Top or bottom? image quality analysis
- Thread starter Snarfy
- Start date
Bottom looks much better. The top looks like an FX card running at FP16, all the blockiness on the tiles from the flashlight.
________
Ford towcommand picture
________
Ford towcommand picture
Last edited by a moderator:
The key is that either it's using FP16 in the wrong place, or it's something else entirely. Either way it is ludicrous to think that this is indicative of how the GeForce 6800 should look if programmed for properly. It's a bug, and it will be fixed, at little to no performance hit.SsP45 said:
I hope this isn't a defensive response to Sniping's link to MikeC's IQ comparison. Obviously we can't fully judge AA with static screenshots, whereas we can with texture quality.
If NV40 is running NV35's hacked paths, then it's doubly unfair to compare IQ. I'm curious why Crytek was able to implement PS3.0 in three weeks and we haven't seen it in a patch in time for the NV40 launch, though I suppose timing was tight. We'll see soon enough if NV40 is as fast with equivalent IQ in Far Cry. I didn't read any IQ complaints for Halo, though, and NV40 performed just as well there.
If NV40 is running NV35's hacked paths, then it's doubly unfair to compare IQ. I'm curious why Crytek was able to implement PS3.0 in three weeks and we haven't seen it in a patch in time for the NV40 launch, though I suppose timing was tight. We'll see soon enough if NV40 is as fast with equivalent IQ in Far Cry. I didn't read any IQ complaints for Halo, though, and NV40 performed just as well there.
jimmyjames123
Regular
Quite obviously the bottom image is better because it is using PS 2.0 while the top image is using some form of the NV3x settings. And also, obviously NVDA is going to fix these IQ issues in their newer Forceware drivers. The NV40 does not need to use NV3x reduced PS settings
Snarfy said:waiting for the appropriate drivers for the nv40 before comparing far cry screenies is like waiting for the x800 to be launched before we assume nvidia has won =) sure, we'll find out the truth in a bit, but has that stopped the rumor mongering?
touche :?
Well, that makes about 0 sense. But hey, more power to you.
jimmyjames123
Regular
LOL, i agree with you, that statement doesn't make any sense. Looks like you did not order your words properly at the end of the sentence, but mistakenly reversed the order of what you were trying to say.
oeLangOetan
Newcomer
Chalnoth said:
read the b3d review, it's a 33% hit when you switch to FP32
(of couse there are many more factors here but there will be a hit in performance)
It's not as simple as that on the NV40. The NV40 doesn't appear to have any register size limitation (though I'm not sure anybody's yet tested to the maximum number of registers yet....). Using _pp still helps for nrm and sqrt functions, and may help for other special functions (sin/cos, ln/exp). These are functions that are frequently used in lighting, and _pp is enough for these situations when used in lighting, so I see no reason why fixing the shader in this way would cause a performance hit on the NV40, and I doubt it would even cause much of a hit on the NV3x.oeLangOetan said:
Anyway, I thinking I was wrong on it being _pp. It's probably a texture lookup that's causing the errors: the errors do look a lot like point sampling (which doesn't mean that they are caused by point sampling, so I could still be wrong on the exact cause). A per-pixel lighting shader can use a cube map lookup to get the normalized light direction. If the shader uses point sampling for this lookup, you could get the blocky lighting shown. I think the NV40 would actually be faster if it switched to calculating the normalized vector. I think the NV3x should have similar performance with the cube map or the calculation.
davepermen
Regular
Lezmaka said:Pete said:
why do all people think this? ps3.0 is in dx since the start of dx9. the ONLY thing that is not in yet, is a compiler for hlsl to ps3.0. this is only a developer thingy, not at all a runtime thingy. people can write and run ps3.0 now, having proper hw (or software emulation).
is that really that difficult to grasp? (seen this stated that often i think it is
Hey, I care. Perhaps you won't believe this, but I really do notice image quality during game play.Diplo said:As someone who enjoys playing video games my answer would be : who cares, you don't walk around in an FPS staring at the floor
btw, the bottom image looks better imo, and I would notice that when playing.
