NVIDIA's GeForce 8800 (G80): GPUs Re-architected for DirectX 10
by Anand Lal Shimpi & Derek Wilson on November 8, 2006 6:01 PM EST- Posted in
- GPUs
All GPUs are Created Equal: Say Goodbye to Cap Bits
DX9 allows quite a bit of flexibility in implementation. ATI and NVIDIA are free to do things a little differently as they see fit. In order for software to understand how fully the hardware supports the required and optional features of DX9, the hardware has specific capability bits set that describe its features. Microsoft has eliminated this feature from DX10. Software written for DX10 will not have to worry about checking cap bits for DX10 hardware. This is due to the fact that Microsoft has been much more specific about the features required to support DX10. There will still be differences in implementation, optimizations, performance characteristics and the like, but all DX10 hardware will have the same basic feature set to draw from. On the down side, hardware vendors who want to add custom features will have to rely on OpenGL (which allows custom vendor specific extensions to the API).
This will make things much easier for game developers, as they won't have to worry about not having a specific feature around to use for an effect or rendering technique. This is also another step in the direction of eliminating the need for multiple GPU specific rendering paths. We can't say that developers won't write different code for different hardware, because we don't know anything about the differences in performance characteristics at this point. We do know from past experience (with NV30) that even something as simple as the order in which code is executed can make a significant difference in performance. We would like to think that issues like this won't present themselves, but we'll have to wait and see when more hardware and software comes along.
In order to avoid programming issues like the initial NV30 + SM2.0 problems, Microsoft will only allow HLSL (High Level Shader Language) to be used with DX10. This means no low level shader ASM optimization, but it also means that each graphics hardware maker will have full control over how shaders get compiled. There is certainly a trade off here, but this should help keep developers from inadvertently doing something that severely hampers performance on any given architecture.
If DirectX 10 sounds like a great boon to software developers, the fact that DX10 will only be supported in Windows Vista is certain to curb enthusiasm. Other than Vista-only games, all developers will still be required to support DX9 in order to keep the installed Windows XP user base as part of their target market. Some developers have actually made comments to the effect that DX10 is more of a headache than a help right now, and that won't change until they are able to abandon support of older hardware. Hopefully, the DX10 performance and feature benefits will be enough to encourage people to upgrade sooner rather than later, but if the past is any indication it could be several years before DX9 is abandoned by the majority of users and developers.
Unified Shaders
Unified shaders aren't actually a feature as much as a result of DX10. This is a small point that seems to get lost in the shuffle, but Microsoft doesn't require a specific implementation for DX10 compliance: they simply made a better implementation more feasible. Until now, building a GPU with unified shaders would not been have desirable, let alone practical, but Shader Model 4.0 lends itself well to this approach.
We haven't seen unified shaders yet because we didn't need or want them. Up to SM2.0, vertex shaders had a higher precision requirement than pixel shaders. While 32bit floating point was required for compliance at the vertex level, 24bit was all that was needed for full precision in pixel shaders. Partial precision hints were added to accommodate 16bit pixel shaders on NVIDIA hardware. It wouldn't have been practical at the launch of DX9 to require that all shader units be 32bit. The same goes for including pixel oriented features in the vertex shader hardware: the API didn't support it, so there was no need to include it. The R300 GPU is 218mm^2 with only 107 Million transistors, and adding any more complexity than necessary would have certainly produced a much larger chip than they would have been able to handle on the 150nm process employed at the time. These days, we are able to do much more in the same space: ATI's latest chip, the RV570, is about 230mm^2 and has 330 Million transistors.
It is much cheaper, easier, and more efficient to build hardware to fit exactly what is required of each step in the rendering pipeline. This is as true with older hardware as it is with G80. Now that DX10 calls for full 32bit in each shader and nearly the same functionality for both vertex and pixel shader units, it doesn't make sense to duplicate and segregate the hardware. Now that functionality can't be excluded from either vertex or pixel processing, hardware designers are optimizing their parts to make the most efficient use of space. It just so happens that the best way to do this and meet the requirements of DX10 is with unified shaders.
Facebook
Twitter
RSS
111 Comments
View All Comments
Nightmare225 - Sunday, November 26, 2006 - link
Are the FPS posted in this article, Minimum FPS, Average FPS, or Maximum? Thanks!multiblitz - Monday, November 20, 2006 - link
I enjoyed your reviews always a lot as they inclueded the video-capbilities for a HTPC on previous cards. Unfortunately this was this time not the case. Hopefully there will be a 2. Part covering this as well ? If so, it would be nice to make a compariosn on picture quality as well against the filters of ffdshow, as nvidia is now as well supporting postprocessing filters...DerekWilson - Tuesday, November 21, 2006 - link
What we know right now is that 8800 gets a 128 out of 130 on HQV tests.We haven't quite put together an HTPC look at 8800, but this is a possibility for the future.
epsil0n - Sunday, November 19, 2006 - link
I am not agree with this:"It isn't surprising to see that NVIDIA's implementation of a unified shader is based on taking a pixel shader quad pipeline, and breaking up the vector units into 4 scalar units. Now, rather than 4 pixel quads, we see 16 SPs per "quad" or block of stream processors. Each block of 16 SPs shares 4 texture address units, 8 texture filter units, and an L1 cache."
If i understood well this sentence tells that given 4 pixels the numbers of SPs involved in the computation are 16. Then, this assumes that each component of the pixel shader is computed horizontally over 16 SP (4pixel x 4rgba = 16SP). But, are you sure??
I didn't found others articles over the web that speculate about this. Reading others articles the main idea that i realized is that a shader is computed by one and only one SP. Each vector instruction (inside the shader) is "mapped" as a sequence of scalar operations (a dot product beetwen two vectors is mapped as 4 MUD/ADD operations). As a consequence, in this scenario 4 pixels are computed only by 4 SPs.
DerekWilson - Tuesday, November 21, 2006 - link
Honestly, NVIDIA wouldn't give us this level of detail. We certainly pressed them about how vertices and pixels map to SPs, but the answer we got was always something about how dynamic the hardware is able to dynamically schedule the SPs optimally according to what needs to be done.They can get away with being obscure about how they actually process the data because it could happen either way and provide the same effect to the developer and gamer alike.
Scheduling the simultaneous processing one vec4 MAD operation on 4 quads (16 pixels) over 4 groups of 4 SPs will take 4 clock cycles (in terms of throughput). Processing the same 16 pixels on 16 SPs will also take 4 clock cycles.
But there are reasons to believe that things happen the way we described. Loading components of 16 different "threads" (verts, pixels or whatever) would likely be harder on the cache than loading all 4 components of 4 different threads. We could see them schedule multiple ops from 4 threads to fill up each block of shaders -- like computing 4 consecutive scalar operations for 4 threads on 16 SPs.
At the same time, it might be easier to maximize SP utilization if 16 threads were processed on one block of SPs every clock.
I think the answer to this question is that NVIDIA knows, they didn't tell us, and all we can do is give it our best guess.
xtknight - Thursday, November 16, 2006 - link
This has been AT's best article in awhile. Tons of great, concise info.I have a question about the gamma corrected AA. This would be detrimental if you've already calibrated your display, correct (assuming the game heeds to the calibration)? Do you know what gamma correction factor the cards use for 'gamma corrected AA'?
DerekWilson - Monday, November 20, 2006 - link
I don't know if they dynamically adjust gamma correction based on monitor (that would be nice though) ...if they don't they likely adjusted for a gamma of either (or between) 2.2 or 2.5.
Also, thanks :-) There was a lot more we wanted to pack in, but I'm glad to see that we did a good job with what we were able to include.
Thanks,
Derek Wilson
bjacobson - Sunday, November 12, 2006 - link
This comment is unrelated, but could you implement some system where after rating a comment, on reload the page goes back to the comment I was just at? Otherwise I rate something halfway down and then have to spend several seconds finding where I just was. Just a little nuissance.Thanks for the great article, fun read.
neo229 - Friday, November 10, 2006 - link
This is a very suspect quote. A card that requires two PCIe power connectors is going to dissipate a lot of heat. More heat means there must be a faster, louder fan or more substantial and costly heat sink. The extra costs associated with providing a truly quiet card mean that the bulk of manufacturers go with the loud fan option.
DerekWilson - Friday, November 10, 2006 - link
If manufacturers go with the NVIDIA reference design, then we will see a nice large heatsink with a huge quiet fan.Really, it does move a lot of air without making a lot of noise ... Are there any devices we can get to measure the airflow of a cooling solution?
We are also seeing some designs using water cooling and theres even one with a thermo-electric (peltier) cooler on it. Manufacturers are going to great lengths to keep this thing running cool without generating much noise.
None of the 8 retail cards we are testing right now generate nearly the noise of the X1950 XTX ... We are working on a retail roundup right now, and we'll absolutely have noise numbers for all of these cards at load.