Silicon Graphics, long known for its MIPS processor-based and IRIX operating system imaging systems, Monday announced a line of Itanium-processor, Linux-based visualization products.
SGI's aim with these new systems is to address visualization performance issues that are increasingly becoming a challenge in day-to-day life science research operations.
"The bottleneck used to be turning data into (images) on the screen," says Simon Hayhurst, product line manager, in SGI's Visual Systems Group. "But now, the size of the data is the issue. The challenge is, 'How can I interact with all of my data, all of the time?' " says Hayhurst.
He notes that because there is so much data, researchers often look only at a small sample of what's available. And while this approach provides useful information, it often does not yield the full value that could be derived from all of the available data. For example, he equates many visualization efforts to picking a book or two out of the Library of Congress. With the information garnered from a single book, "you appear smarter, but you are only accessing a small amount of the information," says Hayhurst.
The new Prism systems address the data bottleneck by combining high computational and graphic processing power with the ability to move a lot of data into memory. Specifically, the Linux-based Prism systems use Intel Itanium2 CPUs, SGI's advanced graphics technology, and SGI's NUMAflex shared memory architecture, which allows large datasets to be held in memory.
SGI believes the Prism's ability to visualize larger data sets opens up new areas of research. For example, confocal microscopy is one area where the combination of high computational and graphical processing power and a large memory to hold data could change the way research is conducted.
A confocal microscopy system produces 2D images. One common life science use of the microscope is to use fluorescent probes to examine gene expression under different conditions in tissues. Often, the microscopy systems come with a PC to view the 2D images. And researchers typically look at a few slices and ignore the rest of the data they've collected. With the Prism system, a researcher could use the memory capacity and processing power to take the 2D slices and build them into 3D images in real time.
Similar to SGI's Altix line of higher-performance computing systems, the Prism line of visualization systems is scalable. The product line includes the Power Level system that supports up to 8 CPUs, 4 ATI FireGL graphic processing units (GPUs), and 96 GB of memory; the Team Level that supports up to 16 CPUs, 8 GPUs, and 192 GB of memory; and the Extreme Level that supports up to 512 CPUs, 16 GPUs, and 6.1 TB of memory. The entry-level price for the Prism line is US$30,000.
SGI also announced today that it has tapped Transitive for its QuickTransit product, which allows software applications compiled for one processor and operating system to run on another system without any source code changes. SGI claims that using QuickTransit, researchers currently running applications on other SGI visualization systems can run these applications on the Prism systems.