The Week in Vis

Randall Hand from VizWorld.com, the web’s best site dedicated to computer graphics and scientific visualization, recap’s the week’s best stories related to supercomputing in the visualization and graphics industries. This week it’s a little bit of science, a little bit of flash, and a whole lot of CUDA.

Quick VizPics

Georgia Tech using GPGPU for Military Defense

Researchers at the Georgia Tech Research Institute (GTRI) are looking at new applications for NVidia’s CUDA libraries and GPU’s for improving performance of common military applications.

“As radar systems and other sensor systems get more complicated, the computational requirements are becoming a bottleneck,” said GTRI senior research engineer Daniel Campbell. “We are capitalizing on the ability of GPUs to process radar, infrared sensor and video data faster than a typical computer and at a much lower cost and power than a computing cluster.”

They’re primarily using the Vector, Signal and Image Processing Library (VSIPL) to run GPU’s, and seeing improvements of 20x to 350x after porting them to GPU.

The Quest For Fire in the Next Harry Potter

It’s no secret that rendering Fire is one of the trickiest possible materials in all of computer graphics. Wired sits down with ILM’s Tim Alexander to talk about how they used some classic supercomputing CFD for the upcoming Half Blood Prince.

Once Half-Blood Prince director David Yates signed off on a visual template, ILM deployed supercharged Linux machines, each loaded with 16 processors and 4 gigs of RAM. “We emulated all these fire parameters: heat ripples, smoke, buoyancy, viscosity, opacity, and brightness,” Alexander says. Processing the massive particle simulations for the 100- by 300-foot firewall was burning up days of data crunching for each frame. So computer graphics artist Chris Horvath spent eight months obsessing over a faster way to conjure impressive flames. “Chris figured out that a lower-resolution particle set still had a fluidy flow,” Alexander says. “The effect looks as if you sprayed propane and then lit it.”

SGI Legal Bankruptcy Recap

It’s been a while since we’ve talked about the events surrounding the SGI Legal Case. What, you thought it was over? Far from it, actually. There was a hearing this week (June 24th) at 10AM to address some more issues, big items on the docket included objections raised by Intel and Oracle about transfer of contracts. Also, a new set of Monthly Operating Reports have been published in dockets 480-493 showing that the CEO’s pulled in a nice $190k this month, down from last month’s $223k. It also shows that SGI operated at a net loss of $15Million this month, not surprising for a company that’s really nothing more than a dying husk at this point as the actual business is being done in the new SGI company.

It’s worth nothing that Eliot Bernstein (iViewIt) and William Kuntz III are still working the system with several objections and legal motions surrounding the event. Check back later for an update of Eliot Bernstein’s objections, direct from the man himself.

More Information on the SuperMicro 1U GPU/CPU System

SuperMicro is showing its new hybrid SuperServer systems at the International Supercomputing Conference in Hamburg, Germany this week. The Register managed to get some details on the SuperServer 6016GT-TF-TM2, the previously announced half-Tesla/Xeon system that offers 2TeraFlops in a 1U rack.

The motherboard used in this machine is a modified version of the normal Twin board using Intel’s “Nehalem EP” processors and “Tylersburg” chipset since that regular Twin mobo only has one x16 slot for graphics cards or GPU accelerators like the Tesla M1060. When you are a motherboard maker with server aspirations, as Super Micro is, you can crank out a new motherboard pretty much at the snap of the fingers to suit your needs. This particular mobo supports 96 GB of DDR3 main memory and has room for three 3.5-inch disks, an integrated Matrox G200eW graphics card, and two Gigabit Ethernet ports.

The Register also brings up some interesting information such as the fact that two of these new hybrid 1U superservers are almost identical to a combination of their standard 2U Server box and a 2U Tesla. You can get the full lineup of SuperMicro’s GPU-enabled systems here.

The Massive Disk Arrays behind Transformers 2

Scott Farrar at ILM is the guy in charge of the special effects of next weekend’s Transformers 2, and he sits down with the San Francisco Examiner to talk about some of the VFX shots:

While the first “Transformers” spawned 14 robots, there are 46 this go around; all were created at ILM.

If all of the effects’ texture maps were printed on one-square-yard sheets, they would cover 13 football fields.

The first film took up 20 terabytes of disk space; this film takes up 145 terabytes (seven times larger). It would take 35,000 DVDs to store 145 terabytes; stacked one on top of the other without storage cases, they would be 145 feet tall.

ELSA, Teradici, and LucidLogic build a new Remote Display System

At the upcoming 17th Industrial Virtual Reality Expo at Tokyo Big-Sight (June 24-26), ELSA will be demonstrating a new External PCI-Express Remote 3D Graphic Solution. The demo uses the ELSA Vridge X100 Quad 8 PCI-Express extension system powered by Lucid’s Hydra 100 to enable high-quality 3D graphics and HD video to be streamed remotely through an IP network using Teradici’s PC-over-IP technology. In a single Vridge system you can install two NVidia Quadro fX5800’s and 2 ELSA VIXEL H200 remote host cards. Ironically, the demo will be in the SGI Japan Booth (23-14).

The goal is to create super-powered workstations that can be centrally located in single datacenter, with displays remotely provided over LAN or high-latency WAN links to users.

Google’s new massive Image Landmark Recognition

Google has released a new paper that they presented at the recent Computer Vision and Pattern Recognition (CVPR) conference in Miami, FL. Called “Tour the World: building a web-scale landmark recognition engine”, it’s about a new tool they’ve built for image pattern matching to identify images of landmarks.

Our research builds on the vast number of images on the web the ability to search those images and advances in object recognition and clustering techniques. First we generated a list of landmarks relying on two sources 40 million GPS-tagged photos from Picasa and Panoramio and online tour guide webpages. Next we found candidate images for each landmark using these sources and Google Image Search which we then “pruned” using efficient image matching and unsupervised clustering techniques. Finally we developed a highly efficient indexing system for fast image recognition.

They’re adamant that this is not a new google product, just some research, but it’s not hard to imagine this being rolled into something like Picasa in the near future.

The ScienceSim Virtual World at SC09

A beta project named “ScienceSim”, aimed at building 3D technologies for visualizing 3D information on the web, will be operating a virtual world for Supercomputing09. Based on the open source package OpenSim, it’s effectively a Second-Life world for discussion of topics related to SC09.

Full instructions for connecting are available on their website.