Powerwall with 250,000,000 pixels [UPDATED]

Print Friendly, PDF & Email

That’s .25 billion pixels people. We’ve got a (smaller) powerwall at my work, and it gets pretty hot. I imagine you’ll have to wear a heat suit to use this thing.

Here is the lede

Developed by scientists and engineers in the NASA Advanced Supercomputing (NAS) Division at Ames, the 128-screen hyperwall-2, capable of rendering one quarter billion pixel graphics, is the world’s highest resolution scientific visualization and data exploration environment.

And some stats

Designed and developed by the NAS visualization team in partnership with Colfax International, Sunnyvale, Calif., the system is powered by 128 graphics processing units and 1,024 processor cores, with 74 teraflops (one teraflop equals one trillion floating point operations per second) of peak processing power and a data storage capacity of 475 terabytes (one terabyte equals one trillion bytes).

HPCwire is carrying the full news release.

[Updated: pics of the thing here.] 


  1. […] that NASA, in a fit of showboating, has built a really big power wall: 250M pixels in a gigantic tiled display that, as it turns out, they often don’t use as a […]


  1. Sure it’s cool, but is it useful? How are 250M pixels “easier for the mind to grasp”?

  2. Ahhhh….the $10M question. The technical argument for more pixels in general is that more pixels allow more fidelity to the original data – no smearing small features. This argument holds, up to a point. 250M exceeds the bandwidth of the human I (as far as I remember, 10 Mbits per second). At some point it makes sense to spend the money to stop spending money on brute force and start investing in algorithms that help researchers find features.

  3. John Leidel says

    Indeed, 250M dots is a bit much for one to take in. Now, I’ve actually stood in front of this thing and it quite impressive. In speaking with the NAS head of visualization, Chris Henze, I’ve learned they have quite a few proprietary apps that power this thing. Apparently, they will rarely ever run visualizations across the entire screen. Information overload. Instead, they can run any number of smaller, high resolution scenarios on subsets of the screens simultaneously.

  4. OK, so you’re saying it’s impressive, but its full capability will rarely be used. How much taxpayer funding was that again? 😉