Nvidia GPUs To Be Repurposed as Supercomputers, Number Crunchers

NVidia's branching out. The chipmaker, best known for its partners' video cards and a sexy fairy, is to repurpose its GPUs for use in high-performance computing. Financial modeling, petrochemical exploration and medical research are some of the applications touted as targets.

Codenamed "NVIDIA Tesla," the project aims to make traditional supercomputers a thing of the past; most of us are already aware, thanks to projects like folding@home, that you can grid up with consumer junk to create number-crunching systems. Now they want to deliver a finishing blow to the expense of single, massive monster PCs. Here's how the fit and finish will work out:

The Tesla family of GPU computing solutions spans PCs to large scale server clusters.
The new family includes:

• NVIDIA Tesla GPU Computing Processor, a dedicated computing board that scales to multiple Tesla GPUs inside a single PC or workstation. The Tesla GPU features 128 parallel processors, and delivers up to 518 gigaflops of parallel computation. The GPU Computing processor can be used in existing systems partnered with high-performance CPUs.

• NVIDIA Tesla Deskside Supercomputer, a scalable computing system that includes 2
NVIDIA Tesla GPUs per system and attaches to a PC or workstation through an industrystandard
PCI-Express connection. With multiple deskside systems, a standard PC or workstation is transformed into a personal supercomputer, delivering up to 8 teraflops of compute power to the desktop.

• NVIDIA Tesla GPU Computing Server, a 1U server housing up to eight NVIDIA Tesla
GPUs, contains more than 1000 parallel processors, to add teraflops of parallel processing to clusters. The Tesla GPU Server is the first server system of its kind to bring GPU Computing to the datacenter.

GPUs are interesting because such chips are tailored to perform the specific calculations useful to rendering visual scenes. Such chips are the first mainstream commodities, for example, to casually break the teraflop barrier. Naturally, there are many other applications that benefit from this approach.

Check out our earlier coverage here. AMD/ATI is already lining itself up on similar lines. Here's the full press release, in PDF format due to footnotery.