NVIDIA 344.48 WHQL GPU Drivers

By the way, if anyone wants a simple, replicable jaggies test spot, just do the following:

Set AI to 20.
Go to Practice Mode 9:00 to 10:00 am start time ISI's Autodromo Portugal (Estoril) and sit in the garage for a moment.

Rev the engine. Do the railings right across the pit lane shimmer with the slight movement of the car?

Drive out of garage onto pit lane and observe the railings that are now on your left and the garage stall fence separations. Both of these can be a festival of jaggies and movement, or, solid as a rock (and realistic looking), if you are willing to suffer the performance loss.

It will be extremely difficult to get this scene looking good. The AI are there so you can also see how bad the FPS will get when they are in view and compare one vehicle to another. Shadow optimization (or not), reflections, etc., will all affect the frame rate dramatically from one car to the next.

It is, however, a very good and easy place to notice the sometimes glaring differences in all the settings.

Follow the instructions precisely. Even using an afternoon lighting changes the results dramatically. This is the "worst-case" scenario for testing as far as I am concerned.

The test of which settings you prefer can only be done with a test like the one above that includes some driving and movement of the car (albeit very slowly). Static screenshots CANNOT demonstrate the merits of settings with rF2. Sorry.
 
"Static screenshots CANNOT demonstrate the merits of settings with rF2."

I agree one hundred percent. I tend to judge settings IQ by what I see when the car is in motion. I look at white lines, clarity of cars in the distance, lettering on cars, chain link fences and numbers on instruments in the cockpit.

Marc, what settings in Inspector are you testing?
 
Yeh, some colors are overly saturated on some cars with HDR, but I think this is mainly a fault of the mods and not rF2. Lighting, reflections, terrain, etc. all look better and more natural with HDR.
 
Carlo said:
"Static screenshots CANNOT demonstrate the merits of settings with rF2."

I agree one hundred percent. I tend to judge settings IQ by what I see when the car is in motion. I look at white lines, clarity of cars in the distance, lettering on cars, chain link fences and numbers on instruments in the cockpit.

Marc, what settings in Inspector are you testing?
Click to expand...

8xQ with 2xSGSS looks best on my system. I was testing various options to see if any significant performance gains could be made without losing too much on the visual side. I think the answer is no, but 16xCSAA with 8xSS may come closest.
 
Pretty much what I've discovered. Either that, or 4xSGSS and 4xMS. With 2xSGSS and 8xQ, the image appears sharper (which I like), but does introduce minimal aliasing that is hardly noticeable when driving.

With both, I do set json ts to 0 and set the correct LOD for the SGSS mode in Inspector. Is this what you do?

Also, I don't seem to notice much FPS difference between 4xSGSS/4xMS and 2xSGSS/8xQ.
 
Carlo said:
Pretty much what I've discovered. Either that, or 4xSGSS and 4xMS. With 2xSGSS and 8xQ, the image appears sharper (which I like), but does introduce minimal aliasing that is hardly noticeable when driving.

With both, I do set json ts to 0 and set the correct LOD for the SGSS mode in Inspector. Is this what you do?

Also, I don't seem to notice much FPS difference between 4xSGSS/4xMS and 2xSGSS/8xQ.
Click to expand...

Yes, JSON TS to 0, or use the special .dll file (see other thread). I have not set the "correct" LOD in Inspector. What should that be?

Also, what visual difference do you notice between those two?
 
In Inspector:

Texture filtering - Driver Controlled LOD Bias = off
Texture filtering - Negative LOD Bias = allow

Then for 2xSGSS ...

Texture filtering - LOD Bias (DX) = -0.500

For 4xSGSS ...

Texture filtering - LOD Bias (DX) = -1.000

This should eliminate any blurring, but you may find the image too sharp for your liking.
 
Carlo said:
In Inspector:

Texture filtering - Driver Controlled LOD Bias = off
Texture filtering - Negative LOD Bias = allow

Then for 2xSGSS ...

Texture filtering - LOD Bias (DX) = -0.500

For 4xSGSS ...

Texture filtering - LOD Bias (DX) = -1.000

This should eliminate any blurring, but you may find the image too sharp for your liking.
Click to expand...

Thanks! I don't see any troublesome blurring now, but it will be interesting to compare.

What should these be set to for non-SGSS? Or does every individual setting (like 8xSS) have its own corresponding entries?

And if driver control of LOD Bias is "off" doesn't that mean the app is controlling it (with JSON file TS setting in our case)?
 
2x MSAA + 2x SGSSAA: -0.500
4x MSAA + 4x SGSSAA: -1.000
8x MSAA + 8x SGSSAA: -1.500

"The formula for determining the correct LOD bias is "y = -0.5 * log, base 2, of (n)" where n is the number of samples and y is the correct LOD bias. This function simply means that every time you double the number of samples you subtract another 0.5 from the LOD bias."

Time to get out the calculator ....

As far as I know, the nVidia driver will take of LOD properly when not using SGSS modes (although you might want to play with it). In this case, however, clamp would likely be advantageous with better IQ.
 
Carlo said:
2x MSAA + 2x SGSSAA: -0.500
4x MSAA + 4x SGSSAA: -1.000
8x MSAA + 8x SGSSAA: -1.500

"The formula for determining the correct LOD bias is "y = -0.5 * log, base 2, of (n)" where n is the number of samples and y is the correct LOD bias. This function simply means that every time you double the number of samples you subtract another 0.5 from the LOD bias."

Time to get out the calculator ....

As far as I know, the nVidia driver will take of LOD properly when not using SGSS modes (although you might want to play with it). In this case, however, clamp would likely be advantageous with better IQ.
Click to expand...

Except clamp doesn't work, so you have to use the special DLL file for non-SGSS...???
 
Yes, that is my understanding. Texture Sharpening in JSON (setting other than zero) may, or may not, have an effect similar to clamp. It's never been clarified, to me at least, whether Texture Sharpening affects the whole image or not.
 
Marc Collins said:
Thanks! I don't see any troublesome blurring now, but it will be interesting to compare.

What should these be set to for non-SGSS? Or does every individual setting (like 8xSS) have its own corresponding entries?

And if driver control of LOD Bias is "off" doesn't that mean the app is controlling it (with JSON file TS setting in our case)?
Click to expand...

I dunno

This is Standard 4xS , no SGSS clamp and stock rF2 sharpening.

please explain what is wrong with this shot ? power lines smooth ? fast fps, blurry graphics is minimal.


BIG thing is you get a perfect moving images no little shimmers or sparkles whatsoever.

Looking at static shots won't achieve anything

Sure it may not look the best but fps is solid and reliable and smooth latency is more important.


People go ramping up settings too much it just runs like a dog and their laptimes will suffer.


Try turning settings down a tad you will most likely have better control.
 
I agree with what you've said DD. It's just that people like Marc and I are probably IQ nuts.

What we're trying to do is get the best IQ obtainable with no latency given the hardware we have. I'm one of the people that ranks GFX almost as high as physics. You are the opposite.

I know, that's sacrilegious! :)
 
Carlo said:
I agree with what you've said DD. It's just that people like Marc and I are probably IQ nuts.

What we're trying to do is get the best IQ obtainable with no latency given the hardware we have. I'm one of the people that ranks GFX almost as high as physics. You are the opposite.

I know, that's sacrilegious! :)
Click to expand...

I am saying you will get the best latency my settings.

Period.

I have tried all methods if I can't do my personal bestlap, forget it.



I don't believe it anyways 3/4 of all these Inspector settings tests is just chasing quality not laptimes.

Want good latency dumb down the settings, all she wrote.
 
DurgeDriven said:
I dunno

This is Standard 4xS , no SGSS clamp and stock rF2 sharpening.

please explain what is wrong with this shot ? power lines smooth ? fast fps, blurry graphics is minimal.


BIG thing is you get a perfect moving images no little shimmers or sparkles whatsoever.

Looking at static shots won't achieve anything

Sure it may not look the best but fps is solid and reliable and smooth latency is more important.


People go ramping up settings too much it just runs like a dog and their laptimes will suffer.


Try turning settings down a tad you will most likely have better control.
Click to expand...

Are you talking about 4xS, or 4x Multisampling? 4xS is actually a seriously heavy-duty setting with far lower performance than most anyone would want. 4x Multisampling alone is a low quality setting that is too ugly for the majority of people.

And the whole point of this exploration as Carlo just said is to get the best IQ WITHOUT compromising control. If we didn't care about control, we'd just crank it up to the highest settings, which always look better.
 
You must log in or register to reply here.
Back
Top