.15 and 15.20 on the Fury X.

Running a replay to benchmark pCARS WILL NOT give you representative figures for how the game runs in real-time. Replays do not use the 600Hz physics engine, a replay will reduce the likelihood of CPU bottlenecks.

Have you put "-pthreads 4 -dx11mt" (without the "") in the launch commands for pCARS in Steam this time? By default, the physics run on 2 threads and the graphics rendering runs on a single thread, my old pc benefits from these commands, on your pc with more cores I imagine they will have a big gain.

r7slayer · 9 Jul 2015 at 19:36

From gregster's video you can actually see performance is slightly worse on the updated version of the driver. But nothing which is noticeable ingame tbh but by watching the FPS counter you can see the dips and slightly lower FPS. Weird!

Gregster · 9 Jul 2015 at 19:40

UTmaniac said:
Running a replay to benchmark pCARS WILL NOT give you representative figures for how the game runs in real-time. Replays do not use the 600Hz physics engine, a replay will reduce the likelihood of CPU bottlenecks.

Have you put "-pthreads 4 -dx11mt" (without the "") in the launch commands for pCARS in Steam this time? By default, the physics run on 2 threads and the graphics rendering runs on a single thread, my old pc benefits from these commands, on your pc with more cores I imagine they will have a big gain.

All my vids are run in vanilla form and with maxed settings. Joe Public will see what to expect and what they can achieve with a similar system. If I start altering things, it is no longer "out the box". I have already changed the default NCP to "prefer max quality" and that wasn't the idea either but to put it more on a level playing field, that was what I did.

Sorry but I shan't be changing anything else and this is the way it will be done.

Mei · 9 Jul 2015 at 19:54

greg hijacked my help thread with his dumb car games!
rocket league > p00py pcars!

Gregster · 9 Jul 2015 at 19:56

Mei said:
greg hijacked my help thread with his dumb car games!
rocket league > p00py pcars!

lol.

LoadsaMoney · 9 Jul 2015 at 20:01

Performance drop in Witcher 3 is normal with these drivers, its in the release notes

Gregster · 9 Jul 2015 at 20:03

LoadsaMoney said:
Performance drop in Witcher 3 is normal with these drivers, its in the release notes

And I showed it as proof that the release notes weren't telling porkies. Well done AMD on a successful downgrade

MyBrainz · 9 Jul 2015 at 20:10

Ran the API Overhead bench again with Intel iGPU disabled in BIOS.

Win 10 (build 10166) 1046 drivers.. DX11 & Mantle scores have gone up quite a lot from my first run..

With iGPU enabled..

DX11 Multi-threaded draw calls per second 1 067 561
DX11 Single-threaded draw calls per second 1 099 283
DX12 draw calls per second 12 480 841
Mantle draw calls per second 11 361 807

With iGPU disabled..

DX11 Multi-threaded draw calls per second 1 044 912
DX11 Single-threaded draw calls per second 1 087 773
DX12 draw calls per second 14 410 974
Mantle draw calls per second 13 026 674

And GPU-Z is showing more Dynamic Memory being used than Dedicated Memory whether iGPU is enabled or not. I'll show a screenshot of GPU-Z once Imgur site loads for me.

What the hell is going on here? This never used to happen on Win 7..

Im :confused:

Maybe this API bench is just hit n miss.