Anti-Aliasing Levels Correspond to What?

you mentioned Monaco...that's old content, that's what I was quite clearly referring too.

Technical answer is to just play the game as there is no fix (quite clearly) instead of nit-picking, sure the trees flicker slightly, annoying it may be (to some), unplayable? certainly not. Are the ISI Track Team aware of it? I would like to think so, therefore it may very well get fixed in the future so everyone's happy...agree?

Well are you enjoying the game? because form your constant need to point out flaws it would appear not, especially when those flaws are so minor. Again, every game has it's little issues so if they affect you THAT much then I suggest you relax with a beer and take your mind off it!

 

Since there seem to be only two categories of state for you, unplayable/needs fixing and ignore the problem, I will refrain from replying further or reading anything you write (anywhere, not just here). Any normal conversation is impossible when the frame of reference is incompatible.

 

lol ok then yeah that's right I said ignore didn't I?! some people are just never happy!

 

I think this, in the same way as separately rendered white lines, will be unavoidable without supersampling. A thin object will just be rendered in 'bits' when close to horizontal or vertical, and no amount of normal AA will be able to fix it (white lines and static 'shadows' that are drawn in the texture don't have the problem, because as the texture itself is sampled to be drawn in place the line/shadow will be partly brought in smoothly, just like every other part of the texture). The only remedy is to render at higher resolution and resample to lower - just as supersampling does. Supersampling is actually the most obvious method of AA; every other method has been derived for performance reasons.

 

I think you may be right. Maybe I should say I have feared for a long time that this is ultimately the reality we are dealing with since no combination of any settings would ever produce a solid image without supersampling (including SGSS on NVIDIA). I guess I will need a next (or next-next) generation NVIDIA card to be able to maintain a minimum 60 FPS at my monitor's native resolution

 

AMD ppl have it made, just choose 4xAA and supersampling type, done. Even better image quality than 4xMSAA + 4xSGSSAA.

 

There is something called Temporal Anti-Aliasing or TXAA, I don't know if it specifically solves this but it seems that some games support it and isn't a performance killer as supersampling...

 

From a quick read it seems like it could potentially improve things in motion, though specific scenarios might reduce its effectiveness and presumably it wouldn't help when stationary. nVidia actually says it's quality at the expense of performance, so I guess you'd put it somewhere between the other AA schemes and supersampling.

The second vid here demonstrates it pretty well (the first vid shows TXAA on/off, but chooses a bad scene for it). Write up also confirms it won't do anything when stationary which makes sense, which might seem ok for a driving game but sometimes objects such as fences may appear quite stationary despite your movement parallel to them. Of course some people don't notice or care about what Marc has shown above, so it's all personal preference

http://www.geforce.com/hardware/technology/txaa/technology

 

That's what I usually use. Great image quality. But with an R9 290X it cannot maintain 60 FPS at 1920x1200. FPS can drop into the high 30's or 40's with a bad combination...like Honda NSX in pit lane at any decent track

4xMSAA + 4XSGSSAA is what I used when I had NVIDIA. It's good enough image quality for me with significantly less performance hit than full SSAA. Either of them are head and shoulders better than ANY other settings, including all the fancy new stuff like TXAA.

 

I agree.

Have you tried giving 2xSSAA a shot?

If you're running maxed in-game graphics, have you tried 4xSSAA w/ shadows and circuit detail both @ high, AF @ 8x, shadow blur @ fast, and maybe the car-body reflections @ low or off?

 

2xSSAA looks bad.

Interesting suggestion on the settings. Here's the result, using one of my "worst-case" scenarios for FPS:

4xSSAA with all maxed: 35-39 FPS
Level 8 (and MS) with all maxed: 48-67 FPS
4xSSAA with your modifications: 41-50 FPS
Level 8 (and MS) with your modifications: 58-97 FPS

What are your conclusions?

 

Hmmm, I'm suprised because in ISI based sims (RF1, RF2, GTL, R07, etc.) 2xSSAA looked better than any non-SSAA setting (but noticeably not as good as 4x). Of course drivers aswell as RF2 itself could have changed since then so my experiences could be irrelevant here.

I wouldn't want to play at nothing less than my display's refresh rate.

If using SSAA, maybe try disabling transparency AA in your RF2 player.json file since the SSAA will take care of that. The framerate increase might not be much though.

I would definitely put circuit detail and shadows on high instead of full, AF at 8x instead of 16x, and shadow blur on fast instead of quality or optimized. There is hardly a decrease in IQ. Then from there use the highest AA you can get away with.

If you still need more frames then perhaps try lowering shadows to medium, maybe some of the reflection settings and opponent detail to high instead of full.

Try to keep texture detail at full though because even going down one click to high results in a noticeable decrease in IQ including that of your cockpit.

What GPU do you have? Make sure RF2's AA and FXAA are disabled aswell as any kind of sync modes (e.g. video sync).

 

Agree with all you said (except the 2xSS...it just is a bit rough looking). R9 290X. All settings as you suggest already.

It's a trade-off between some flickering with all maxed (using Level 8 MS) and no flickering, but losing some details and features. Can't have both.

The FPS above are lowest possible. Even with that combination, out on the track, it is usually above 60 FPS. But any time there is dust or a race start or getting near certain pits, etc., you risk diving below 60 and then also have the problem of the screen transition as it crosses 60 (causes a visible glitch on each pass, up or down).

 

If you get drops during pits then lower the pit LOD setting, can make a big difference in RF1 so I would think it makes a big difference in RF2 aswell. It's in the player.json file.

 

Doesn't help if the car has improperly implemented LODs (like the NSX)

 

Interestingly, the GTX 970 I mounted today only offers 4 FSAA settings ( Zero to Three).

But it looks absolutly spectacular, and gives 60 steady FPS unless at the end of a 20+ car race start, where it might "only" be 55.

 

Let's go back on to the core topic. Now that we have determined that rF2 uses the AA levels (number of and features of) from your GPU (and its driver), why would there be any advantage or reason to prefer in-sim AA settings over driver-set (GPU control panel) ones? I have heard arguments about this over the years, but those would now appear to be rubbish.

 

There are different settings in the drivers.
There is a setting that forces antialiasing for the entire scene, then there's another setting that applies your desired antialiasing to the objects that the application specifies.
The first one is slower and the second one is useful when you don't know what antialiasing the application uses.
I use (and recommend) the second one so I can exactly specify what antialiasing method to use and let rFactor specify what parts of the scene to apply the antialias to; note that you need to enable the antialiasing in rFactor for this to work, though the level that you specify in rFactor doesn't matter in this case because it will be overridden by the selected one in the driver.
Hope this helps to clarify.
Happy racing!

 

And then there's the option of rendering the screen at a higher resolution with less/none AA and downsampling it. Forget what the option is called by the various brands, but some people report a nicer image & better framerates.

 

You're describing supersampling generally. Doing this for the entire screen (FSAA) gets good results but is very expensive. The other methods either perform supersampling only on selected parts of the image (MSAA for example avoids supersampling where the entire final pixel is contained within a single (geometry) triangle, because there are no edges that will need smoothing), or use various methods of sampling to arrive at a final pixel colour - both intended to improve performance over FSAA but hopefully achieve similar quality.

Here's an old guide that might help:

http://www.tomshardware.com/reviews/anti-aliasing-nvidia-geforce-amd-radeon,2868.html

If you could do something like 32xFSAA the quality would be great. But it would be a slideshow, not a video.