GUIDE AMD Control Center

[Besidilo]

What you'd probably want is a Radeon Pro guide for AMD users, as that one has way more configuration options.

[EisDrache]

Well I edited the wiki page and I think it looks pretty good.

[Farlo]

Well I edited the wiki page and I think it looks pretty good.

It does, thanks for the work!  I'll give it a few edits and see if we can't merge the two pages so they follow one-another at some point today.

[Farlo]

Made some heavier changes than I was expecting to, but I hope it looks even better. Also changed the name to "AMD Control Center" since it could be either. The images might not appear right for a couple minutes, I'm renaming them to something less generic than "Step#".

[Kelmych]

The new set of screens is great. Are the settings in screen 5 true for users in all three STEP categories (performance, baseline, and extreme)? We eventually also need to add some screens with the new GUI (perhaps from z, I have only one card myself) that include crossfire settings.

[EisDrache]

Well the settings that are change able in the AMD Control center most will be fine to run on any system the only setting that will be a noticeable change in performance is Anti-Aliasing but the rest can be used on any system without a real notice in performance drops

 

And thanks Farlo for the little changes that looks alot better i'm a noob when it comes to wiki editing.

[TechAngel85]

My only complaint would be, instead of "recommendations" on the settings, you are telling users to use a specific setting which isn't going to apply to all users. This is why I suggested to use the NVIDIA Inspector Guide as a reference on the OP of the Wiki Editors To-do List thread. I give users an explanation of what the setting does, provide a recommended setting, and provide information (where it's not common sense) as the effect of setting (better performance, FPS loss, etc). I have also provided performance data for most of the settings so users can better make an informed decision on the best setting for them. Telling users to "use this setting" was done very rarely in that guide.

 

This guide is nothing but "use this setting", which is potential for issues with more troubleshooting. "I used your AMD Guide and my game is now really slow. Please help!"... You guys have made it much more useful, but this guide needs to be more informative and geared towards recommendations rather than "set it to this and forget it" or we're going to be getting more troubleshooting issues from users. I would do it, but I'm a NVIDIA user which is why I spent a lot of time writing the Inspector Guide with plenty of details and information.

[DanielCoffey]

I have tried some benchmarks on the New Game cart ride into Helgen since they are fairly consistent and repeatable and have observed a few interesting points which may be worth discussing then adding to the Guide.

 

My Hardware

=========

 

GPU : Sapphire 7950 3Gb OC with Boost (860MHz with auto-OC to 925MHz)*

CPU : Xeon 6x3.33GHz

RAM : 12Gb 1333MHz ECC

SSD : 240Gb SATA-II

Resolution : 2560x1440

OS : Win7-64 Home

Drivers : Catalyst 13.1

 

* Running at PCIE 1.0 speed in a 16x slot

 

Firstly the Guide does not make any recommendations for the hardware to support 4xSSAA and I think this needs revising. I tried setting 4xSSAA on my 7950 3Gb as per the guide and it lowered the frame-rate to an unacceptable degree as well as maxing out the GPU. I was averaging 30fps and the GPU was flat out at 100% with the fans on max. As soon as I went back to 4xMSAA everything was back under control.

 

Secondly it appears that forcing Catalyst to override Skyrim's own AA/AF settings saves on VRAM considerably (by around 350Mb on my settings).

 

EDIT : I had failed to make sure that the SkyrimPrefs.ini AA and AF settings matched the CCC Override settings. Disregard these results.

 

Test 1 : Skyrim managing 4xAA and 16xAF at 1440p

 

VRAM Avg / Max : 1622Mb / 2048Mb

RAM Avg / Max : 1755Mg / 2114Mb

FPS Avg / Max : 56 / 60

GPU Load Avg / Max : 70% / 100%

 

 

 

Test 2 : Catalyst 13.1 forcing 4xMSAA and 16xAF at 1440p

 

VRAM Avg / Max : 1269Mb / 1669Mb

RAM Avg / Max : 1720Mg / 2078Mb

FPS Avg / Max : 56 / 60

GPU Load Avg / Max : 67% / 100%

 

 

I should add that I am using a 2010 Mac Pro to play the game and since I am using a PC 7950 flashed to support the Mac EFI I am currently limited to PCIE 1.0 but I understand that at this frame-rate it will not be the limiting factor. It is still a 16x PCI slot and will only suffer at the initial loading of bulk textures into VRAM.

[DanielCoffey]

No, I edited the ini files through ModOrganizer to turn off the AA and AF within Skyrim and just relied on the TESV.exe profile in CCC. That is correct, yes?

[Kelmych]

I don't see how having AF settings in 3 places changes RAM use; two of these settings are used to let Skyrim and the ENB know what is happening. It is actually happening in the video card and there is only one AF process.

[z929669]

Need to test repeatability of results by running 2 trials of each settings combo using Skyrim only, CCC only and both combined. To test SMAA, run this all again with SMAA/MSAA enabled/disabled.

 

It is pretty well known that SSAA costs much more than MSAA, but SMAA should be better than MSAA in terms of performance.

[DanielCoffey]

OK, I'll have a look tomorrow after installing the new 13.3 beta3 AMD drivers which are supposed to have a fix for some texture issues on my 7950.

[DanielCoffey]

I have re-run the tests on the cart ride and can confirm that when Skyrim is involved in setting the AA and AF, the VRAM usage reported by Skyrim Performance Monitor is higher than when just CCC is used. I also found that setting Triple Buffering was causing the GPU to work much harder which led to some GPU 100% flat peaks with shaky frame-rate so I turned it off for the performance gain. All the tests below were run at 2560x1440 with Triple Buffering OFF.

 

 

1. When CCC is overriding the AA/AF settings, should Skyrim .ini match it or be set to 0xAA 0xAF?

 

Test 1 : Skyrim setting 4xAA, 16xAF

 

VRAM Avg / Max : 1540Mb / 2000Mb

FPS Avg / Max : 56 / 60

GPU Load Avg / Max : 69% / 100%

 

 

 

Test 2 : Skyrim Setting 4xAA, 16xAF, CCC overriding with 4xMSAA, 16xAF

 

VRAM Avg / Max : 1651Mb / 2076Mb

FPS Avg / Max : 56 / 60

GPU Load Avg / Max : 70% / 100%

 

 

 

Test 3 : Skyrim setting 0xAA, 0xAF, CCC overriding with 4xMSAA, 16xAF

 

EDIT : No AA/AF was observed in screenshots in this test. SkyrimPrefs.ini NEEDS to match the CCC Override for AA to be visible.

 

VRAM Avg / Max : 1318Mb / 1702Mb

FPS Avg / Max : 57 / 60

GPU Load Avg / Max : 57% / 100%

 

 

 

 

We can see here that Skyrim Performance Monitor reports that whenever Skyrim is involved in setting the AA/AF (Test 1), more VRAM is used than when only CCC dictates the settings (Test 3). Also it appears that the CCC Override is not as thorough as we would perhaps like since Skyrim does indeed influence the AA/AF workload (Test 2) whenever it is set in the .ini files.

 

CCC Override with Skyrim set to 0xAA 0xAF also shows that the GPU load is reduced which in turn allows a smoother frame-rate.

 

From these tests I would advise that Skyrim .ini is set to 0xAA and 0xAF if the CCC Override is used.

 

 

2. What is the performance impact on the GPU of Super Sampling AA?

 

Now that I had disabled Triple Buffering, I had some performance headroom on my 7950 so I decided to switch back on SSAA as advised by the guide. I found that providing Triple Buffering was off, my card could handle 4xSSAA comfortably. I made sure the Skyrim .ini was set to 0xAA and 0xAF and allowed CCC to force the override for the AA and AF.

 

I have repeated the Test 3 result below for clarity here...

 

Test 3 : Skyrim setting 0xAA, 0xAF, CCC overriding with 4xMSAA, 16xAF

 

VRAM Avg / Max : 1318Mb / 1702Mb

FPS Avg / Max : 57 / 60

GPU Load Avg / Max : 57% / 100%

 

 

 

Test 4 : Skyrim setting 0xAA, 0xAF, CCC overriding with 4xSSAA, 16xAF

 

VRAM Avg / Max : 1329Mb / 1710Mb

FPS Avg / Max : 58 / 60

GPU Load Avg / Max : 53% / 100%

 

 

 

While the reported VRAM remained the same, the GPU usage actually dropped with 4xSSAA and this resulted in slightly smoother frame rates. Just to see what would happen, I put the Skyrim .ini settings for 4xAA and 16AF back into use and saw the GPU struggling to maintain an even frame rate. The VRAM shot back up too of course.

 

Test 5 : Skyrim setting 4xAA, 16xAF, CCC overriding with 4xSSAA, 16xAF

 

VRAM Avg / Max : 1629Mb / 2093Mb

FPS Avg / Max : 57 / 60

GPU Load Avg / Max : 68% / 100%

 

 

 

It seems that if your card can handle it (and you like the different appearance of SSAA over MSAA), there may be no practical performance difference between 4xSSAA and 4xMSAA.

 

The next test for me to perform will be to look at a pair of conditions (just Skyrim controlling AA and AF and then CCC forcing AA and AF) with Triple Buffering on and off.

 

 

EDIT : Apologies but these tests are invalid as no AA was present because I had turned it off in the SkyrimPrefs.ini. Disregard this half of the testing.

[DanielCoffey]

That was my STEP copy with some added Immersion mods.

 

I have no idea why I had turned Triple Buffering on in my earlier tests. I can only assume I had misread the AMD CCC Guide. The most recent five tests above are investigating what the guide advises - Skyrim AA/AF control vs CCC AA/AF control and looking at MSAA vs SSAA. Triple Buffering was off.

 

EDIT : Vanilla Skyrim with HDDLC, DG, HF, DB with CCC controlling 4xSSAA and 16AF shows VRAM avg / max at 1091Mb / 1300Mb, FPS avg / max at 58/60 and GPU load avg / max at 50% / 88%. Nice and comfortable.

[DanielCoffey]

Argh - now I feel like an idiot - you are quite correct! You DO need to mirror the settings in the SkyrimPrefs.ini file. I took some screenshots of sky against a roofline to check and it was immediately obvious.

 

If CCC forces an override but the SkyrimPref.ini is set to 0xAA 0xAF, that is what you get. No AA and no AF. Once the SkyrimPrefs.ini is edited to put back the AA and AF, then AA is visible. I do have to ask if the CCC settings are even applied at all. Some more testing would be needed.

 

Damn - I had got all excited and thought I could save a bit of VRAM and get a steadier framerate but no. Lots of testing and I never even looked at a screenshot to check if the AA/AF was even on at all.

 

Well spotted.