Slashdot Mirror
Nvidia Discloses Details On Next-Gen Fermi GPU
EconolineCrush writes "The Tech Report has published the first details describing the architecture behind Nvidia's upcoming Fermi GPU. More than just a graphics processor, Fermi incorporates many enhancements targeted specifically at general-purpose computing, such as better support for double-precision math, improved internal scheduling and switching, and more robust tools for developers. Plus, you know, more cores. Some questions about the chip remain unanswered, but it's not expected to arrive until later this year or early next."
To the best of my knowledge, double-precision floating point operations are actually pretty important for some scientific applications of GPUs, and as such this is significant for those using GPUs as supercomputers.
Notice the features being marketed: concurrent CUDA kernels, high performance IEEE double-precision floating point performance, multi-level caching and expanded shared memory, high performance atomic global memory operations. NVIDIA doesn't care about you anymore. Excepting a small hardcore, gamers are either playing graphically trivial MMOs (*cough*WoW*cough*) or have moved to consoles.
They won't want to sell you this chip for a hundred bucks, they want to sell it to the HPC world for a couple thousand bucks (or more... some of NVIDIA's current Tesla products are 5 figures). The only gamers they're really interested in these days are on mobile platforms, using Tegra.
I work at a physics lab, and demand for these newer NVIDIA cards are exploding due to general-purpose GPU programming. With a little bit of creativity and experience, many computational problems can be parallelized, and then run on the multiple GPU cores with fantastic speedup. In our case, we got a simulation from 2s/frame to 12ms/frame. It's not trivial though, and the guy in our group who got good at it... he found himself on 7 different projects simultaneously as everyone was craving this technology. He eventually left b/c of the stress. Now everyone and their mother either wants to learn how to do GPGPU, or recruit someone who does. This is why I bought NVIDIA stock (and they have doubled since I bought it).
But this technology isn't straightforward. Someone asked why not replace your CPU with it? Well for one, GPUs didn't use to be able to do ANY floating or double-precision calculations. You couldn't even program calculations directly -- you had to figure out how to represent your problem as texel- and polygon-operations so that you could trick your GPU into doing non-GPU calculations for you. With each new card released, NVIDIA is making strides to accommodate those who want GPGPU, and for everyone I know those advances couldn't come fast enough.
People have been saying that forever now. I think only the first 2 generations of 3D cards were greeted by universal enthusiasm, while everything else since had a number of "who needs that power to run game X" crowd. The truth is, yes, you can run a lot of games with old cards, but you can run them better with newer cards. So, it's just a matter of preference when it comes to the usual gaming.
AMD/ATI is at least doing something fun with all this new power. Since you can run the latest games in 5400x2000 resolutions with high frame rate, why not hook up three monitors to one Radeon 58xx card and play it like this? That wasn't something you could do with an older card.
Similarly, using some of the new video converter apps that make use of a GPU can cut down transcoding from many hours to one hour or less... you can convert your blu-ray movie to a portable video format much easier and quicker. Again, something you couldn't do with an old card, and something that was only somewhat useful in previous generation.
In summary, I think the *need* for more power is less pressing than it used to, but there's still more and more you can do with new cards.
Never underestimate the bandwidth of a 747 filled with CD-ROMs.
Some motherboards have more than one PCI Express slot. Some even come with GPUs built onto the motherboard. In either case, it is entirely conceivable that there may be a GPU present other than the one attached to the display. Then there's the Hydra 200 (look on Anandtech, I'm too lazy to find the link) - a chipset which evenly distributes processing power among multiple GPUs from any vendor to render a scene or do GPGPU computing.
Nvidia just released new drivers which explicitly disable PhysX acceleration in the presence of a GPU from another manufacturer. For the above stated reasons, this is evil.
Now you have done it. Don't you know enough not to wake the fanboys. Now I am gonna have to hear how Jobs is a genius for not allowing people to do things that he believes unimportant with their hardware. That nobody actually needs any more control than Steve gives them. It's not that I agree with them. I just do not want to hear it. Please just let them have their fanboy dreams in peace. Please!
Why is it so hard to only have politicians for a few years, then have them go away?