BTW on gamma corrected AA sometimes it does actually look bad IMO, for example a black power line or something similar, thin and high contrast with background, tends to in my opninion actually look worse with gamma corrected AA. So maybe their could be a toggle, or more to the point something in the drivers that could tell if this scenario was developing, and if so change the AA a bit. There was a thread awhile back about this.
Beyond3D Wolves Needed to Tear This One Apart
- Thread starter JoshMST
- Start date
- Status
Lollerskates!!1The Baron said:
They are when it comes to GPUs. Video cards aren't like politics--there's not much intuitive about edge- and texture-aliasing to the layman. People buy video cards to play games, not to play settings. There's a reason consoles are so popular: buy console, buy game, play. No AA, AF, TAA, GI, bri, etc. There's also a reason devs spend time finding the best default settings for various video cards: so their customers won't get lost/intimidated/angry when their game runs like poop on their card.radar1200gs said:
Choice is nice, but at some point you're going overboard.
I was under the impression R300's AA was jittered, not rotated, and for all modes, not just 4x and up.radar1200gs said:
It's sparse sampled - in all modes (even 2x, Greg). As far as I can figure R300 has enough subsample accuracy for up to 144 subsample positions. NV3x had enough subsample accuracy for 4 subsample positions, so even it is sparse sampled at 2x.
radar1200gs
Regular
I'd have to look it up again to certain, but I think you'll find ATi's AA is based on a rotated grid, with the ability to change (jiiter) the sampling points, which is the basis for temporal AA being feasible.
I think you will find that many ordinary people are better acquainted with their video control panel than you give them credit for. They may not always fully understand why a setting has the effect it does, but that doesn't stop them experimenting and knowing the results the like.
Like I said early on, provide sensible defaults by all means, but allow those who want to change things to change them.
The reason OpenGL guy trys to twist my words and argue with me is because most of ATi's optimizations are built into the hardware and cannot be disabled, there is no alternate method. Most of nVidia's optimizations by contrast are driver level and the hardware is capable of traditional operation. So Ati has little choice but to oppose choice, to coverup the fact their hardware operates only in shortcut mode.
I think you will find that many ordinary people are better acquainted with their video control panel than you give them credit for. They may not always fully understand why a setting has the effect it does, but that doesn't stop them experimenting and knowing the results the like.
Like I said early on, provide sensible defaults by all means, but allow those who want to change things to change them.
The reason OpenGL guy trys to twist my words and argue with me is because most of ATi's optimizations are built into the hardware and cannot be disabled, there is no alternate method. Most of nVidia's optimizations by contrast are driver level and the hardware is capable of traditional operation. So Ati has little choice but to oppose choice, to coverup the fact their hardware operates only in shortcut mode.
If you've looked up anything to do with AA so far then I really would pay any credence to what the hell you are reading!!
And, no, the term is sparse sampling because the hardware has enough subpixel accuracy to sample positions from anywhere in a 12x12 grid. "Rotated" FSAA is just a name to describe the common "off angle" 4x AA you see with sparse and jittered sampling systems, and "jittered" actually relates hardware that that produces FSAA sampling positions by duplicating the geometry, offsetting each of the copies from the original point slighting (AKA "jittering") and then rendering each of the sets of geometry separately.
NVIDIA also features similar optimisations as ATI, such as compression and early ZCULL - these are also hardware optimisations that can't be turned off (at least, by end users), these are not bad things. (And Gamma corrected AA is not an optimisation since there is actually more maths involved, not a lot, but there is more to consider)
radar1200gs
Regular
I never argued that compression early z are bad things. OpenGL guy tried to put those words into my mouth. Go check the posts.
radar1200gs
Regular
Z compression can be turned on and off in rivatuner if desired, Gamma correct AA may not be an Optimization in that it does not speed anything up, but it does modify the image you see. Anything that can modify the image you see should be toggleable.
Any image that is displayed via a CRT has a gamma ramp applied to it - just averaging FSAA subsample values outside of gamma space will give an image that is less appropriate for the display. Gamma corrected AA is actually giving an image closer to what is supposed to be displayed and thats why people view it as looking "better" as it is actually more correct than AA without it! You should be questioning systems without gamma corrected AA as these are "altering the image" (in an incorrect way) more than those that have the gamma ramp applied in the subsample weighting.
Ah, OpenGL guy! Since you're talking about gamma corrected AA anyway, maybe you could answer a question I've been totally unable to find an answer to anywhere.OpenGL guy said:
Is the Gamma corrected AA programmable? As in can you change the level of gamma to correct for?
It really should be an option in the driver where you could fill in the gamma of your screen to get the best results. The lack of such an option makes me belive that it's hardwired at 2.2 or something like that. But a hardwired gamma value would be wrong if you turned up the gamma in game. Like that guy in a D3 thread that said he increased the gamma to 2.2. In that situation no gamma correction would be the right thing to do, any 'gamma correction' would actually be 'gamma error introduction'. This is of course not a problem if the gamma correction is programmable, but I have never seen anything that suggests that it is, even though the issue has been brought up before.
[url=http://www.beyond3d.com/reviews/ati/radeon9700pro/index.php?page=page4.inc said:
Our understanding was that it will use the ingame values if there are any, otherwise apply a default.
DaveBaumann said:
nooo! now you've gone and ruined it!
I wanted to hear it from OpenGL guy. Now he might use that as an excuse to never answer my question, and I won't know for sure and I'll never get to know why there isn't any way of telling the drivers the appropriate gamma for my monitor.
u bastard
Actually radar1200gs is quite good in this game. In few messages the discussion moved from bashing NV3x to talk about gamma corrected AA. So, if you get into a bad situation, just make some crap out of thin air and people will forget the first discussion and move to second, giving you a better position.
Seriously, I never thought I would see a discussion about "is gamma corrected AA a bad thing or a good thing".
Seriously, I never thought I would see a discussion about "is gamma corrected AA a bad thing or a good thing".
Yeah, I almost forgot that this message was originally about the content of the State of 3D article. And here we somehow roped in OpenGL Guy, and so far he hasn't had a chance to comment on the article itself.
Gamma corrected AA is the bomb. Also, it just so happens that the sparse sample pattern that ATI uses for 4X AA looks like a 4 point rotated grid pattern, the 2X looks like a 2 point rotated grid pattern, and the 6X looks like the 6 random blood splatters against my wall.... oh wait
Gamma corrected AA is the bomb. Also, it just so happens that the sparse sample pattern that ATI uses for 4X AA looks like a 4 point rotated grid pattern, the 2X looks like a 2 point rotated grid pattern, and the 6X looks like the 6 random blood splatters against my wall.... oh wait
Ooh, that's an idea! We can each take turns coughing up blood from the laughing, and we can figure out what kind of sampling pattern it is!
I'm now forced to wonder how long this thread could become... it scares me.
I'm now forced to wonder how long this thread could become... it scares me.
We're talking about games here, not mission-critical simulations. If ATi or nV find it's a better speed/perf trade-off to "hard-code" a feature in order to include it at a reasonable price, that's what I'd want. Surely you'd expect driver-enabled optimisations to be slower than hard-coded ones, or more flexible hardware to cost more?radar1200gs said:
I can't fathom why you're picking on gamma correction--well, not in any way that reflects well on you. At any rate, I'm done wasting time arguing with you. Unless you're that rare nVidia employee who can post online, I don't see why convincing you NV30 was inferior to R300 is worthwhile. The evidence is out there; at this point, if you choose to ignore it, it's really your loss.
Dave, so the Voodoo 5 is the only consumer 3D card that did (a form of) JGAA?
OpenGL guy
Veteran
Exactly. And since they don't understand the current settings, why give them more settings they won't understand? Why invite support calls?radar1200gs said:
You do a good job twisting your own words. Your arguments change with every post and you never reply to the points raised but instead choose to respond to the most irrelevant portion of a message. Also, you have no clue how ATI hardware works. You have no clue how the driver works. But by all means keep on posting nonsense.
What does rivatuner have to do with nvidia's driver? Does nvidia offer the same functionality in their control panel? No? Why not?radar1200gs said:
Did it ever occur to you that the optimizations that ATI and nvidia have put in their hardware and software come at from an investment in time and money? To allow your competition to evaluate the effectiveness of said optimizations is foolish at best.
Lastly, the end-user does not benefit from such choices. Disabling these optimizations only makes things go slower and who in the world would want their games to go slower? I guess only "intellectuals" such as yourself.
- Status
Similar threads
