Black and White.

Tuesday, August 07, 2007

Nvidia launches Quadro VCS GPU server

A DAY AFTER DAAMIT revealed its FireGL line-up of professional graphics cards, Nvidia came up with a press release stating that the company is launching the QuadroPlex Visual Computing System, or VCS in short.

The model goes by the name VCS Model S4, which is a short name for a 1U server case featuring four heavily modified Quadro boards creating a very dense GPU array. This product is intended for data-centric apps such as off-line rendering, remote graphics applications, and embedded visual computing. In short, the company is dead set on killing the data centres consisted out of CPUs in movie and TV production industry.

These four modified boards are based on Quadro FX 5600 (has nothing in common with GeForceFX 5600 sans the name), a G80GL GPU powered board, with 1.5 GB of GDDR-3 memory. This monster of a board comes with 71.6 GB/s of bandwidth (memory clock is set at 1.49 GHz), making a combined total of 286.4 GB/s, an astonishing figure by any account.

Overall, the frame buffer reached 6GB, so large textures can be processed with Full-Scene AntiAliasing to extreme details. Nvidia claims that S4 server comes with accelerated 3D texture performance as well, for neat visualization of datasets used in energy and scientific research. Read: we want our piece of GPGPU pie.

Of course, this product is not intended to be used as a single product, but as one of numerous units in something that we could start pronouncing as "Visualization Datacentres" or products for "Visualization server room." Kudos to Nvidia for creating a completely new segment of the market, and once this gets rolling there is no turning back. We figure that by now you have understood why Intel is considering Nvidia to be its biggest threat, right.

This product will be available in Q4 2007, and you can expect a whole lot of Hollywood movies and more importantly, TV series productions to rely on QuadroPlex. Pricing was not announced, but that does not matter anyways - what AMD and Nvidia have done in the past couple of years was total crash of the budgets required to create special FX. From rendering farms built by SGI, we have come at the time when GPU workstations are good enough for full-blown previews, and taking over the rendering segment was just the matter of time.

Sadly, the only thing missing for resounding success are the weird ways how Nvidia treats its partners, regardless of being a software vendor, add-in board partner or member of the press. Nvidia's head of PR stated to us that the company is not launching anything at Siggraph, but relying on what a PR says is equal to believing in the tooth fairy

----------------
Listening to: John Powell - Extreme Ways (Bourne's Ultimatum, by Moby)
via FoxyTunes

No comments: