X1800/7800gt AA comparisons

[_xxx_]

It's definitely the drivers, and it will get better. As for nV's TxAA, I didn't know they had dedicated hardware that does it...? Can someone throw me a link where that is explained, I always thought it was just a driver feature.

I'm not sure either, but that's what they claimed in some interviews

 

[Kombatant]

I'm not sure either, but that's what they claimed in some interviews

Well, if you look at it from a "broader" perspective (aka PR talk), I guess that every single feature is "implemented in HW", since you need the hardware to run it anyway

 

[Jawed]

On my TFT it looks neutral at 2.0 or 2.1 (hard to tell). At 2.2 there is a very slight, very hard to spot coloration. I've tweaked gamma a bit and it makes absolutely no difference wrt to the wire screen shot.

LOL, that's quite funny, the fact that two different gamma values show as "neutral" is not very encouraging.

Jawed

 

[vember]

Be aware that in-game settings can override the gamma LUT.

 

[Nite_Hawk]

Hi Guys,

Thanks everyone for participating in this thread! Thanks also to Sharkfood for more dramatically showing the effect that I was talking about. If any of you guys have a plasma or other high luminance display mind checking out those images?

Jawed:

The LCD display I'm viewing with here at work appears to be calibarted to 2.0. ATI drivers by default assume a gamma-space of 2.2?

Nite_Hawk

 

[_xxx_]

Well, if you look at it from a "broader" perspective (aka PR talk), I guess that every single feature is "implemented in HW", since you need the hardware to run it anyway

I can't find a reference, but I've seen a diagram where it was pictured as a part of the ROP's. But yeah, you're prolly right

 

[L233]

LOL, that's quite funny, the fact that two different gamma values show as "neutral" is not very encouraging.

Jawed

I didn't say they were neutral, I said the difference was hard to tell. It doesn't matter anyway, they're both very close to neutral (maybe the "neutral" with my TFT is 2.05?). Playing around with the gamma settings until 2.20 was neutral did not change the "dotted line" appearance of the wires in the screen shot and that's what we're talking about, isn't it?

It'd say my monitors gamma setting has nothing to do with it. As you can see on the zoomed images, it IS in fact a dotted line of alternating black and gey. It IS there. A high brightness and high contrast monitor just makes it very visible without having to zoom in.

 

[Jawed]

The LCD display I'm viewing with here at work appears to be calibarted to 2.0. ATI drivers by default assume a gamma-space of 2.2?

Windows is defined to operate at gamma 2.2.

ATI assumes that you're running your system at gamma 2.2. Unfortunately this involves the user making sure of this...

Obviously if you run your system mis-calibrated then this will clash with the "gamma-correct AA" that ATI opines about. I don't know how much, because I've not personally experimented - I just keep my system properly calibrated.

Incidentally, the wayward gamma settings being found here is NVidia's argument why "gamma-correct AA" is a waste of time. I dare say, you're sort of proving them correct

Jawed

 

[Nite_Hawk]

Windows is defined to operate at gamma 2.2.

ATI assumes that you're running your system at gamma 2.2. Unfortunately this involves the user making sure of this...

Obviously if you run your system mis-calibrated then this will clash with the "gamma-correct AA" that ATI opines about. I don't know how much, because I've not personally experimented - I just keep my system properly calibrated.

Incidentally, the wayward gamma settings being found here is NVidia's argument why "gamma-correct AA" is a waste of time. I dare say, you're sort of proving them correct

Jawed

That doesn't really make any sense though. A higher gamma space should show a more contrasty image. If anything, I'd think a gamma of 2.2 would make this more apparent rather than less so. Even at that, it's not like this is a terribly dramatic difference. The luminance theory seems more plausible to me at this point.

Nite_Hawk

 

[Jawed]

I didn't say they were neutral, I said the difference was hard to tell.

That's what's worrying me. The difference should be bleeding obvious.

Too low and the right-hand greyscales should be coloured red,green,blue and too high they should be coloured cyan,magenta,yellow. Any doubt about this is what's worrying me...

It doesn't matter anyway, they're both very close to neutral (maybe the "neutral" with my TFT is 2.05?). Playing around with the gamma settings until 2.20 was neutral did not change the "dotted line" appearance of the wires in the screen shot and that's what we're talking about, isn't it?

Yes.

But it's nice to eliminate the fact that your LCD is incorrectly calibrated before assuming that the "gamma-correct AA" is not functioning.

It'd say my monitors gamma setting has nothing to do with it. As you can see on the zoomed images, it IS in fact a dotted line of alternating black and gey. It IS there. A high brightness and high contrast monitor just makes it very visible without having to zoom in.

Now that you've calibrated your LCD "properly" I'm quite happy to accept that your LCD is "revealing too much detail in the AA pattern", hence making it look worse.

So far we've eliminated sharpening (because apparently your LCD doesn't sharpen) and gamma (it's now correct).

So we're left with the dynamic range and tonal range your panel is capable of.

Is it a true 8-bit panel or is it a 6-bit dithering panel?

Jawed

 

[Jawed]

That doesn't really make any sense though. A higher gamma space should show a more contrasty image. If anything, I'd think a gamma of 2.2 would make this more apparent rather than less so. Even at that, it's not like this is a terribly dramatic difference. The luminance theory seems more plausible to me at this point.

Nite_Hawk

NVidia's AA assumes gamma 1. That's what makes the lines look thicker than they should, because the "gradient" from light to dark at the edge of the lines has more pixels "dark" than it should.

As you go up the gamma-scale, the proportion of "dark" pixels reduces - making the lines look thinner. If the gamma goes too high, then the gradient consists mostly of light pixels and the line will break up.

If you have Photoshop then you can experiment with image manipulations (e.g. resizing or rotating) using gamma 1 or gamma 2.2 colour spaces. I wrote about it recently, here:

http://www.beyond3d.com/forum/showpost.php?p=541182&postcount=30

http://www.beyond3d.com/forum/showpost.php?p=541268&postcount=32

The point I've just raised about tonality (6-bit?) is the next interesting question.

Jawed

 

[Nite_Hawk]

Jawed:

It appears the display I'm on is a rebadged samsung display of some model. A couple of places claim 24bit, but dell doesn't appear to say. The only really useful specs on the panel that dell gives are:

25ms
600:1 contrast ratio
250 CD/m ² luminance output

My guess though, is that this one of samsung's 25ms MVA panels. It certainly has the look of one. Those should be full 24bit panels.

Nite_Hawk

 

[L233]

So we're left with the dynamic range and tonal range your panel is capable of.

Is it a true 8-bit panel or is it a 6-bit dithering panel?

Jawed

It's an 8 bit LUT panel and I have carefully adjusted the contrast and color settings to not throw out any colors, except for the first grey tone after black.

The way contrast works on my TFT (and prolly most TFTs) is that if you lower or raise contrast it simply shifts the colors up or down (e.g. 230 -> 229). That means if you lower the contrast one tick and the lowest grey value becomes black. This is why 10 bit LUTs are nice, they don't throw out any colors when you change contrast.

It also means that contrast adjustment is merely a corrective to be used when you change color temerpatures on my TFT.

Brightness on the other hand does nothing but increase or decrease the intensity of background lighting on my TFT which is great because it's one variable less that can fuck up your colors.

Be it as it may, I spent a lot of time in front of grey scale images adjusting my settings to the point where I have the full colour range (except the lowest grey value) and smooth colour transition without any visible banding. I am confident that this is as good as it gets with an 8 bit LUT.

 

[Jawed]

OK, well, I'm not an expert with LCDs so I can't provide any further insight into what's going on.

It'd be interesting if this question gets more prominence and perhaps an article on the subject gets written.

Maybe ATI themselves would have something to say on this topic.

So, in the meantime, we're apparently left with either the maximum brightness or the dynamic range of LCDs as being the reason why "gamma-correct AA" is rendering poorly...

Jawed

 

[Moloch]

while lcds are brighter, they dont have the same ability of crts to produce deep blacks.
LCD specs are all full of shit- pixel response and contrast ratio.
I'd say it's because of the fact that lcds are alot brighter than an avg crt and as some people here said, are calibrated at something other than what ati assumes the user has.
contrast is the DIFFERENCE between the darkest black and the most intense light, not brightness.
Repeat after me
http://www.anandtech.com/displays/showdoc.aspx?i=2428&p=5
Look at the contrast ratio found by "boy wonder"
The highest is 101, by the Samsung 915N, however the darkest black was recorded by the Samsung 193P.

 

[Nite_Hawk]

OK, well, I'm not an expert with LCDs so I can't provide any further insight into what's going on.

It'd be interesting if this question gets more prominence and perhaps an article on the subject gets written.

Maybe ATI themselves would have something to say on this topic.

So, in the meantime, we're apparently left with either the maximum brightness or the dynamic range of LCDs as being the reason why "gamma-correct AA" is rendering poorly...

Jawed

Well, if it is happening on the CRT that is set to a high luminance as well, it doesn't seem specifically like an LCD problem. I'd really like to know if this is a problem on a plasma or TV CRT displays as well. Maybe I can try checking it out later.

It's certainly an interesting problem.

by the way, can ATI's gamma correction be turned off? If so, it would be interesting to see those same shots with ATI's various AA modes without gamma correction and see how they compare.

Nite_Hawk

 

[DemoCoder]

This used to be true. AFAIK good CRT monitors have contrast ratios of 500:1, most have less. My 19" LCD is listed with a 1000:1 contrast ratio. Most recent VA-panel 19" LCDs have contrast ratios of at least 800:1. All modern LCDs have contrast ratios of at least 450:1. Even if the manufacturers' claims are full of shit (they usually are) I think it is safe to say that CRTs don't have an advantage over modern LCDs anymore when it comes to contrast.

And you would be wrong and I say that as an LCD lover. LCD CR specs are some of the most outrageous bullshit. On/Off CR has nothing to do with IQ contrast response. It's the gamma ramp that determines it. Boot up an calibration disk like Avia on a CRT and on a LCD. Calibrate both. Now look at the gamma ramp distribution. The LCD will not look as good as the low and hire IRE values. Maybe I misspoke in the original post "superior contrast" is what I meant. Measuring contrast by measuring On/Off really tells you almost nothing, except what's the brightest white and darkest black, and even then, ANSI CR is a better measure.

 

[Jawed]

Well, if it is happening on the CRT that is set to a high luminance as well, it doesn't seem specifically like an LCD problem.

But when the CRT is set to high luminance it needs to be re-calibrated for gamma 2.2. Otherwise the comparison is invalid.

It may well be that extremely high luminance is the root of this, but we have to be sure that everything else is controlled first.

I'd really like to know if this is a problem on a plasma or TV CRT displays as well. Maybe I can try checking it out later.

Well make sure that you ensure gamma 2.2 is being adhered to, as per those tests I posted earlier.

by the way, can ATI's gamma correction be turned off? If so, it would be interesting to see those same shots with ATI's various AA modes without gamma correction and see how they compare.

I don't think ATI allows it to be turned off.

Jawed

 

[Nite_Hawk]

while lcds are brighter, they dont have the same ability of crts to produce deep blacks.
LCD specs are all full of shit- pixel response and contrast ratio.
I'd say it's because of the fact that lcds are alot brighter than an avg crt and as some people here said, are calibrated at something other than what ati assumes the user has.
contrast is the DIFFERENCE between the darkest black and the most intense light, not brightness.
Repeat after me
http://www.anandtech.com/displays/showdoc.aspx?i=2428&p=5
Look at the contrast ratio found by "boy wonder"
The highest is 101, by the Samsung 915N, however the darkest black was recorded by the Samsung 193P.

Pixel response specifications have nothing to do with this thread.

As far as contrast ratios are concerned, we really don't know much about how their tests compare until they include a CRT do we? They even admitted in the article that their equipment isn't as sensative as those in the manufacturers labs. I don't disagree that the manufacturers specs are over-inflated, but it's meaningless to compare "maybe accurate" real world LCD test results with CRT marketing specs.

Despite this, let me pose a question for you. Do you think that a lack of contrast would make this effect more or less apparent? Do you honestly think that a lack of contrast is going to make the lighter and darker portions of the antialiased line become more visible?

Nite_Hawk

 

[Moloch]

Pixel response specifications have nothing to do with this thread.

As far as contrast ratios are concerned, we really don't know much about how their tests compare until they include a CRT do we? They even admitted in the article that their equipment isn't as sensative as those in the manufacturers labs. I don't disagree that the manufacturers specs are over-inflated, but it's meaningless to compare "maybe accurate" real world LCD test results with CRT marketing specs.

Despite this, let me pose a question for you. Do you think that a lack of contrast would make this effect more or less apparent? Do you honestly think that a lack of contrast is going to make the lighter and darker portions of the antialiased line become more visible?

Nite_Hawk

Someone brought up the fact that lcds have crazy high contrast ratio listed.. they also have overrated pixel responce times listed, get it?
They go together like two peas in a pod.
think the problem is simply ati assumes you have 2.2 gamma and your lcd doesn't use that, like it should.