Dr Wim Vanderbauwhede, of the University of Glasgow, and colleagues at the University of Massachusetts Lowell have created a processor that has 1000 execution units or cores. This would be the fastest processor as of now. Intel also has announced its plans to create a thousand core monster soon.
In a true MIMD fashion, they have sliced up a large FPGA into execution units which can handle thousand different instruction streams. Dr Vanderbauwhede justifies the use of FPGAs thus; "FPGAs are not used within standard computers because they are fairly difficult to program but their processing power is huge while their energy consumption is very small because they are so much quicker - so they are also a greener option."
As a benchmark, the chip was used to run a MPEG movie algorithm and it achieved a throughput of 5 Gigabytes per second. That would be at least 20x faster than current desktop performance. the team is hoping to present the findings at the International Symposium on Applied Reconfigurable Computing in March this year. What would be interesting to find out would be how do they make it all work. Memory contention, system bus bandwidth, communication overheads, cache coherence etc. are some relevant problems.
Subscribe to:
Post Comments (Atom)
Posts
No comments:
Post a Comment