[H]OCP does Radeon9700 I.Q....
- Thread starter martrox
- Start date
I thought that in game gamma settings just was an override for the driver gamma settings, which in turn applies a gamma correction in the RAMDAC. In other words, it doesn't affect at what gamma the image is stored in the frame buffer. So it's only to be used to adjust the gamma compensation for your monitor (at your current res/intensity/brightness).
In that case I don't see why there should be any gamma correction before the FSAA blending, since the framebuffer should be at linear gamma (otherwise you'll get errors in pretty much all lightning math and alpha blending with current cards).
However, if you have the gamma set to make the framebuffer value proportional to intensity as perceived by humans, then you would need the gamma correction in the FSAA blending and all other frambuffer read/writes. (And then on top of that do a correction for monitor gamma in the RAMDAC.) As far as I know, the only card available now that can do that is R9700, so I don't think games are made for that.
Now the questions I want to ask is:
How does in-game gamma correlate to driver-gamma?
Does R9700 force a gamma != 1 for the framebuffer, and in that case what gamma? (2.2?)
Using the in game gamma for it just don't make sense to me, since I believe that would mean that it has a different meaning than for other cards.
In that case I don't see why there should be any gamma correction before the FSAA blending, since the framebuffer should be at linear gamma (otherwise you'll get errors in pretty much all lightning math and alpha blending with current cards).
However, if you have the gamma set to make the framebuffer value proportional to intensity as perceived by humans, then you would need the gamma correction in the FSAA blending and all other frambuffer read/writes. (And then on top of that do a correction for monitor gamma in the RAMDAC.) As far as I know, the only card available now that can do that is R9700, so I don't think games are made for that.
Now the questions I want to ask is:
How does in-game gamma correlate to driver-gamma?
Does R9700 force a gamma != 1 for the framebuffer, and in that case what gamma? (2.2?)
Using the in game gamma for it just don't make sense to me, since I believe that would mean that it has a different meaning than for other cards.
Full screen gamma correction could fix the edges and other places.
However most content is developed with average PC monitors in mind (around 2.2 gamma).
So applying full screen gamma correction would make that content look washed.
On the other hand gamma correction on AA does nothing to the image except at the edges.
For example the monitor might have a 2.2 gamma, that means it maps input signals to output brigness approximately with the following function:
g(x) = x^2.2
where x is the input signal (0..1)
and g(x) is the output signal (0..1)
Standard AA does simple averaging of the pixels.
For example 2x AA does:
(sample1 + sample2) / 2
Unfortunately this is done in the input signal's domain, so it won't look correct on the monitor.
Gamma corrected AA does:
g'( (g(sample1) + g(sample2)) / 2 )
where g'(x) is the inverse gamma function.
Notice that the function does nothing when sample1 == sample2 so it doesn't do anything inside polygons - only at the edges.
If you don't like math here's a picture to demonstrate:
The upper line is normal, the bottom is gamma corrected. It meant to look good around 2.2 gamma. (Lines are where the effect is most evident.)
Without correct gamma correction a line would look like a dashed line (look at it from a further distance if your monitor is too sharp.)
However most content is developed with average PC monitors in mind (around 2.2 gamma).
So applying full screen gamma correction would make that content look washed.
On the other hand gamma correction on AA does nothing to the image except at the edges.
For example the monitor might have a 2.2 gamma, that means it maps input signals to output brigness approximately with the following function:
g(x) = x^2.2
where x is the input signal (0..1)
and g(x) is the output signal (0..1)
Standard AA does simple averaging of the pixels.
For example 2x AA does:
(sample1 + sample2) / 2
Unfortunately this is done in the input signal's domain, so it won't look correct on the monitor.
Gamma corrected AA does:
g'( (g(sample1) + g(sample2)) / 2 )
where g'(x) is the inverse gamma function.
Notice that the function does nothing when sample1 == sample2 so it doesn't do anything inside polygons - only at the edges.
If you don't like math here's a picture to demonstrate:
The upper line is normal, the bottom is gamma corrected. It meant to look good around 2.2 gamma. (Lines are where the effect is most evident.)
Without correct gamma correction a line would look like a dashed line (look at it from a further distance if your monitor is too sharp.)
Doomtrooper
Veteran
Mize said:
No I have not, I installed one for a good friend of mine the other day..MSI KT333 Xp 2000 combo..went very smooth...good quality stuff Enermax 430 watt Powersupply.
I will be getting mine somtime in the next month once I find a buyer for my Radeon 8500...
I have had alot of bad luck as of late...cars breaking down, health problems etc...so money has to be prioritized....and with a wife...well
Bigus Dickus
Regular
Thanks for the clarification Hyp-X.
BTW, the upper line is precisely what AA'ed wireframe models look like in CAD...
BTW, the upper line is precisely what AA'ed wireframe models look like in CAD...
Doomtrooper said:.I must say I'm glad to see ATI jumping right on these issues..well done guys.
Ditto...
http://www.ati.com/support/infobase/4080.html
Ahh there's the catch. I assumed that games set the gamma compensation to 2.2 to make the frambuffer linear. If you set gamma correction in the driver panel to 2.2, then the upper line looks good, and the lower line looks (slightly) dashed. If the games is made for gamma=2.2, then you won't need gamma compensation in the FSAA. Furtermore, if you don't program for gamma=2.2, then all lightning will be wrong (again, unless you compensate on every FB read/write).
So does R9700 force gamma correction for all FB read/writes?
So does R9700 force gamma correction for all FB read/writes?
Doomtrooper said:
Thanks, I try
RussSchultz said:
JPEG Compression, no touching up was done, I simpy took the shots, saved em in Paint, then put our logo and text on them in Photoshop and saved them in the same full quality .bmp format, and then used Ulead's Smart Saver Pro to compress them.
Bigus Dickus said:
Yep, JPEG compression, Unfortuntely it's not practicle to have them all non-compressed
Believe me I know it's not perfect, but can you imagine dl'ing 5MB files for wach picture? Especially if your on dial up, and not to mention our bandwidth usage would go through the roof! What I can do though is say if you want non-compressed images, email me, I have all of them still in their non-compressed state.Reverend said:Whoa... using a warez copy of UT2003 in an article = big ballz!
GF4 4x is better than 9700 2x based on the shots posted in this thread (just going with the flow here... dunno why anyone started this comparison in the first place).
Image compression artifacts as noticeable as the ones here (in a IQ article) = bad !
But a good read nonetheless, Brent! Looks like you've been doing a better job at most things than Kyle since you joined him!
"I've got big balls, I've got big balls, And they're such big balls, Dirty big balls, And he's got big balls, And she's got big balls, But we've got the biggest balls of them all! "
LittlePenny said:
Let me get em all together and see how much space it will require for all the non-compressed images.
McElvis said:
I took the screenshtos at different times of the day, and in Flight Sim 2002 it uses the time from your computer to decide where in the sky the sun (light source) is. Because I took them at different times of the day the sun was in a different place in the sky, thus casting the shadow differently. This was simply a mistake on my part, i should have manually selected the time of day in the game so that the sun would be in the same spot in all the shots, I simply forgot about that when I was taking the shots.
As for AA in general on the R9700 Pro I guess you just have to see it in person compared to other cards, play games, notice edges etc...
The 9700 really does look better at AA then the GF4, it's VERY noticable in games, and games are were it counts!
I have been playing FS2002 a lot here recently, I find that with a GF4 Ti 4600 i have to run at 2X AA and 4X AF at 1280x1024x768 to be smooth enough for gameplay, and even still you can see aliasing on the planes... however with a radeon 9700 pro on the same system i can run 6X AA and 16X AF at 1280x1024x32 and it's smooth... when i drop to 2X AA just to compare IQ you can just see that the R9700 Pro is so much better at AA'ing the edges, most noticeable from a spot view of the planes.... you just have to see it for yourself!
As for the JPEG compression of the screenshots, again I am sorry about that, but it could not be helped
I am though offering the non compressed images through email, though they are very large.
Hyp-X said:
Thanks for the explaination, Hyp-X. Now I have something real to look at.
Here's what I did:
I loaded up Morrowind on my GeForce4 Ti 4200, and selected wireframe mode. Indeed, I could see the artifact where some of the lines, depending on angle, looked "dotted."
Then I decided to play around with the gamma settings. From the very brightest to the dimmest gamma, the amount of "dotted lines" that I saw did not change in the least (This was using in-game gamma, of course). Driver gamma settings were at default.
OpenGL guy
Veteran
And this proves what? That this technique is not effective for the GeForce 4?Chalnoth said:
Doomtrooper
Veteran
Let me answer that... YES.
Chalnoth seriously is it your goal in life to badmouth, spread rumors, state obvious flawed information about ATI's hardware.
I have no idea why you are so concerned about ATI cards at all, you will never own one..so give live in your fantasy world.
Chalnoth seriously is it your goal in life to badmouth, spread rumors, state obvious flawed information about ATI's hardware.
I have no idea why you are so concerned about ATI cards at all, you will never own one..so give live in your fantasy world.
It tells me that the problem that was described by Hyp-X is not tied to a specific gamma setting (otherwise there should be a gamma I could select that would make it disappear with a non-corrected implementation...which is not the case).
Assuming, of course, that the GeForce4 does not use a similar technique as that seen in the Radeon 9700. Since we've never heard of it, I doubt that it does use a similar gamma technique.
Assuming, of course, that the GeForce4 does not use a similar technique as that seen in the Radeon 9700. Since we've never heard of it, I doubt that it does use a similar gamma technique.
OpenGL guy
Veteran
There's a little more to the algorithm than can be simulated this way. I guess you'll just have to take my word for it (of course, you won't) and take the words of people who have said that the Radeon 9700 gives better AA quality than the GeForce 4 (of course, you won't).Chalnoth said:
