Hm.. i do against my will (3Dfx fanboys will love this) but..
http://www.tcinternet.net/users/pctoolbox/benchmarks.htm
http://www.tcinternet.net/users/pctoolbox/benchmarks.htm
Interesting Voodoo6000 benchmarks
- Thread starter Mummy
- Start date
LittlePenny
Regular
How old was the guy that wrote this webpage? He talks like Hellbinder 
Hmm, a guy on a forum claims to have emailed those ppl and they told him the benchmarks where done with 4XAA and 8XAA enabled, is it possible that with aa enabled that the card is so fast ?
I think i read from some ppl here that Rampage would have complete 4X free AA, was it the same for the VooDoo 6000 ?
Dave / Reverend maybe can help me?
I think i read from some ppl here that Rampage would have complete 4X free AA, was it the same for the VooDoo 6000 ?
Dave / Reverend maybe can help me
?
No - it was 'effectively' fillrate free under certian texturing conditions. It wasn't bandwidth free.
Grrr! It would have been 'real' nice if people actually gave me some sort of credit when they use screenshots that I took. The 3 'good' screenshots were of course taken by me. It's easy to tell because of the custom q3 slash skin. It was a little trick I did so I would always know if someone had taken my shots. I made the skin and never distributed it.
Of course the other 2 shots are the worst things i've ever seen. If there weren't so much Jpeg artifacting, it would be seen that only half of the samples are there. It's just a driver which most people don't notice.
Anyway, 8x on the V5 6000 is slow... real slow.
Here's a bunch of shots I took ages ago with cg_drawfps 1
19 FPS - http://www.users.on.net/triforce/6000/glide0000.jpg
19 FPS - http://www.users.on.net/triforce/6000/glide0002.jpg
13 FPS - http://www.users.on.net/triforce/6000/glide0008.jpg
26 FPS - http://www.users.on.net/triforce/6000/glide0008.jpg
As you can tell, "real slow"
-Colourless
Of course the other 2 shots are the worst things i've ever seen. If there weren't so much Jpeg artifacting, it would be seen that only half of the samples are there. It's just a driver which most people don't notice.
Anyway, 8x on the V5 6000 is slow... real slow.
Here's a bunch of shots I took ages ago with cg_drawfps 1
19 FPS - http://www.users.on.net/triforce/6000/glide0000.jpg
19 FPS - http://www.users.on.net/triforce/6000/glide0002.jpg
13 FPS - http://www.users.on.net/triforce/6000/glide0008.jpg
26 FPS - http://www.users.on.net/triforce/6000/glide0008.jpg
As you can tell, "real slow"
-Colourless
Mine is 4x1, like all the other real cards. That 2x2 pic is just a mock up as far as I know. http://www.users.on.net/triforce/6000/front.jpg There is also a few other pics in http://www.users.on.net/triforce/6000/ as well
Hellbinder
Banned
I am 33 years old..
I guess that anyone who gets angry at the rediculous levels of Pro Nvidia Bias on the internet is an "immature" kid...thank you.. in that case ill stick with my immature ways...
Unlikely, 4x9tap is almost as bad as quincux (because of the blur filter apllied afterwards) and even ignoring that you'll only get 3 intermediate color graduations on the edges that are overlapping because of the shared color values (gf1/2 had a similar mode except with supersampling in the det2/3s). 8xRGSS will have 7 intermediate shades on near horizontal and vertical edges and will not overblur.
No, this is not true. Personally, I don't like the blur filters, as they do mess up text, but they are good for fully-3D scenes.
As far as textures are concerned, 8x FSAA is very roughly equivalent to 3-degree anisotropic. 8-degree aniso with 4x9 is closer to 6-degree aniso. In other words, I'm sure the 4x9 shots would have superior texture clarity, despite being a little bit worse in texture clarity than the GeForce3/4 is capable of.
Additionally, looking at edge AA, two adjacent pixels share three color values, and thus each pixel has six color values to call its own. This means six color variations in an optimal situation.
For low-angle surfaces, there will be three color variations, due to the OGMS method.
V56000 8xFSAA isn't Rotated Grid, it's more like Jittered grid. This is the sampling pattern used:
And that is exactly where 8x Jittered Grid produces the most variations. Low-angle surfaces are also the surfaces with the most visible amount of aliasing. Big Stair Steps = Very Noticable
-Colourless
Chalnoth said:
And that is exactly where 8x Jittered Grid produces the most variations. Low-angle surfaces are also the surfaces with the most visible amount of aliasing. Big Stair Steps = Very Noticable
-Colourless
The 4x9 shots would be worse in clarity for flat on shots as the v6 could use lod for a screen near 3x the res here whereas the 4x9 would have to use normal lod and then it would be blurred slightly on top.
For edge AA the 8xJGSS is a clear winner because it is able to give 7 evenly spaced color graduations, the 4x9 will only give 3 with the odd different pixel inbetween.
Even then the lines are overblurred:
4x9 will give something like the left side. True 3x3 supersampling sould give something more like the right side. The 8xjgss will give much cleaner and smoother lines than 4x9tap will because there is no colour sharing between pixels.
8xjgss may even be better at surfaces that need aniso depending on how the lod levels are calculated. Its possible to get near 8x aniso from the sampling pattern as when projecting the pixel onto an angled surface each sample occupies a differnent point lengthways through the pixel and while this isnt as perfect as the straight line of samples that aniso uses lod levels for 6 or 7 x aniso could be used on sloped surfaces.
For edge AA the 8xJGSS is a clear winner because it is able to give 7 evenly spaced color graduations, the 4x9 will only give 3 with the odd different pixel inbetween.
Even then the lines are overblurred:
4x9 will give something like the left side. True 3x3 supersampling sould give something more like the right side. The 8xjgss will give much cleaner and smoother lines than 4x9tap will because there is no colour sharing between pixels.
8xjgss may even be better at surfaces that need aniso depending on how the lod levels are calculated. Its possible to get near 8x aniso from the sampling pattern as when projecting the pixel onto an angled surface each sample occupies a differnent point lengthways through the pixel and while this isnt as perfect as the straight line of samples that aniso uses lod levels for 6 or 7 x aniso could be used on sloped surfaces.
