If you're looking for an excuse to buy a top-end graphics card or four for your gaming rig, you'll want to check out the latest happenings at the University of Antwerp.
It seems that the techies over in Belgium have plonked
four GeForce 9800GX2 graphics cards in a quad-SLI motherboard – plus the requisite glowing blue fans – to create an eight-GPU desktop supercomputer, according to an article on The Tech Report
. The pint-sized processing powerhouse – which came in at a svelte €4,000 – has the same performance of approximately three hundred and fifty desktop PCs based on 'normal' microprocessors.
Sadly, the Nvidia-based supercomputer won't be used for cranking up the frames-per-second in the latest first-person shooters: ASTRA, the group behind the system, plans to use it to process tomographic data from medical scanners to rapidly produce three-dimensional models of patients' internal bits and bobs. Whereas traditional PCs at the €4,000 mark would take a week to crunch through the data, the power of the GPU on the Nvidia cards is such that the same data can be processed in a matter of hours.
Not happy with just having a single octo-core GPU-based supercomputer, the University team is planning an entire cluster
of these machines – a project looking to create a system of such power it will be able to render a 3D tomography model in real-time.
The system, built by a team from the Belgium computer shop Tones.be
was made using entirely off-the-shelf commercial parts – meaning repeating the stunt to create an ultra-powerful cluster that can also be used as a kick-ass LAN for deathmatches will be well within the reach of the ASTRA group.
If you want to see more of the beast, a gallery is available on the ASTRA website
Tempted to see if you can utilise the CUDA toolkit to create your own ultra-fast GPU-based processing system, or do you just use your graphics card for gaming? Share your thoughts over in the forums