Slashdot Mirror
NVIDIA's Lead Scientist Interviewed
rtt writes "bit-tech.net has up an interview with NVIDIA's chief scientist, David Kirk, about the PlayStation 3, next-generation architectures and what to expect in PC gaming. From the article: 'We're going to see the next generation of shader-based games. At the first generation, we saw people using a shader to emulate the hardware pipeline, and finding "Hey - this really is programmable". After that, they tried to do a few things with more lights, using perhaps eight instead of ten. Then they started to write material shaders, and they made great cloth and metal effects that we saw. People are now starting to change the lighting model, and are exploring the things that they can do with that.'"
The real life picture is, however, exactly oposite. On XBox you will have to re-design your game to use 3 threads(!) (not 2, not 4) to get predictable fluid parallel performance. This is *very* difficult to do (debugging nightmare). Game (and other) developers are very much used to single thread. Sony came up with better idea: Cell chip has parallel vector units that will be used by low-level libs (well tested and stable). Libs will be both provided by Sony and later by engine companies themself. Game programmer is simply writing single-threaded app, as he always did, but using these libs as much as possible (even OpenGL libs will use them). Your app is under the hood running 1-8 way parallel, depending on how much you use those libs. Isn't that better idea then 3-thread SMP approach? For me as a developer yes!
839*929
I don't think he's saying that graphics don't matter... just that they might not be the biggest part of what makes a game fun to play.
In other words, Need For Speed: Dumb Ricer Edition is going to be lame no matter how pretty they make it.
Actually, they already are considering it. The 7800GTX has 50% more transistors than the 6800 Ultra, but runs cooler.
_ speed/
Basically they're shutting off portions of the chip when not in use to cut down on power consumption.
This is mentioned briefly at http://www.bit-tech.net/news/2005/07/07/g70_clock
and also at http://www.hardocp.com/article.html?art=Nzg0LDI=
One example of a reasonably priced, fanless GPU is the FX5200, which can be had at electronic stores for $50-$70. The plain FX5200 is passively cooled, and most manufacturers include only one video output on it. The slightly faster FX5200 Ultra requires a fan for the increased heat, and would probably include two video outputs. The ones I've seen with two outputs had one VGA and one DVI. Surely someone is producing one with two DVIs.
I just purchased a FX6600GT for $165. For its performance, I'd call that reasonably priced, and it includes two DVI outputs, but has a fan.
I'd prefer to see video cards with passive heat sinks too, but the silicon process just isn't there yet. It is getting closer, however.
Like http://www.commell.com.tw/Product/SBC/LV-672.HTM which is Mini-ITX form factor, if you're also into space-saving designs.
-Charles
Learning HOW to think is more important than learning WHAT to think.
I wouldn't say its a niche, since its still in very wide use. Research, special effects production, etc. use it primarily.
The issue is that the game industry is incredibly massive now, so much that its the primary driver in developing new graphics cards. And of course, since DirectX is used by the overwhelming majority of computer games, of course is going to be focused on by nVidia. Gotta sell your product.
PlayStation 3's development environment is based on OpenGL. That alone makes it hardly a "niche" API. I believe the GameCube also uses OpenGL and Revolution probably will too.
main(c,r){for(r=32;r;) printf(++c>31?c=!r--,"\n":c<r?" ":~c&r?" `":" #");}
I know several mathematicians that would disagree with your assessment that coming up with new algorithms is engineering. Algorithm development is typically in the domain of scientists working on better computational tools.
If astrophysicists, computational fluid dynamicists, and/or computational plasma physicists (all of them are scientists, by definition), all got together with game developers and swapped algorithm ideas, I guarantee that there would be improvements seen in games. Those scientists are constantly developing new techniques and algorithms to refine their computations.
An engineer would come along, read the papers those scientists published, gain an understanding of the algorithms and techniques discussed, then go implement them in some code.
Not to diss any engineers out there (I am one), but that's basically what engineers do: take the work of others and implement it, usually in a practical manner. I've worked the other side of the situation as well, working in CFD and controls (amongst other computational things).
You think GLSL is more risky to use than HLSL? Bullshit. It's not really that fundamentally different. Neither is Cg. It's like comparing C and Pascal. In fact, NVIDA's shader compiler is the same for all three languages. It's abstracted into a backend and a set of frontends for each language: Cg, GLSL, and HLSL. So, for NVIDIA hardware all three basically perform identically.
It shouldn't mean more programmers at all. When he is talking about lights he is talking about implementing them with shaders, which an artist can conceptualize and create. Then its essentially drag and drop into the game and it just works. No more programmers required.
10: SIN 20: GOTO HELL