Today, scientific research is carried out on supercomputing clusters, a shared resource that consumes hundreds of kilowatts of power and costs millions of dollars to build and maintain. As a result, researchers must fight for time on these resources, slowing their work and delaying results. NVIDIA and its worldwide partners today announced the availability of the GPU-based Tesla™ Personal Supercomputer, which delivers the equivalent computing power of a cluster, at 1/100th of the price and in a form factor of a standard desktop workstation.
- Enjoy this article? Help vote it up the 'Vine.
- Public Discussion (2)
This is certainly good for cash-strapped research labs, but I'm not sure the general public has any demand for the processing capacity of a personal supercomputer. Perhaps it will follow.
- 4 votes
PCs already have a unit similar to this except with only four vector registers, as opposed to the hundreds on the Tesla GPU.
I myself am a programmer and have known about these projects for quite some time - if you've ever spent several hours hacking away at a real-time raytracer you'll understand why. Larrabee is Intel's equivalent project.
Because I know a considerable amount, I will divulge a little about it. The Tesla GPU and Intel's Larrabee (soon in the coming I should suppose) will be able to perform a ridiculous amount of arithmic operations simultaniously. This is because they carry hundreds of vector registers. In each of these 'vector' registers four floating point numbers can added, subtracted, multiplied, divided, shuffled, etc with a single instruction. The technical term for this is SIMD - single instruction, multi-data. Anyway, the power and genius is in the SIMD and the sheer number of registers. We're talking thousands of floating point numbers operated on within a fraction of a second so small that human beings don't even experience it. Totaly crazy.
I would recommend anybody interested in this post to look up real-time raytracing. Raytracing is how Pixar makes its videos - that's why they look so damned real. However raytracing a scene composed of thousands of triangles with the added costs of computing lighting effects takes a ridiculous amount of time to compute (It takes several months for Pixar's raytracing technology to make a single movie) and is not currently a candidate for real-time applications - I'm talking video games. Even though there have been some significant advancements in the field lately on the software side they have not been enough to increase framerates to a reasonable amount (~60 fps for a typical videogame). However, with these new hardware advancements it will be more than possible. But don't think video games will be the only things to benefit from this.
As a side not, one should also mention the tension between NVIDIA and Intel. There are several differences between Intel's new approch to graphics and NVIDIA's. Intel is putting their money in raytracing while NVIDIA is sticking with rasterization. Intel has never been a huge competitor in the graphics industry and hopes that raytracing will take over because they will never be able to compete with NVIDIA who has spent years researching and optimizing the rasterization approch. It's a big gamble for Intel.
There are also several differences between the Larrabee GPU and the Tesla themselves. Because Intel has a tremendous amount of experience making CPU's, the Larrabee GPU will consist of hundreds of programmable CPU-like cores, all working in parallel - the key word is programmable.
Well that's a little and its time for supper. Catchya later :)
- 3 votes
You're in Easy Mode. If you prefer, you can use XHTML Mode instead. |



