NVidia and HP have teamed up to create a new GPU “Starter” kit aimed at bringing GPU computing to a wider audience without getting into all the fine details of system configuration. However, their new “starter” kit is anything but entry-level.
The system contains eight ProLiant SL390 G7 servers, packed full of 24 M2070 GPUs, 16 CPUs, and its preconfigured with CUDA 4.0. The servers, presumably loaded with quad-cores, offer a respectable 32-cores of additional CPU power in addition to the copious amounts of GPU performance. The M2070 GPU that’s included in the package is a Fermi-based part, with 6GB of RAM per GPU.
With a price tag of $99,000, it’s a bit beyond your average hobbyist. It’s a great steal for companies tho as the total system is offered at about a 50% discount of the total package, plus contains their special GPU development tools.
The Wall Street Journal has an article today on HP showing a new “wall of touch”. This is a standard display wall comprised of six LCDs. Each LCD can be anywhere from 43″ to 46″ and is capable of 1080p resolution. Philip McKinney, Vice President and Chief Technology Officer of H-P’s Personal Systems Group, showed the setup to the Journal.
He said “wall of touch” will not be widely available to consumers until 2011, and would likely come with a hefty price tag: anywhere from a couple thousand dollars up to $100,000 for more advanced systems with technologies like HD video conferencing.
I would expect that the cost would be a couple thousand dollars for a single screen, since 42″ LCDs are hovering around one thousand dollars.
Hess Corp., a global independent energy company, had a problem. Their visualization software and their simulation software ran on separate machines and separate operating systems (Windows and Linux) and moving the massive amounts of data between them was becoming a problem. They turned to Nvidia for a rather unorthodox solution, using Parallels to emulate one OS on top of another. Thanks to NVidia’s new MultiOS GPU Support, it wound up being a perfect solution.
“Instead of running applications on different machines, and moving data back and forth over the network, SLI Multi-OS radically streamlines users’ workflows and instantly enhances productivity,” said Jeff Brown, general manager, NVIDIA Professional Solutions Group. “We believe Hess’ use of fully virtualized workstations will makes their employees more efficient, while lowering the total cost of ownership.”
Scientists at HP have demonstrated a technique for using ordinary, unmodified flatbed scanners to scan an image and then detect and correct defects such as creases.
On the surface, the technique appears relatively simple. Most flatbed scanners use two separate light bulbs to accurately capture all the colour in a photo. By controlling these independently of each other, two slightly different images (each taken from different directions as the bulbs move under the photo) can be captured of the same photograph. From these, rudimentary 3D information can be generated.
With the defect — a crease in our example — identified, software can artificially mask it entirely. Known as ‘infilling’, each pixel inside the scanned crease is replaced by a new one generated from pixels just outside of the crease. The software makes sure the two pixels are similar, to avoid sticking a giant red pixel in the middle of a bunch of green ones.
The main limitation is that it requires a 3D defect, defects that lie perfectly against the glass can’t be detected by the different lights.
HP has just announced a new workstation, the Z800, that is now available with Dual NVidia Tesla units, making it one of the most powerful workstations I know of. From the press release:
In molecular dynamics, AMBER, a public research code with more than 60,000 users, has been written to leverage the massively parallel CUDA architecture to deliver a 50X speed up of simulations. The result is faster scientific insights for researchers. In finance, Numerix and CompatibL have announced CUDA support for a new Counterparty Risk application of their Numerix 7 analytics solution used by over 375 financial institutions. This derivatives pricing application is today experiencing an 18X boost in performance in the calculation of complex pricing models with NVIDIA GPUs.
The Z800 is available for as low at $1800, although that’s without the Tesla’s. I was unable to find a configuration with dual-Tesla’s using the provided configuration tool. If anyone gets a price on one of these, let us know!
HP has just announced a new display technology they’re calling “eSkin”. What is actually is? Well, listen to this comment from an HP spokeswoman:
“HP eSkins is in fact a dynamic digital surface (and not just a static display) that can be controlled to address up to 80 segments to give the perception of movement and eye-catching motion,” an HP spokeswoman said in an email. “The segmented display can be turned off and on to create visual effects.”
Looks like it’s somewhere between eInk and LCD, with baked in images that can be turned opaque/transparent via an electrical impulse. They’re formally presenting the technology later this week at Display Week2009 in San Antonio, just like the new Sharp 5-color Display Technology.
Just a few days remain in Rackable’s 25-day wait for acquisition of SGI, and while no-one has officially made an offer, it seems that a few of them are considering it. John West over at InsideHPC has pulled some strings and gotten the scoop on a few vendors, including one very interesting possibility:
HP: Although HP and IBM dominate the Top500, HP hasn’t had a strong presence in the top 10 for a while. SGI’s R&D in HPC could reinvigorate their offering. HP has a market cap of just about $85B right now, with $10B of cash and equivalents on hand. [Update: Readers have also pointed out that HP is the only other large-scale Itanium vendor in the US. Buying SGI would be a consolidation of technology and talent surrounding the platform.]
An HP-SGI consolidation would be an interesting one. While they both do alot of work with Itanium, Itanium simply hasn’t “taken off” as the CPU of choice in HPC. HP does, however, seem to have the money if it wanted. Their Q1`09 statement shows a $28.8B profit and $100B in assets, along with a $1B Net gain in Cash.
Update: Just got word from a reputable source that the SGI Auction closes today at 5pm Eastern. So we should be hearing something soon one way or another.
If you’ve ever wondered just how much horsepower it takes to make a feature-length animated film, HP has a press release touting their involvement in the new “Monsters vs Aliens”, with some statistics for you:
40million Computing Hours, 8x more than Shrek
Render the nearly 100 terabytes of disk storage.
Render more than 30 sequences in the movie that would have taken more than 1,000 years to render on a single workstation.
Create one of the most technically challenging sequences of the film involving a flyover of a town, including houses, hills and trees. In this shot, the background trees had to be broken up into more than 300 layers to render.
Stage an explosion in one of the battle scenes, which required more than three terabytes of disk space alone.
VizWorld.com We cover visualization and graphics news from around the internet, including Scientific Visualization, Visual Effects, and Graphics Hardware. Read more on our About Page or learn about our Advertising Options Get updates via twitter from @VizWorld.