Experimenting with AMD graphics card settings...

I'm also surprised, in the past I loved the look of SSAA but gave up on it because of the performance hit. I tried it again and I am now able to get decent performance even with 4xSSAA using Catalyst Control Center to override settings. Maybe something is different in the current build? Thanks for your post!

 

Hi,

Thanks for the tips, I've got the fps up a fair bit by using ccc for af 8x/adjusting bezel also dropping the res down slightly.

Cheers,

Sean

 

playing with CPU overclocking and notcied this:

2014-01-29 21:49:39 - 4ghz
Frames: 4443 - Time: 60000ms - Avg: 74.050 - Min: 57 - Max: 102

2014-01-29 23:01:56 - 4.100ghz
Frames: 4313 - Time: 60000ms - Avg: 71.883 - Min: 51 - Max: 98

2014-01-30 15:55:49 - 4.200ghz
Frames: 4246 - Time: 60000ms - Avg: 70.767 - Min: 56 - Max: 92

2014-01-30 16:00:17 - 4ghz
Frames: 4426 - Time: 60000ms - Avg: 73.767 - Min: 57 - Max: 100

Only changed the Clock so far (no voltage or Bus change or anything else), anyone have a clue why a higher overclock = lower avg/max FPS? I'm not totally clued up on CPU overclocking so maybe missing something! ta

 

RFactor engine based games recommended AMD Catalyst Control Panel (CCC) Settings

I have noticed over the years that getting ISI based games to look great while using AMD graphics cards is very straightforward and always follows the exact same formula....

NOTE: If you don't feel like reading everything below then just basically make sure the in-game/game config AA is set to none/0x, and in your graphics card's control panel set AA to override app settings, with 2x, 4x or 8x, and type being supersampling. The key here is supersampling.


A. Anti-Aliasing Mode - Override Application Settings - Unlike most games, make sure to disable all game controlled Anti-Aliasing and leave it up to your graphics card's control panel.

B. Anti-Aliasing Samples - 2x, 4x, or 8x - 8x will most likely kill your framerate (supersampling AA method, see "D" below) even on new hardware and an older sim, but looks pretty much perfect. Everything, and I mean everything is sharp, clear, clean, defined. Fences, trees, painted lines on the ground, leaves, poles, wires, even when they are far off in the distance all look great, like a photoshopped/edited picture but in live gameplay motion of course . It makes, for example, Toban at RFactor 1 (w/ high res texture mods) look beautiful and real. Don't worry though because 4x is a VERY good compromise. 4x gives you much more frames and less slowdowns than 8x but it is very, very close to the image quality of 8x. 2x gives even more frames and is what I mostly use because I play with triple monitors and like frame rates to be between 90 & 120. Unlike the IQ of 4x which is hardly a reduction in IQ compared to 8x, 2x does see a noticeable decrease in IQ and you do start noticing the jaggies. I would always though, 100%, rather have 2x supersampling over even 16x or 32x non-supersampling (multisampling or adaptive multisampling) in most ISI based sims, see option "D" below for the reason why.

NOTE: Experimented tons with quality versions of AA for each respective AA sample amount, for ex. 4xEQ instead of 4x, etc. doesn't do much, if anything, for ISI engine IQ.

C. Filter - Standard - Experimented tons with edge-detect, doesn't do anything for ISI engine IQ, regardless of AA mode.

D. Anti-Aliasing Method - Supersampling - The image quality of ISI's engines hardly improve when using regular multi-sample AA even if you're using something insane like 16x or 32x, and unlike most DX9 games the adaptive AA setting (I believe NVidia calls this Transparency AA) hardly improves the image quality as well. This is where Super-sampling AA comes to the rescue, and boy does it ever. The problem with non-supersampling is that many of the things on screen don't get affected by your AA, so even setting the amount of samples to 10000x wont do anything because certain objects/things on screen are just not affected by the AA, period. Supersampling is different because it doesn't care about what's on screen, it affects the entire picture. It uses a technique where it sort of upscales the entire image to a higher resolution and that's why everything looks incredible with supersampling, but also why it kills frames especially with 8x.

NOTE: Raceroom Racing Experience doesn't seem as dependant on supersampling AA in order to get a clean picture as other ISI engine based games. Simbin seem to have worked the ISI engine in RRRE very nicely in terms of getting a pretty clean picture while not using supersampling.

E. Morphological Filtering - Off - AMD's version of FXAA. Generally speaking, I'm not a big fan of post-processing AA. They are POST-processing which means they affect the image after it's already been drawn so they don't really "talk" to the game much and therefore they affect and blur out entire objects/surfaces and textures themselves. FXAA in Battlefield 3 at 1080p turns the ground from insanely sharp, detailed and beautiful to looking like a blurry, console, low resolution, game from 2005. I might as well just buy a console or turn down my game's resolution to 800x600 if I am going to be using FXAA/Morph AA. SMAA is the only one that seems "intelligent" to not affect textures too much while still getting rid of jaggies, but it has to be "injected" with Radeon Pro and can be picky depending on the game you're using, settings, etc. Some games that have badly designed AA programming, and are too intensive to use supersampling AA with, do need a bit of FXAA/Morph AA though. For example, the FXAA in BF3, I still have to keep it on "Low", (1 tick from off) rather than off like I really want to, because BF3 is like RFactor, you can use 10000X regular AA but some things like power lines and many other things just don't get affected by the regular AA mode. Also, if you don't mind/don't notice the lower IQ of FXAA/Morph AA then, for the most part, it is a great and extremely efficient way (fps wise) of getting rid of jaggies. I imagine in a racing sim, as opposed to a shooter, that you would hardly notice or care for blurred textures if it means getting rid of jaggies without having to resort to frame-per-second killing supersampling.

F. Anisotropic Filtering Mode - ...
1. Use Application Settings - Set it in-game for the sims that have that option like RFactor 1, RFactor 2 and sims based on those engines (ex. Game Stock Car 2013, ARCA Simracing, Simulador Turismo Carretera, Formula Truck 2013, etc.)
2. Override Application Settings - For sims based on an older/custom version of ISI's engine with no in-game AF setting (ex. GTR2, GT Legends, Race 07 series, possibly RRRE, etc.).

G. Anisotropic Filtering Level - Use Application Settings unless if "F" (above) is set to 2 then set to 16x or 8x. In some games 8x may give a boost in frames while hardly being a reduction in IQ from 16x. I can rarely notice a difference in far off angled texture sharpness/blurriness (ex. the ground as it gets further and further away from you) between 8x and 16x, but the reason I keep it on 16x is because most games nowadays get like literally a 2 fps improvement when going from 16x to 8x so if that's the case then I might as well max the IQ. There are rare exceptions though, there is a sim (not ISI/RFactor based) that, to my surprise, receives some pretty large gains in fps when going from 16x to 8x, and even from 8x to 4x. Real rare though but it happens.

H. Texture Filtering Quality - High Quality...
I. Surface Format Optimizations - Off...
- My OCD makes me always put these to their highest IQ settings for just about all games, period (unless there's a special circumstance for a specific game, which I don't know of any as of now). However, don't be shy in experimenting. You may not notice IQ reduction and may notice fps improvements from changing those settings to "Standard"/"Performance" and "On".

The last 4 options are for vsync, openGL, and tessellation. Pretty much leave vertical refresh to "Off, unless application specifies" as the game itself should handle vsync on/off unless a special circumstance makes you need to force it from the control panel (very rare). OpenGL triple buffering I leave to default Off, but it's mostly irrelevant as the setting only affects OpenGL based games while most games (including just about every race sim out there) are DirectX based. The 2 tessellation options I actually keep on "Use application settings" instead of the default "AMD Optimized". It won't make a difference to almost any current racing sim, but for any games and benchmark programs that do use some tessellation I like to just use the pure/raw way the tessellation was designed in-game, without any graphics card control panel driver "optimizations" going on.


NOTE FOR RFACTOR 2 USERS: When using supersampling AA you can go ahead and disable Transparency AA in your PLR file in order to save some FPS. Supersampling AA takes care of all this, so using both offers no image quality benefits while still slightly negatively affecting your FPS (post #4 HERE confirms this, includes pics. Thanks Prodigy). Go into your PLR file and make sure the specific line says "Set Transparency AA=0" (you want it to equal "0" so it's disabled), then save. Make sure you remember to re-enable it (set it back to equal "1" ) if you decide to not use supersampling, or else your trees, fences, crowds, etc. will look much worse. Aside from disabling this while using supersampling AA in order to save some FPS, this should otherwise always be enabled.

 

Will have to give that list a run through later today Spinelli and see what happens. Will post any improvements I see.

 

The image quality of supersampling should be better than any other type of non-supersampling settings you use unless maybe you're using injectors, third party tools, etc. etc, then who knows what you're up to, but if only adjusting the in-game graphics options and the options available in your GPU control panel then I'm pretty sure you'll be satisfied with the IQ .

 

@Spinelli - Very good and informative post, thankyou

 

anyone? need help understanding this.


also during benchmark tests with the replay file I noticed that Soft Particles (dust) was different sometimes, as in replay files don't always show the same...which could cause the above but it's a big fps difference regardless.

 

I can't explain why you're losing frame rates but I will say that don't mess with voltage unless you've done quite a bit of research. If you're able to OC without a voltage adjustment, that is a good thing. Voltage adjustments without experience do like to cause system instability.

 

I've done a PRIME95 Benchmark Stock & Overclock (4.3ghz) and that shows a vast difference so the OC is working but strangely not for RF2 with the numbers posted above. Is it a Bottleneck? even with an OC'ed GPU?

yeah I dont mind small voltage tweaks, already doing that with my GPU.

 

Appears to be an RF2 thing, with AC I see an FPS increase with a higher overclock.

 

That's really super duper weird MarcG, I honestly don't know what to tel you other than use a monitor logging program to make sure your CPU isn't dropping/throttling.

 

Yes that does seem a little strange but you also said all you've done is change the multi so I assume all the other bios settings are on AUTO. If so as you increase the multi the board will increase the voltages (and adjust other stuff too).

What do you get at stock, although with turbo boost that will actually be 3.6-3.8GHz depending on cores used. You'll probably find FPS the same meaning you're GPU limited but it could be an interesting exercise if you like to fiddle around.

 

I followed this guide:
http://www.overclock3d.net/reviews/cpu_mainboard/gigabyte_z77_overclocking_guide/1

But as above the benchmark in prime showed improvement as did AC, cpu-z shows it's working in Windows.

 

The tests are good
I followed your instructions to the letter and I have rediscovered this beautiful post simulation.Ce deserves to be STICKY
Thank you

 

Two things:

- I've had to return my CPU to Stock settings as RF2 CTD on random occasions and never lasted more than 20minutes, disappointing as with AC I got 10 FPS extra (start of race) and that ran for over an hour without issues. Returning to Stock and RF2 is just fine, wonder if I can load a BIOS Profile per Game I'm using!?

- Tried Spinellis (decent) tips and even though I get an FPS loss it's not too great and does indeed make the game look better, thanks for that

 

For those looking for more performance and slightly better look, I've tested these two drivers over the weekend;

This was a Beta Candidate.
http://forums.guru3d.com/showthread.php?t=385614

and this driver...

http://support.amd.com/en-us/kb-articles/Pages/latest-catalyst-windows-beta.aspx

Both seemed to provide between 15-50 more FPS. Cars and tracks where I was only getting around 65-80FPS are now at 70-128FPS. I haven't seen rF2 hit consistent, over 100FPS on my system since getting it and my system, while not super powerful, is decent enough.

These are my settings;
View attachment 11600

View attachment 11601

View attachment 11602

 

Excellent detail! Do you write manuals for a living? You should! d'_'

I tried these settings. They were close in performance to the things I've tried. As mentioned by someone else, I think, I lost about 20FPS at times while viewing or playing but overall I think the graphics look better with your settings.

Now I have two options! This is good.

 

Thank you Spinelli, we did that tonight with friends (everyone with ATI) and that was better.

 

You're all welcome, you're all welcome