cthellis42 said:
Well, from what I've heard OEM's already have it. Although, its a completely different R420 to the one we'll see!
what is the current state of NV50 ?
- Thread starter Megadrive1988
- Start date
Well, from what I've heard OEM's already have it. Although, its a completely different R420 to the one we'll see!
cthellis42 said:
No, I don't. Is it really any secret my ATI sources always, well, sucked?
If you want ATI info, ask Wavey, not me!
Megadriver: Oh no, NOT again! I must have explained the meaning of ILDP at least 10 to 20 times in my life! Do a forum search, damnit, hehe.
Uttar
Gunhead said:
True, but one comment though.
Even though gates have varying number of transistors, it's quite often that a "gate" in such cases are equal to 4 transistors (like a NAND gate). In fact, I'vee seen tools that reported number of transistors and number of gates for the synthesized design. But it looked a little bit suspisious so I tested it. And indeed, the "gate count" was actually transistor count/4, not the actual number of gates.
Still, 120M*4 is a lot.
Shadow Volume generation is done in software in DOOM3. I doubt JC will implement a hardware Shadow Volume generation technique.Hyp-X said:
It is currently possible to generate shadow volumes in hardware with current hardware (at least with VS 2.0, probably with earlier versions as well).
arjan de lumens
Veteran
How? If i have understood the DOOM3 shadow volume technique correctly, it basically scans an object for polygon edges between lit and unlit polygons and extrudes those edges away from the light source, which at the very least implies the creation of new vertices/polygons, something which is not within the capabilities of VS2.0.Chalnoth said:
arjan de lumens said:
Yes and no.
It can be done if you precreate all possible vertices and triangles.
That resulting mesh has approximatly 6x vertex count and 4x triangle count.
There are two set of vertices 1 for the front and 1 for the back of the volume. The vertices include the face normal to correctly test it it's in front of the light or not.
There are quads (2 tris) at every original edge between the two triangles at the sides of the edge. If the vertices at the sides of the edge turn out to be at both front/back then this quad will be degenerate (a line), otherwise it will be a side part of the volume.
Hyp-X said:
Wow, if image quality would improve by a signifigant margin and Nvidia can get a NV40 card to Carmack soon enough, that would be really great to see such an advanced technology taken advantage of right away with a big game like Doom 3. A while back Carmack was complaining about certain harware vendors approachs to high order surfaces. Does the PPP fit into the approach Carmack wanted to see?
Not only the Doom 3 engine gains from the PPP, but obviously all the other games based off the engine. I really hope the PPP happens in Doom 3.
The PPP myth has been lurking around for more than one generations now; I'd say it would be safer to start asking around how things look like with recent DX-Next revisions and said programmability, before jumping to any preliminary conclusions.
Also the rumours about a pure stochastic algorithm don't sit quite well with my "instincts" either. I have severe doubts that a low sample amount stochastic makes more sense than current sparse MSAA algorithms. At the other we all know what higher than say 16/32 samples mean in terms of bandwidth and memory footprints.
Just my 2 cents.
Also the rumours about a pure stochastic algorithm don't sit quite well with my "instincts" either. I have severe doubts that a low sample amount stochastic makes more sense than current sparse MSAA algorithms. At the other we all know what higher than say 16/32 samples mean in terms of bandwidth and memory footprints.
Just my 2 cents.
I think stochastic should start to look good closer to 8 samples than 16-32. So, it might be about time to start offering stochastic, but if stochastic is offered, then a static FSAA algorithm should also be offered (and should be trivial to support if stochastic is supported...).Ailuros said:
The point is that 8x sample stochastic would have to deliver better results than 8x sparse MSAA to justify the increased hardware cost of the first. So far I haven't seen or read a single commentary in these forums and in more than one occassions that less than 16x samples are worth the effort. Experiments and their results so far don't seem to have shown different results either.
Even more supporting my point above. Then why develop and support two different algorithms when you can have just as good if not better overall results up to a certain threshold with just one algorithm?
Keeping the existing algorithm wouldn't help much either hypothetically (to combine with 8x sample stochastic). 4x sample AA would continue to be widely used in the majority of cases, OGMS would continue to lackluster against the competition and mainstream and value products would have an even bigger problem.
Even more supporting my point above. Then why develop and support two different algorithms when you can have just as good if not better overall results up to a certain threshold with just one algorithm?
Keeping the existing algorithm wouldn't help much either hypothetically (to combine with 8x sample stochastic). 4x sample AA would continue to be widely used in the majority of cases, OGMS would continue to lackluster against the competition and mainstream and value products would have an even bigger problem.
What we have seen, however, is that even with ATI's 6x sparse AA, there are angles where significant aliasing occurs (If I remember correctly, the most severe aliasing is on a diagonal line from top left to bottom right, or 135 degrees from the positive x axis), and somewhat less aliasing perpendicular to that line.Ailuros said:
A stochastic or procedurally-changing AA sample pattern could utterly eliminate any angle preference to the AA algorithm, which would make it close to perfect (instead of jagged lines, you'd get single-pixel aliasing issues that would be far less noticeable).
But, it may be that 8 samples is too few for a purely stochastic algorithm. It might be better, in terms of visual quality and hardware implementation, to instead use a procedural method (that would use one of several sparse patterns for each pixel).
That's not the issue here. The issue here is to effectively surpass the competition (in this case ATI) - preferably whatever improvements they will have in their next generation products - on a quality/performance combination ratio. I don't see the NV3x having there anything to match ATI's 6x sparse sollution from that perspective, let alone the more commonly used 4x sparse option.
jvd said:
If you use some sort of analytical anti-aliasing approach, then certain edges should get varying amounts of samples anyway. Those edges will cross in most occassions more than one tile; what exactly have I missed here? (not to speak that I don't see why the above shouldn't be more or less the same with IMRs and tiled back buffers).
Similar threads
