R520 Running

[Ailuros]

I do own a 21" CRT and you don't actually need much of antialiasing in that case since the monitor is doing some sort of "upsampling" by itself.

The manufacturer lists only the horizontal dot pitch, but I'm guessing that the vertical dot pitch must be at 0.25.

355mm/0.25 = 1420

In other words even 1920*1440 is already "stretched" for such a monitor. Even with a 340MHz video bandwidth it yields only a 75Hz refresh rate, which is perfectly tolerable on one side, the viewable space is just too small for such a resolution. You can read any form of text or handle 2D flawlessly up to 1600*1200, above that: guess again.

Yes current 256MB high end cards can handle under most occassions 2xMSAA in 2048, but performance still is way behind what I'd consider playable.

Now I personally had the impression that CRTs are a dying breed (and along some people with similar preferences to me). I do hope that you don't mean purely hypothetical TFT/LCDs with a 2048 native resolution on the other hand.

I think you're measurements are off there with respect to the vertical length and dot pitch of your monitor (or you have one very oddly shaped and spec'd 21" monitor). Either way bringing this back to my point, R420 = still slow at extreme resolutions in the more stressing games. R520 = hopefully must faster at extreme resolutions and could make gaming at them more feasible in these games.

I blame it on the typo-devil; it's actually 305mm which makes the result of any possible calculation even worse.

http://monitor.samsung.de/article.asp?artid=869FFA74-D1F0-4522-B5CD-9B608B834905&show=specs

Eizo has CRTs under the discontinued section for the record:

http://www.eizo.com/support/discontinued/crt/t966.asp

Yes bringing it back to the point, real time experience would show you that a 21" monitor is in no way adequate for optimal output. The difference can be seen in real time. In order for such a resolution to make sense we need bigger display areas and video bandwidth on monitors, and for the time being I cannot see anything in sight when it comes to TFT/LCDs unless there is a sollution out there that costs as much as a small island.

Up to 1600 is perfectly enough IMHO; I'd rather have antialiasing algorithms with much higher sample densities than today, than resolutions scale endlessly, but that's just me.

16x AA is insane. 8x would be fine.

8x sparced MSAA would be wonderful; I wonder if upcoming sollutions will still only be capable of 2xAA per cycle.

 

[jvd]

I dunno , i see no problems with temporal aa on the x800xt pe . Once in awhile i get some flashing . But nothing major and it definitly looks much better than with out it on .

 

[ChrisRay]

16x AA is insane. 8x would be fine.

I agree. 16x just isnt needed. Nvidia already supports a hybrid RGMS+OGSS version of it and it just doesnt seem that practical to me. Then again.. neither does 8xAA, Something which people need to understand, (Which Ailuros has always pointed out in the past). There are other effective ways of improving quality without adding more samples.

 

[digitalwanderer]

There are other effective ways of improving quality without adding more samples.

I don't care what number they give it or what they call it, that is the result I'm hoping to see...from both sides.

 

[Reverend]

I agree. 16x just isnt needed.

16x super-sampling AA at 1600x1200 would probably be almost nearly perfect, edge-wise and spatial-wise.

We pursue and need perfection in order to progress in the right direction, right?

 

[ChrisRay]

I agree. 16x just isnt needed.

16x super-sampling AA at 1600x1200 would probably be almost nearly perfect, edge-wise and spatial-wise.

We pursue and need perfection in order to progress in the right direction, right?

I think there are better uses for the rendering power than 16xAA super sampling could provide. Not only are we talkign ridiculous fillrate requirements, but framebuffer requirements as well. We need to worry about improving performance in our current methods, And perhaps do single cycle 4x AA before I'd even begin to worry about 16xAA.

I would rather have the illusion of perfection if it comes at a practical cost, Than perfection at an impractical cost. I am viewing this from a more practical view than anything. I feel 16xAA in our current form of multisampling/supersampling is fairly impractical and resources could be better spent.

 

[Reverend]

Oh, I didn't know you were talking about what's currently available.

 

[Hellbinder]

No..

What i was refering to is Completely Different than MSAA or SSAA or Temporal or any mix of the above.

Its something That will allow for perfect AA of Pixel Shader Heavy titles, as well as other things. The Performance should be very, very, very, Good.

That is all i am going to say.

 

[digitalwanderer]

No..

What i was refering to is Completely Different than MSAA or SSAA or Temporal or any mix of the above.

Its something That will allow for perfect AA of Pixel Shader Heavy titles, as well as other things. The Performance should be very, very, very, Good.

That is all i am going to say.

One not allowed follow-up then, will it only work on hardware that isn't available yet or do I have hopes with me X800 pro of playing with it? :|

 

[karlotta]

I agree. 16x just isnt needed.

16x super-sampling AA at 1600x1200 would probably be almost nearly perfect, edge-wise and spatial-wise.

We pursue and need perfection in order to progress in the right direction, right?

Gimmie now!

 

[AndrewM]

Hellbinder smoking shit again?

 

[Ailuros]

I agree. 16x just isnt needed.

16x super-sampling AA at 1600x1200 would probably be almost nearly perfect, edge-wise and spatial-wise.

We pursue and need perfection in order to progress in the right direction, right?

You know better than me that finding the fillrate/bandwidth and framebuffer requirements for 16x sample SSAA in 1600 is going to be close to impossible for the foreseeable future. SA proposed once sample positions for either 8x or 16x sparce-whatever. Somehow I have the feeling that 16x sample (non-ordered) patterns are at the edge of where we'd rather have far more sophisticated algorithms instead. I'm not even sure anymore if we ever reach such sample densities before switching to entirely different algorithms.

I see Falanx's Mali100 being capable of single-cycle 4xRGMS, on a ultra small chip and hence I'm wondering if and when it might appear on PC high end architectures.

Just for the record there's a quite old fillrate tester from zeckensack, named Archmark. With 16xSSAA and trilinear quad texturing I get a pathetic 88 MPixels/sec; that compared to the theoretical maximum bilinear fill-rate of 5.6 GPixels/sec (2.8 tri) is quite frankly ridiculous.

 

[Ailuros]

Someone correct me if I'm wrong, but AFAIK ATI hasn't used anything alike read@resolve/filter@scanout on PC products have they?

I saw a read@resolve mention for the eDRAM module in the ancient Xenon diagram; I can't really figure as a layman why anyone would want such a technique with eDRAM, yet if there's any merit to it, there must be a good reason why they're using it.

By the way I'm not in the least interested in any (let me call them) "software optimisations" when it comes to antialiasing. I'm not throwing them all into the same pot, yet I still prefer regular 2xRGMS over Quincunx and I still prefer 4x or even 6x sparced in 1280 or higher than any so called "temporal-AA".

I don't see many (besides obviously people like Reverend) caring about anything but polygon edges/intersections. There's more to full-scene antialiasing than those; frankly I could easily live with "only" 2x or 4x sparced MSAA in high resolutions as long as I'd see far better texture filtering implementations than today. Compared to AA, the AF topics are far too few IMHO.

People around here used to dig up some highly interesting whitepapers that follow any kind of research/advancements with anisotropic filtering. Uhmm anyone?

 

[Geo]

What about the Matrox approach to 16x? Would that be workable with a large dose of driver monkeys banging away at fixing its problems? Could you mix-mode it, laying it over 4xmsaa? That would give you 64x on most edges, and hide the worst of your edge-intersection issues as they'd still be 4xmsaa there?

Of course we are well into dangerous territory for my level of expertize, so I'll apologize in advance if that is an eye-rolling suggestion for those of you who actually know what the hell you're talking about.

 

[_xxx_]

There were some fundamental problems with that algorithm. I can't remember what it was, just that it was something pretty bad. It was probably discussed here before, maybe a search could help?

 

[Geo]

There were some fundamental problems with that algorithm. I can't remember what it was, just that it was something pretty bad. It was probably discussed here before, maybe a search could help?

Intersections (which I mentioned) was a big one. Some stencil issue too, tho there were also reports they fixed it on a later rev of the chip.

At least for edges I think some kind of adaptive technique like FAA could be a big win at relatively small performance cost --but would need to be combined with other techniques as well. At least short-to-mid term. It seems like a 4x base level of AA has been pretty much been driven even to the mid-range at this point, so it ought to be time for a nice step-up on the high-end.

 

[Hyp-X]

FAA and Z3 can be viewed as a lossy compression for MSAA.
Above a certain number of samples lossy compression should provide better quality/performance ratio than a lossless one.

 

[digitalwanderer]

Hellbinder smoking shit again?

Sorry HB, I HAD to!

Mebbe, but what does that have to do with the discussion? :|

 

[tEd]

Where is sireric when you need him. hint hint

We didn't get a christmas present yet? hint hint Please??!!

The lack of information at this point is so uncool.

I drink too much coffee....

 

[digitalwanderer]

Where is sireric when you need him. hint hint

We didn't get a christmas present yet? hint hint Please??!!

The lack of information at this point is so uncool.

Secondededededed!!!!! :?

I drink too much coffee....

That's impossible if your hands are still steady enough to type.