Nvidia brings supercomputing to desktops

Set to ship in August, Nvidia's GPU Processor and Server packages are bringing supercomputing power to the desktop and datacenters. The scalable systems are ideal for research and analysis projects, especially in the biosciences and physical/geophysical domains.

The Tesla family of products includes:

  • Nvidia Tesla C870 GPU Computing Processor ($1,499): A dedicated board that scales up to 128 parallel processors and delivers up to 518 gigaflops of parallel computation
  • Nvidia Tesla D870 Deskside Supercomputer ($7,500): A scalable computing system with two Nvidia Tesla GPUs and can be wired to a PC or workstation via an industry-standard PCI-Express connection, delivering up to 8 teraflops of computing power to the desktop
  • Nvidia Tesla S870 GPU Computing Server ($12,000): An 1U server housing up to eight Nvidia Tesla GPUs, i.e. 1000+ parallel processors, and it is the first server system of its kind to bring GPU computing to the datacenter

The details gleaned from the iTnews write-up on high performance number crunchers also details that the software bundled along with the system includes a C compiler, debugger, and libraries for Linux and Windows XP.

More information is available at:

Nvidia touts desktop supercomputer (vnunet)

Nvidia expands into high performance chips (InfoWorld | IDG News Service)

Also, an interview with Dave Kirk, Chief scientist at Nvidia (Beyond3D) regarding Tesla and the future of GPU computing.

As processing goes more multi-core and multi-threaded, GPUs are also going from being niche to multi-purpose. Applications that exploit the full potential of parallel processing systems have still not gone mainstream, since multi-threaded programming is a significant shift from the sequential single-thread models.

Recently, Google purchased PeakStream, a firm that engaged in abstracting the task of running multiple threads to software. Definitely, the potential for parallel processing systems is huge, but are there enough applications out there to take it mainstream and make it more appealing to businesses other than just research firms? Join the discussion.


What are the measures to be adopted to make parallel processing mainstream in businesses as well.


i got my casio 702p 2kb memory on 1983(?) , and saw apple 2e's dancing demons(graphics). i wonder how would flip flopping pixels would bring us to developments that time. now, we have gyroscopes and much more on galaxy and many computer smart phones. we could now map(?) humans and paste it on other places(star trek patent ?). im tinkering the advents of holographic tvs and movies to actually manipulating atoms into physical travels(light travels= brave new baby). sounds like Odyssey2010. then mobile phones again becomes penta chunker,faster than brains.

Editor's Picks