Experimenting with AMD graphics card settings...

I decided to sit down and do some tests to see if i could both improve the visuals of RF2 on my system and optimize my FPS at the same time and ended up with some interesting results which i hope will benefit other AMD card users.

My basic system specification is: i7 860 at stock speeds, R9 280X 3GB, 4GB ram, W7 64bit O/S.

I started by creating a replay of a race at Mills Metropark with 9 AI drivers which i used for my base testing. I made sure the load was high on the system by beginning my FRAPS benchmark from the moment the start lights went green and starting from last place. I then recorded FPS results for a sixty second period, tweaking settings until i found a good compromise between FPS and image quality. I made a custom game profile in CCC for Rfactor (pointed to the .exe in the RF2 core folder) and set the following settings:

AA to override App settings,
AA samples to x4,
Filter standard,
AA method Super sample
Morphological AA disabled.

AF to override App settings,
AF level to x8 (x16 lost me an average of 5fps measure with fraps with little to no visual improvement)
Texture filter to high quality,
Surface format optimization on.

In the RF2 display options i am running everything on the best/highest quality settings. I made sure to set the ingame AA and/or FSXX to off and also setting the AF to bilinear in game to off. Controlling them via CCC not only improved image quality but also improved FPS on my system. I quickly found that as in previous builds environment and road reflections are a FPS killer, as expected. My average FPS increased from 41 to 54 by just turning off these two settings. Reducing shadows to high and shadow blur to optimal increased FPS another 9 FPS to give a highly playable average of 63 FPS, with Super sample AA enabled. Tests at other tracks are similarly encouraging and if FPS still a little low (like Silverstone for example) i reduced shadows again to medium and shadow blur to fast, or turned off soft particles, which smoothed things out.

My test at Loch Drummond was particularly eye opening, as i have always struggled for FPS there. I used the same method of creating a replay to use as a base test. Keeping the same settings in CCC as above, SSAA enabled and with everything maxed in game i achieved an average of 43 FPS with a ten car grid, 55 FPS with reflections off and 67 FPS with shadows set to high/optimal. 67 FPS average at Loch Drummond is a real result and even better when it looked as good as it did!

It was curious to note that despite many previous posters noting a huge FPS drop by enabling SSAA instead of MSAA my FPS drop was small between the two settings. On the Mill Metro Park race for example my average FPS only increased by 7 FPS by using MSAA, and the lowest FPS in that replay increase by only 1 FPS from 41 to 42.

Using these settings injected a previously unseen level of detail to all the ISI based tracks and for the first time i found myself thinking how good RF2 looked, even more so at certain times of day such as late afternoon! Probably not much use for those of you who like to run larger grids as FPS suffer as you add more AI but if you are content with around 10 AI and haven't time to experiment yourself i hope maybe with these or similar settings you can unlock the potential which for me at least this game was visually hiding until now.

 

IMHO the point is that every PC with different software/hardware has its own unique tweaks and settings(which are working), some may be similar on similar systems but on my two amd PCs, different tweaks worked and some worked at both.
I am speaking about grafical settings, different drivers with different settings and tweaks with radeonpro.
It's interesting to hear the tweaks and settings of different users, to know how rF2 performs with them.

For example, I got 55-130FPS with two cards while one has 99% usage and the other one has 30-40% usage but with one card I had 28-63FPS while the gpu was at 99% usage. This is changing a bit with different builds.
That means if rF2 would be optimised to use both gpus, I may get 90-180FPS or something like that and 45-100FPS with one card.

Your post sounds like you found a way to let rF2 use the whole GPU-Power.

Congratulations!

 

Excellent! AMD HD7950 here. I just quickly tried using your CCC settings with my existing rF2 settings at Loch Drummond with 15AI and whilst FPS does drop below my 86fps cap a bit easier, crucially it doesn't drop any lower than it does without the CCC profile (45-50fps at worst), or at least the difference isn't worth noting. The image quality is much improved! My rF2 settings are a mixture of med/high with reflections off and single monitor.

Thanks for sharing!

 

4x SuperSampleAA is a blast!

I have recently discovered it too, like Jayblue, I've fiddled around with different settings and antialiasing and after seeing beauty of nice crisp SSAA, I've never looked back

2x SSAA didn't look special and 16x SSAA were too much. 4x was golden middle for me.

I'm having 2x 6870 in CF, had to turn off some stuff like reflections and mirrors to have constant 60fps. But graphics are smooth as butter like OP said, especially afternoon to night.

 

Well this really improves the image. No jaggies anywhere. Looks quite cool.

However I have quite a big FPS drop but what's more interesting for ISI maybe is the fact that tuning mirrors off and on doesn't have such a big impact on the FPS like with the in game controlled settings. Just a few frames like the nVidia users are reporting.
Maybe something worth to look at ISI.

But the FPS drop for me is too big here, maybe I can find a sweet spot otherwise I'll switch back again.

 

Although i don't use amd, i decided to finally try SSAA because of your thread. I've just tried using 4xMSAA with 2xSGSSAA (Sparse Grid SSAA) and it really does look better than the ingame AA setting (4xSGSSAA looks even nicer...wire meshes are then pretty much 100% clean and sharp looking but the performance hit is too much for me...i want 100+ fps). The improvements are most noticeable when in motion....the general scenery is much crisper and the track edges, walls, even the cars have much less jaggies to them. You drive around corners or long straights and the whole scenery feels hazzy due to the aliasing without SSAA. Very, very nice improvement!

 

Some interesting posts above, really i guess what they prove is that a certain amount of performance can be unlock from AMD graphics cards users by simply taking the time to methodically tweak their settings and find out their systems sweet spot.

I am glad that at least a few others have had similar joy to me when tweaking and i hope that maybe others too will try these suggestions and see for themselves that the game is actually very capable of producing some great visuals, even if it takes some work to get them!

Going to run some tests on my sons PC now and see if i get the same results - his PC is older than (and mine is hardly the latest and greatest) He has C2D e6600 overclocked to 3.0ghz, 3gb ram, ATi HD 6950 2GB, Vista 32bit. I know the number of AI is already limited on his system by his relative lack of system ram, he always tries to run big fields but gets a runtime error or CTD because he runs out of ram!

Jonrait your triple screen tweak should be very useful for those who run that configuration if it is repeatable across all systems, nice find!

 

I'm not willing to turn road and car reflections completey off, but it sounds interesting to try and override FSAA settings.

OTOH, shouldn't the 8 available in-game settings actually be like that? I.e. it should be possible to select 4x SSA or whatever by picking the right number... problem is knowing the right number.

Did this improve distant jaggies like they are so present on Silverstone for example?

 

@Jayblue: Your son's system would benefit more by upgrading to a 64bit version of Windows. The 3.5Gb memory ceiling includes the memory on the graphics card, so the system suffers very quickly as the graphics memory gets used. With the 64bit OS, it will open up about 2Gb more ram available for system and physics work.

@GCCRacer: It used to be that turning reflections off helped with FPS, but you still got some reflection from both surfaces. It might still be the same so check it out and see.

 

That is useful to know, thanks! I am waiting until Intel release either the Haswell refresh or Broadwell later this year for an upgrade, at which point he will inherit my older i7 860 based system.

 

This has been said many times, "reflections off" doesn't mean you don't have reflections, you still have the SAME reflections as the other settings, the effects is ONLY on the update of the reflection map that is done less frequently, let say once a second, so it could happen that you notice it, try historic Brianza for example.

From a performance standpoint use whatever AA level you like EXCEPT "level 3" that is extremely detrimental for AMD cards.

 

On my GTX 770, tested on the Loch Drummond with clio's benchmark replay, my fps went from 120 fps with no environment reflections, 80fps (33% drop) on low environment reflections and 70fps (42% drop) on high environment reflections.

The performance losses (from memory) are no more than half these amounts with no/low/high road reflections. I keep road reflections on high because with just a little bit of heat haze you can see reflections in the road and it's nice (even more so in rain ofc). Environment reflections off because although it looks nice it kills performance and using "low" is not a nice compromise to me because you can see the evironment reflections inside the car update at something like 5hz...very stuttery. With it off, you still get environment reflections but they update every 20 seconds or so and looks much nice than low even though low is real-time but it's stuttering as hell.

edit: you were talking about road reflections, lol.

On another note, using SSAA removed the jaggies compared to ingame aa settings. All ingame aa settings still had aliasing. 4xMSAA + 4xSGSSAA (nvidia) removed it completely. I use 2xMSAA + 2xSGSSAA however because again, performance drop is too much for my taste. 70fps for 4x, 110 for 2x....still looks nice with 2x, much better than ingame aa but will be nice to be able to use 4x whenever i get a newer card (780 users have no problem here).

 

I use this settings already from very early builds(CCC 4xSSAA) and with FXAA on(don't know reason for this behavior), i get a lot less stuttering than with any other settings(despite FPS may be a little lower, but of course image quality is on another level compared in-game settings(more crisp and details are more pronounced).

 

Did all this ages ago but soon realised a new Build changes FPS in different areas and I was essentially wasting time doing Benchmarks when I was quite happy with ingame settings, not to mention some tracks requiring settings to be lowered or raised and It all became a PITA fiddling and testing and not actually racing!


Having said that I could'nt resist the temptation to have a fresh little go this evening just for kicks, and here's the results I got:

HD7850 overclocked = 1200v, +20 Power, 1150 core, 1450 mem

Normal settings ingame with RF2 controlling AAx2 AFx4
Frames: 4442 - Time: 60000ms - Avg: 74.033 - Min: 55 - Max: 90

CCC Override both - MultiSampling 2x AF4x
Frames: 4504 - Time: 60000ms - Avg: 75.067 - Min: 57 - Max: 96

CCC Override both - SuperSampling 2x AF4x
Frames: 3837 - Time: 60000ms - Avg: 63.950 - Min: 48 - Max: 81

A little gain with CCC taking over MS whilst not really noticebale in game, SS obvious smoothness of edges ingame but too costly on FPS for my liking.

 

True Dat

Its nice to fiddle and tweak, especially when it gives good results, but it gets tiresome. Sometimes I think we have more tweak-threads in here then threads about simracing, cars and tracks

I do serious tweaking every 3 builds or so, the rest is just auto-updating. I refuse to become an expert in rf2 tweaking

 

For me the biggest gain in FPS was the (worthwhile) overclocking of my GPU, above a standard HD7870 now and I can push it a little more but found a "safe" zone so it's stayed there.
Still got the CPU to overclock too so may have a play with that this weekend, overclocking is more satisfying when it comes to fiddling & tweaking to get more FPS I find.

 

I'm draining recourses anyways with my system so that's no option for me I think.

Still, bookmarked this thread for future reference, i'm sure I will return and test the stuff that is discussed here.

 

I have an i7 920, with a modest overclock (liquid cooling system) and a Radeon HD5800 with it overclocked through the AMD Overdrive function in CCC. I am also running triple screens through a Triplehead2Go. I think the overclocking helps pretty well, plus these new settings really helped. I am getting between 54 to 93 fps in game. Best I have ever seen.

 

aaaargh back to the drawing board guys ........went home to tinker last night and discovered a glaring mistake in my findings, so much so I've deleted my original post so as not to cause confusion.

You may recall I got serious FPS increase after readjusting the bevel compensation on my triple monitor & resetting the resolution .....well of course in doing this it automatically unchecks the "Multiview" option .... spotted it last night and FPS back down to unplayable levels (for me anyway)...... sorry chaps.

On the other hand, with my FOV and using the howston cars I didn't really notice the screen distortion....... so I've left it unchecked so at least I can run it. Must say, the more time you spend driving it, RF2 is fantastic in terms of physics, feedback and dynamic track.

I also took the plunge and bought Assetto Corsa as well ...... Ok it is prettier and manages to hold higher FPS but RF2 has the edge on the physics IMHO, RF2 is raw and edgy, unforgiving for the heavy footed but so rewarding for the purist where as AC is .....polished and smoothed.

It will be interesting to see how they both develop over time.

JR

 

Just as a note to you guys, I tried changing my system to use the suggested settings and saw little benefit in the realm of fps, but had worse graphics in game than using my normal settings. With CCC controlling the AA and all that, the graphics looked horrible and jaggy. With in game settings the game looked a lot cleaner and smoother. Not sure if I missed anything, but that has been my experience so far.