SGI debuts acceleration technology
- By Michael Hardy
- Sep 19, 2005
Silicon Graphics Inc. unveiled new technology today that the company claims can exponentially accelerate the speed of applications using SGI's existing Itanium 2-based servers. Called Reconfigurable Application-Specific Computing (RASC), the rackmounted hardware module used field-programmable gate arrays (FPGA), which are semiconductors that users can program for specific applications.
The FPGA essentially becomes a dedicated computation engine for specified routines, lightening the load on the main processors and performing its specified tasks at high efficiency. SGI's technology makes it easier for users to reprogram the FPGAs and also integrates the RASC device into SGI's NUMALink network fabric.
Ron Renwick, SGI's product manager of reconfigurable computing, said speed gains can be more than 100-fold, but even the low-end results multiply speed by more than 42 times.
“Even at the lowest number we have here, there are some huge performance gains to be seen," he said.
FPGA technology has been around for some time, but the challenge has been making the reprogramming feasible, he said. Most programmers are not used to thinking in the hardware terms that FPGA configurations require. Third-party tools make the process easier for programmers, but Renwick said the technology isn't quite where the company envisions it yet.
“The longer run is that, let’s say a couple of years down the road, it’s going to be seamless," he said. "For this to be a standard part of anybody’s portfolio, it’s got to be seamless. Today it’s not."