Radeon 9700 NDA Lifted
- Thread starter SanGreal
- Start date
Heres a key part:
That'll explain the 8 pipelines. If they are working on a .13um part I wonder if they will add that back in.
That'll explain the 8 pipelines. If they are working on a .13um part I wonder if they will add that back in.
Looks like we might get another DX 1.1 vs 1.4 problem here.
On the other hand, we're also moving towards shading languages which might make this a rather small problem.
Edit:
First of all, any scores/benchmarks without aniso and FSAA is imo not even interesting for next gen cards.
And the R9700 absolutely rocks with 4XFSAA and aniso.
(Although they don't mention if the score was with trilinear + aniso or bilinear + aniso)
Typedef Enum
Regular
I must say that I'm quite impressed. I might just have to pick up one of these things when they ship.
Not to take anything away from this chip, but I thought it was rather profound that Anand stated something like, "Believe it or not, NV30 will be faster than R300...etc"
I mean, look @ that performance...To think that another chip is coming out that will crank out even higher performance is just staggering.
Not to take anything away from this chip, but I thought it was rather profound that Anand stated something like, "Believe it or not, NV30 will be faster than R300...etc"
I mean, look @ that performance...To think that another chip is coming out that will crank out even higher performance is just staggering.
All of the current cards have queryable resource limits. DirectX only specifies the minimum neccessary. For example, although DX8 limited the number of constant registers to 96 and program length to 128, I believe the GF3 could actually handle many more registers and program lengths up to 160.
I think what Anand was saying is that currently, the VS2.0 is limited to 1024 instructions and PS2.0 to 96 instructions, the R300 can do more than that (160 pixel shading instructions) and that the NV30 limits may be much higher.
However, 1024 instruction VS and 96 PS are effectively "unlimited" for what most people want to do. While the true hardware may support more, and NVidia might be shipping hardware acceleration for Renderfarms doing Renderman in hardware, for most games, they will stick with the DX9 limits.
The issue with PS1.0,1.1,1.2,1.3,1.4 wasn't instruction limits, or # of registers, it was new and differing instructions and differing ways of using the texture ops. Hopefully, the only real difference between DX9 boards will be in the maximum resource limits.
I think what Anand was saying is that currently, the VS2.0 is limited to 1024 instructions and PS2.0 to 96 instructions, the R300 can do more than that (160 pixel shading instructions) and that the NV30 limits may be much higher.
However, 1024 instruction VS and 96 PS are effectively "unlimited" for what most people want to do. While the true hardware may support more, and NVidia might be shipping hardware acceleration for Renderfarms doing Renderman in hardware, for most games, they will stick with the DX9 limits.
The issue with PS1.0,1.1,1.2,1.3,1.4 wasn't instruction limits, or # of registers, it was new and differing instructions and differing ways of using the texture ops. Hopefully, the only real difference between DX9 boards will be in the maximum resource limits.
Typedef Enum said:
NV30 is (also according Anand) AT LEAST 5 months away, that's almost half a year!!! Don't you think a chip coming two quarters later definitly should be faster? Even think about the improvmenents the memory semiconductor companies achieve in this timeframe (2 ns instead of 3 ns memory ...).
Pretty impressive chip by all accounts. It sounds as exciting to me as way back in the day when I bought my Orchid Righteous 3D then oohed and aahed at the demos that came with it.
I was thinking about picking up a cheap Radeon 9000 but I'm sorely tempted to go for the R300!
One quote in the Anandtech review which I found a bit peculiar was:
If they didn't encounter problems with performance, compatibility or image quality, where are the problems!
I was thinking about picking up a cheap Radeon 9000 but I'm sorely tempted to go for the R300!
One quote in the Anandtech review which I found a bit peculiar was:
If they didn't encounter problems with performance, compatibility or image quality, where are the problems!
You probably won't be able to pick up an R300 in a store for almost 30 days. That places the real (non-paper) launch around August 18th. The NV30's paper launch will probably be in Nov, with availability in Dec, or, they might opt to skip the paper launch and just launch everything in Dec. In any case, the distance between the R300 and NV30 in terms of real delivery is probably more like 3-4 months, but that's a guess.
One thing I'd like to see with these new DX9 cards is tests of *long* complicated vertex/pixel shaders, especially pixel shaders which use lots of texture reads and preferably 64-bit/128-bit FP framebuffers. The FP framebuffers will expose any memory bandwidth problems (if NV30 doesn't have a 256-bit bus, it better be a tiler or eDRAM architecture!) and long pixel shaders with lots of texture lookups will expose any inefficiencies in the pipelines. Perhaps the R300's single texture per pipe design will leave them at a disadvantage? Who knows, but we need better benchmarks than Quake3 and UT2003 to answer these questions for DirectX9.
One thing I'd like to see with these new DX9 cards is tests of *long* complicated vertex/pixel shaders, especially pixel shaders which use lots of texture reads and preferably 64-bit/128-bit FP framebuffers. The FP framebuffers will expose any memory bandwidth problems (if NV30 doesn't have a 256-bit bus, it better be a tiler or eDRAM architecture!) and long pixel shaders with lots of texture lookups will expose any inefficiencies in the pipelines. Perhaps the R300's single texture per pipe design will leave them at a disadvantage? Who knows, but we need better benchmarks than Quake3 and UT2003 to answer these questions for DirectX9.
Hercules UK are saying September for their 9700 boards.
Well, with those kinds of scores (10K 3dmarks w/ AA and AF?) I will hold off popping any corks until I see the final product and its IQ.
The specs and details from the two links look very promising... but we all should know that what benchmark graphs dont show is usually the most important thing.
One thing I did find interesting about Anand's article (well, besides it's attempt to take the wind out of the R300's sails at every available opportunity to taint the reader towards the NV30..) was the blurb about SV-II:
This truly sounds like a SmoothVision-II, with absolutely no idea what will really ship like the 8500. hehe. We heard the same kinds of things on pre-release R200, and what we got was first OGSS, then later *mostly* OGSS with slight jittering when circumstances were perfect (i.e. no fog, title specifity, etc.etc.).
I am a little bummed on the onset with Toms eluding to the card being only MSAA as having SV now is a godsend for sims, RPGs and other games that have those aweful aliased textures that just seem to keep popping up in even new games. So far the only commentary on SV-II quality is using the MS utopian spoke demo... would love to hear some commentary on a real 3D scene (like one with textures and everything). It will be critical for their "quality" AF to eleviate texture aliasing if this is to be a targetted "gamers" card IMO.. and even moreso critical if SV-II wont have supersampling to come to the rescue like it does today.
The specs and details from the two links look very promising... but we all should know that what benchmark graphs dont show is usually the most important thing.
One thing I did find interesting about Anand's article (well, besides it's attempt to take the wind out of the R300's sails at every available opportunity to taint the reader towards the NV30..) was the blurb about SV-II:
This truly sounds like a SmoothVision-II, with absolutely no idea what will really ship like the 8500. hehe. We heard the same kinds of things on pre-release R200, and what we got was first OGSS, then later *mostly* OGSS with slight jittering when circumstances were perfect (i.e. no fog, title specifity, etc.etc.).
I am a little bummed on the onset with Toms eluding to the card being only MSAA as having SV now is a godsend for sims, RPGs and other games that have those aweful aliased textures that just seem to keep popping up in even new games. So far the only commentary on SV-II quality is using the MS utopian spoke demo... would love to hear some commentary on a real 3D scene (like one with textures and everything). It will be critical for their "quality" AF to eleviate texture aliasing if this is to be a targetted "gamers" card IMO.. and even moreso critical if SV-II wont have supersampling to come to the rescue like it does today.
Impressive launch indeed. I like it. Aniso sounds fine which removes the last beef I had with Radeon in the past, so yeah--as soon as the .13 version is out, that's definetely something to look into for my home system. Until then, the driver question should have been answered as well, even though I concur with the Mariner and am wondering, too, what's the "problem" with the current set as presented by Anand.
On a side note, I really liked the bits about RenderMonkey. True RenderMan and 3dsmax support on a graphics board is VERY nice.
ta,
-Sascha.rb
On a side note, I really liked the bits about RenderMonkey. True RenderMan and 3dsmax support on a graphics board is VERY nice.
ta,
-Sascha.rb
Similar threads
