AMD ponders maths boost to quad-core Opterons

Talking to Clearspeed

AMD is in talks with UK co-processor company Clearspeed in a bid to boost the performance of future multi-core x86 CPUs, it has emerged. It's not hard to see why. Clearspeed's current CSX600 design can provide a huge floating point boost.

Clearspeed's co-processors accelerate compute-intensive maths library operations in much the same way GPUs accelerate graphics operations. They not only bring their own processing horsepower to the table, but free the CPU up for other tasks. Time was when maths co-processors were commonplace add-ons to integer maths-focused CPUs.

The CSX600, launched in June 2005, contains an multi-threaded array of 96 processor cores each with 6KB of memory and sharing 128KB of scratchpad RAM. Clearspeed says the part consumes no more than 10W yet delivers 250,000 1024-point complex floating point operations each second. The company targets high-performance computing applications - the kind of area AMD is keen to get Opteron chips into.

Interestingly, it's not just AMD who's taken an interest in Clearspeed. Last week, Intel co-published a paper on using Clearspeed's twin-chip Advance board to boost Intel Math Kernel Library operations significantly.

Clearspeed last year also announced a huge supercomputer win in Japan by teaming with Sun on what should become the country's largest cluster. ®

Sponsored: Designing and building an open ITOA architecture