nVidia release new all singing all dancing dets.
- Thread starter BoardBonobo
- Start date
Doomtrooper
Veteran
I have no problem with that statement....the obvious optimizations on BENCHMARK appz only
Gotta get back to my game of 3Dwinbench...
Gotta get back to my game of 3Dwinbench...
Doomtrooper
Veteran
Which one. I have so many I can't keep track..damn it I was just getting into the 3Dwinbench...
Oh well..
Oh well..
Doomtrooper said:
"point sampling offering which lowers IQ yet inflates 3Dmark scores (default graphic setting I might add)" is wrong, because it doesn't inflate the scores (compared to bilinear filtering). The image quality/performance issues are not related to the point sampling issues. I thought I'd point it out clearly to give you an opportunity to admit that part was wrong since the popular act seems to be to bait you instead.
What? You have no problem with saying something that is wrong? Point-sampling doesn't help the 3DMark score. I pointed this out somewhere in this thread but I'll do so again, if you like (whether anyone will believe the figures or not is another subject though).
It doesn't even help a texture heavy benchmark like PVR Temple (again, look back earlier in thread). I tried the PVR Fortune fill rate test, and once more, there's no difference between nearest-point and bilinear.
Edit: Damn, I'm too slow at this typing malarky...
It doesn't even help a texture heavy benchmark like PVR Temple (again, look back earlier in thread). I tried the PVR Fortune fill rate test, and once more, there's no difference between nearest-point and bilinear.
Edit: Damn, I'm too slow at this typing malarky...
Doomtrooper
Veteran
I have so many friends, whats the matter boys..truth hurt. Yes I may have been wrong with performance gains but I wasn't wrong with IQ..or optimizing for benchmarks.
You see, I've seen these drivers before alot of you...
You see, I've seen these drivers before alot of you...
Doomtrooper
Veteran
walkndude said:
How would you know kid, you just joined or are you another fan site alias member.
Doomtrooper
Veteran
Inside information on Nvidia..no way..this came by email..a little heads up for some honest reviews for a card that was just released, I really don't care if you don't beleive me.
Well, shall we recap all the various discussions and issues and topics that have been coming up here?
1) NVIDIA added anisotropic filtering to the control panel for D3D (a very GOOD thing as it's loooong, loooong overdue), and by futzing around with the settings there, a user can wind up with "no af" settings wrongfully being in the registry as point sampling. The worst case scenario of this "bug" is there are bound to be websites reporting point-sampling benchmarks wrongfully as "application preference." All the GF4 users in the world are going to use RivaTuner or NVMax like they have since the days of the GF3, so it really doesnt effect them.
2) NVIDIA has, over the past several revisions, proceeded to slowly reduce the quality/effectiveness of 8x anisotropic filtering to improve performance. As 8x AF performance was the keypoint to be harped on from product release, as well as competition from other IHVs, it would make sense to find some way to get these benchmarks to improve. The little 4x/8x gallery I provided was at the 29.42 level (i.e. the "follow up to AF" drivers from the 28.32's that got so harped on this feature). It appears the 4x->8x transition has been reduced even further in the 40.xx drivers.
3) Some people believe that games like Flight Sim 2002, Motocross Madness 2, Motoracer or F1/racing sims are a good way to measure bandwidth/fillrate performance in order to try and prove superfluous points concerning AF. The rest of the community knows well enough that certain tests within 3dmark, as well as heavily CPU bottlenecked games dont make much sense nor are good methods to pinpoint anisotropic filtering performance, especially when combined with benchmark accuracy (say upto +/- 4% for lots of benchmarks with same settings can be yielded).
4) Many people have noted how it is interesting that the Nature test in 3dmark2001 seems to be the most dramatically effected by the 40.xx detonator drivers. To date, no one has found anything else to illustrate a performance edge with these new drivers so the pursuit continues.
You can make up your own motives for the above. Trying to determine motive isnt a cut and dry thing, regardless of how obvious or how much evidence there is in support of a given motive. It is much easier to instead simply assume it was a "lucky coincidence" or similar for issues to work in favor of one result or another.
Cheers,
-Shark
Doomtrooper
Veteran
I posted a link to Aquamark a few previous pages back, since no one will show screenshots comparions and speed differences for the different modes, and it is a pretty stressing benchmark with PIXEL SHADERS like 3Dmark..maybe you can Shark.
68Megs
68Megs
Sharkfood said:
Well, if websites reported point-samplinc scores that would indeed be sad for NVIDIA since it's a little slower than the DEFAULT setting.
Sharkfood said:
Please back up that last claim. 8X might have been buggy for some time, but from my knowledge the other settings (and ALL settings in OpenGL) are just fine and the way they have always been.
Sharkfood said:
Yes it's very interesting from a technical standpoint. Considering what type of effects Nature use, and what types of effects a game lika Aquanox use. But "optimizing" Nature is in itself very dumb because it doesnt add very much to the final score. How silly of NVIDIA.
Typedef Enum
Regular
I would just like to take this opportunity, once again, to thank Doom for ruining yet another thread...turning into a "I have contacts, I know what I'm talking about" or flat out not understanding basic fundamentals...or even better, resorting to "look kid..."
Do you not understand that this point sampling setting does not taint performance numbers? Do you not understand that this is not the default setting? It's abundantly clear to anybody who doesn't have a closet full of ATI T-shirts that these are the facts...and yet, you still insist on things which are not true.
The only thing that you can say about point sampling is that it should never have been there in the first place, and it's an obvious error...nVidia has nothing to gain...and everything to lose...by having it in there. But as it relates to benchmarks, it's a non factor. Got it?
Do you not understand that this point sampling setting does not taint performance numbers? Do you not understand that this is not the default setting? It's abundantly clear to anybody who doesn't have a closet full of ATI T-shirts that these are the facts...and yet, you still insist on things which are not true.
The only thing that you can say about point sampling is that it should never have been there in the first place, and it's an obvious error...nVidia has nothing to gain...and everything to lose...by having it in there. But as it relates to benchmarks, it's a non factor. Got it?
Doomtrooper
Veteran
Typedef Enum said:
I'm not referring to point sampling, I'm referring to optimizing for benchmarks...This thread was runined along time ago...
These drivers were released a couple weeks back....take that to the bank
Oh my "take it to the bank" hehe.
Well, here are some Aquamark numbers if you like. Sorry don't know how to take screenshots at the same place in that benchmark. If you wanted numbers compared to older drivers I'm sorry I can't give them to you. I have changed drivers back and forth enough.
no aniso: 68,7
0X 68,9
1X 68,1
2X 36,9
4X 25,1
8X 25,1
"take it to the bank" hehe.Well, here are some Aquamark numbers if you like. Sorry don't know how to take screenshots at the same place in that benchmark. If you wanted numbers compared to older drivers I'm sorry I can't give them to you. I have changed drivers back and forth enough.
no aniso: 68,7
0X 68,9
1X 68,1
2X 36,9
4X 25,1
8X 25,1
Similar threads
