Applications of GF100’s Compute Hardware
Last but certainly not least are the changes to gaming afforded by the improved compute/shader hardware. NVIDIA believes that by announcing the compute abilities so far ahead of the gaming abilities of the GF100, that potential customers have gotten the wrong idea about NVIDIA’s direction. Certainly they’re increasing their focus on the GPGPU market, but as they’re trying their hardest to point out, most of that compute hardware has a use in gaming too.
Much of this is straightforward: all of the compute hardware is what processes the pixel and vertex shader commands, so the additional CUDA cores in the GF100 give it much more shader power than the GT200. We also have DirectCompute, which can use the compute hardware to quickly do some things that couldn’t be done quickly via shader code, such as Self Shadowing Ambient Occlusion in games like Battleforge, or to take an NVIDIA example, the depth-of-field effect in Metro 2033.
Perhaps the single biggest improvement for gaming that comes from NVIDIA’s changes to the compute hardware are the benefits afforded to compute-like tasks for gaming. PhysX plays a big part here, as along with DirectCompute it’s going to be one of the biggest uses of compute abilities when it comes to gaming.
NVIDIA is heavily promoting the idea that GF100’s concurrent kernels and fast context switching abilities are going to be of significant benefit here. With concurrent kernels, different PhysX simulations can start without waiting for other SMs to complete the previous simulation. With fast context switching, the GPU can switch from rendering to PhysX and back again while wasting less time on the context switch itself. The result is that there’s going to be less overhead in using the compute abilities of GF100 during gaming, be it for PhysX, Bullet Physics, or DirectCompute.
NVIDIA is big on pushing specific examples here in order to entice developers in to using these abilities, and a number of demo programs will be released along with GF100 cards to showcase these abilities. Most interesting among these is a ray tracing demo that NVIDIA is showing off. Ray tracing is something even G80 could do (albeit slowly) but we find this an interesting way for NVIDIA to go since promoting ray tracing puts them in direct competition with Intel, who has been showing off ray tracing demos running on CPUs for years. Ray tracing nullifies NVIDIA’s experience in rasterization, so to promote its use is one of the riskier things they can do in the long-term.
NVIDIA's car ray tracing demo
At any rate, the demo program they are showing off is a hybrid program that showcases the use of both rasterization and ray tracing for rendering a car. As we already know from the original Fermi introduction, GF100 is supposed to be much faster than GT200 at ray tracing, thanks in large part due to the L1 cache architecture of GF100. The demo we saw of a GF100 card next to a GT200 card had the GF100 card performing roughly 3x as well as the GT200 card. This specific demo still runs at less than a frame per second (0.63 on the GF100 card) so it’s by no means true real-time ray tracing, but it’s getting faster all the time. For lower quality ray tracing, certainly this would be doable in real-time.
Dark Void's turbulence in action
NVIDIA is also showing off several other demos of compute for gaming, including a PhysX fluid simulation, the new PhysX APEX turbulence effect on Dark Void, and an AI path finding simulation that we did not have a chance to see. Ultimately PhysX is still NVIDIA’s bigger carrot for consumers, while the rest of this is to entice developers to make use of the compute hardware through whatever means they’d like (PhysX, OpenCL, DirectCompute). Outside of PhysX, heavy use of the GPU compute abilities is still going to be some time off.
Facebook
Twitter
RSS
115 Comments
View All Comments
x86 64 - Sunday, January 31, 2010 - link
If we don't know these basic things then we don't know much.1. Die size
2. What cards will be made from the GF100
3. Clock speeds
4. Power usage (we only know that it’s more than GT200)
5. Pricing
6. Performance
Seems a pretty comprehensive list of important info to me.
nyran125 - Saturday, January 30, 2010 - link
You guys that buy a brand new graphics card every single year are crazy . im still running an 8800 GTS 512mb with no issues in any games whatso ever DX10, was a waste of money and everyones time. Im going to upgrade to the highest end of the GF100;s but thats from a 8800 GTS512mb so the upgrade is significant. Bit form a heigh end ati card to GF 100 ?!?!?!? what was the friggin point in even getting a 200 series card.!?!?!!?1/. Games are only just catching up to the 9000 series now.Olen Ahkcre - Friday, January 22, 2010 - link
I'll wait till they (TSMC) start using 28nm (from planned 40nm) fabrication process on Fermi... drop in size, power consumption and price and rise is clock speed will probably make it worth the wait.It'll be a nice addition to the GTX 295 I currently have. (Yeah, going SLI and PhysX).
Zingam - Wednesday, January 20, 2010 - link
Big deal... Until the next generation of Consoles - no games would take any advantage of these new techs. So? Why bother?zblackrider - Wednesday, January 20, 2010 - link
Why am I flooded with memories of the 20th Anniversary Macintosh?Zool - Wednesday, January 20, 2010 - link
Tesselation is quite resource hog on shaders. If u increase polygons by tenfold (quite easy even with basic levels of tesselation factor) the dissplacement map shaders needs to calculate tenfold more normals which ends in the much more detailed dissplacement of course. The main advatage of tesselation is that it dont need space in video memmory and also read(write ?) bandwith is on chip but it actualy acts as you would increase the polygons in game. Lightning, shadows and other geometry based efects should act as on high polygon models too i think (at least in uniengine heaven u have shadows after tesselation where before u didnt had a single shadow).Only the last stage of tesselator the domain shader produces actual vertices. The real question would be how much does this single(?) domain shader in radeons keep up with the 16 polymorph engines(each with its own tesselation engines) in gt300.
Thats 1(?) domain shader for 32 stream procesors in gt300(and much closer) against 1(?) for 320 5D units in radeon.
If u have too much shader programs that need the new vertices cordinations the radeon could end up being realy botlenecked.
Just my toughs.
Zool - Wednesday, January 20, 2010 - link
Of course ati-s tesselation engine and nvidias tesselation engine can be completly different fixed units. Ati-s tesselation engine is surely more robust than a single tesselation engine in nvidias 16 polymorph engines as its designed for the entire shaders.nubie - Tuesday, January 19, 2010 - link
They have been sitting on the technology since before the release of SLi.In fact SLi didn't have even 2-monitor support until recently, when it should have had 4-monitor support all along.
nVidia clearly didn't want to expend the resources on making the software for it until it was forced, as it now is by AMD heavily advertising their version.
If you look at some of their professional offerings with 4-monitor output it is clear that they have the technology, I am just glad they have acknowledged that it is a desire-able feature.
I certainly hope the mainstream cards get 3-monitor output, it will be nice to drive 3 displays. 3 Projectors is an excellent application, not only for high-def movies filmed in wider than 16:9 formats, but games as well. With projectors you don't get the monitor bezel in the way.
Enthusiast multi-monitor gaming goes back to the Quake II days, glad to see that the mainstream has finally caught up (I am sure the geeks have been pushing for it from inside the companies.)
wwwcd - Tuesday, January 19, 2010 - link
Maybe I'll live to see if Nvidia still wins AMD / Ati, a proposal which is as leadership price / performance, or even as productivity, regardless of price!:)AnnonymousCoward - Tuesday, January 19, 2010 - link
They should make a GT10000, in which the entire 300mm wafer is 1 die. 300B transistors. Unfortunately you have to mount the final thing to the outside of your case, and it runs off a 240V line.