NVIDIA GF100 & Friends speculation

yes I did, and I responded to your BTW. This is the same place where the HD5970 gets around 65 FPS ;), that also tells you there isn't much tessellation going on.

you did what? have you post a 480 screen shoot with tessellation off at this place?
 
you did what? have you post a 480 screen shoot with tessellation off at this place?


I just explained to you where tessellation was happening and not happening, its really easy to see, if you don't have experience with tessellation fire up 3dsmax (don't use sub-d, although it does somethings like tessellation its a bit different), and do some scan line renders with AA off, you can tell the difference between textures, aliasing, aliasing due to tessellation, etc, if you know what to look for its like night and day what the artifacts are when tessellation with no AA.
 
The sad part? Not that big of a change, double the SIMDs, add another piece to the front end, spin spin spin.

Aren't you being a little unfair here? Besides the new tessellation/rasterization stuff, and doubling SIMDs, didn't they also change the cache architecture, add fast DP, and rumored to have changed the warp scheduling stuff? I mean, by these standards, AMD hasn't done anything significant since the R500.
 
I think it's fair to say that any game shipping in near term that supports DX11 tessellation is probably sporting it as a tacked-on afterthought, and that we will not see what taking full-advantage of tessellation looks like for 18 months. Ditto for DC. Ungine looks cool and all that, but it's simultaneously under-kill and over-kill. Under-kill because it doesn't look like their Hull Shaders/Domain Shaders are doing anything fancy other than displacement mapping. Over-kill because they appear to be tessellating too much overall.

Here's an example of an effect that I think would rock. Imagine an explosion going off like in "The Hurt Locker" and the physics engine senses a simulated ground-shock wave that travels outward displacing level geometry, and/or kicking up objects in contact with rigid structures. If you've seen the slo-mo explosion, you know what I mean. And this is just an effect, not even general purpose application of TS to enhance game quality. Take idTech 6/Rage, using mega-texture to apply super-high detailed displacement maps over the entire level. I think this would make Ungine look like a joke with the repetitive brick/cobblestone displacements.
 
Aren't you being a little unfair here? Besides the new tessellation/rasterization stuff, and doubling SIMDs, didn't they also change the cache architecture, add fast DP, and rumored to have changed the warp scheduling stuff? I mean, by these standards, AMD hasn't done anything significant since the R500.

Not to mention DX11, significant changes to the SMs, ECC, the global hierarchy... It's really hard to argue that Fermi is anything less than a major evolution. Whether that evolution turns out to be successful is another story.
 
I just explained to you where tessellation was happening and not happening, its really easy to see, if you don't have experience with tessellation fire up 3dsmax (don't use sub-d, although it does somethings like tessellation its a bit different), and do some scan line renders with AA off, you can tell the difference between textures, aliasing, aliasing due to tessellation, etc, if you know what to look for its like night and day what the artifacts are when tessellation with no AA.

what a pity razor1, I guess you still have sth to say


480:

http://bbs.pczilla.net/attachments/month_1003/10030907221f7737ee82e7a341.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video with Tessellation on

http://bbs.pczilla.net/attachments/month_1003/10030907221756b9994258a635.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video only Tessellation off

http://bbs.pczilla.net/attachments/month_1003/100309072332b70c94d96c3616.jpg

5870 1GHz/1250MHz (core/mem) Tessellation on/off comparing

http://bbs.pczilla.net/attachments/month_1003/100309072324c44bc75a2f937a.gif
 
Not sure this has been posted before, look like gtx470 SLI
1003080032f3987bf567ef4af6.jpg
 
Power is converted to heat neliz, not voltage. Physics 101. No matter what voltage targets they missed the result has to come in under their total power budget.

and one way to do that is lowering Tj such that you generate lower power. You lower Tj via more aggressive cooling. With something the size of 470, the difference between Tj of 80 and 70 C could be 20-25W!
 
what a pity razor1, I guess you still have sth to say


480:

http://bbs.pczilla.net/attachments/month_1003/10030907221f7737ee82e7a341.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video with Tessellation on

http://bbs.pczilla.net/attachments/month_1003/10030907221756b9994258a635.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video only Tessellation off

http://bbs.pczilla.net/attachments/month_1003/100309072332b70c94d96c3616.jpg

5870 1GHz/1250MHz (core/mem) Tessellation on/off comparing

http://bbs.pczilla.net/attachments/month_1003/100309072324c44bc75a2f937a.gif

My guess is the gtx480 wouldn't get to 66fps at this part of the benchmark under any reasonable circumstance. The card really gives off the impression of being a stayer rather than a sprinter, which is no bad thing in itself.
 
You should be able to change from fly through mode into free camera mode, like in Petersen's video, and see if you can really home in on the absolute worst case frame :p

Jawed

I don't have access to that pc right now but I've been spending a little bit of time getting into the guts of unigine.

Around about here is the gtx480's worst frame, which according to Peterson's video is near 155 seconds, and a mere 33/34 fps. The wireframe pic I took here on my 4770 is only dx9 without tesselation, but I noticed how my 5770 struggled at this point too - interestingly it wasn't as bad here for the 5770 as it was at the dragon, but this is around where the gtx 480 hit bottom fps. - a good 10fps less than at the dragon I think.

00036.jpg




comparisonbenchmark.png
 
Last edited by a moderator:
Yep, all I'm saying is that simply applying a LOD system to today's models isn't going to benefit anyone much. We need to see a tangible increase in the maximum LOD as well and displacement mapping is going to have a much bigger impact than simply smoothing out pointy heads and elbows.



Well LOD popping today isn't even a matter of triangle density. We have entire boulders snapping into view in front our eyes so the problem is far worse than just the lack of tessellation :)

The problem is also that significiant factor (much bigger than the hardware which was the oposite years ago) of the visuals depends on the developers and artists and not perfect looking lifeless engines.
If u watch God of War 3 or Final Fantasy 13 on PS3 than u need to wonder what level graphic can they reach with such limited hardware but with that amount of developers, time and excelent artists. Its hard to even imagine what would they reach with a 1GB 5870 card in that console.
And as the graphics will increse even more burden will fall into developers hands.(the tesselation is one of them :LOL:)
 
I don't have access to that pc right now but I've been spending a little bit of time getting into the guts of unigine.

Around about here is the gtx480's worst frame, which according to Peterson's video is near 155 seconds, and a mere 33/34 fps. The wireframe pic I took here on my 4770 is only dx9 without tesselation, but I noticed how my 5770 struggled at this point too - interestingly it wasn't as bad here for the 5770 as it was at the dragon, but this is around where the gtx 480 hit bottom fps. - a good 10fps less than at the dragon I think.
Maybe there's lots of blending going on, with the grass and smoke? Or the volumetric shadows (god rays) shader is consuming lots of shader time? Or both. etc.

Overall the shapes on the graph of the two cards match very closely - the divergence here is notable. But it would need more detailed testing to say why they diverge, I can't think of anything definite.

Jawed
 
Back
Top