The Inquirer-Home

Researchers use CPUs to feed GPUs

Claim 20 per cent performance boost
Thu Feb 09 2012, 14:20

RESEARCH in part funded by AMD has found that by using the CPU to pre-fetch data for a GPU performance can be increased by 20 per cent.

Researchers at North Carolina State University (NCSU) claimed to have measured a 20 per cent performance boost by having the CPU fetch data from RAM to feed the GPU in chips that have on-die CPU and GPUs, such as AMD's Fusion processors. The team claims to have set the GPU to do heavy computation while the CPU acted as the controller behind the GPU's raw computing power.

Huiyang Zhou, an associate professor of electrical and computer engineering at NCSU commented that while firms like AMD are putting CPUs and GPUs on the same die, "the CPU cores and GPU cores still work almost exclusively on separate functions. They rarely collaborate to execute any given program, so they aren't as efficient as they could be. That's the issue we're trying to resolve."

Zhou continued, "Our approach is to allow the GPU cores to execute computational functions, and have CPU cores pre-fetch the data the GPUs will need from off-chip main memory [...] This is more efficient because it allows CPUs and GPUs to do what they are good at. GPUs are good at performing computations. CPUs are good at making decisions and flexible data retrieval."

Although AMD and Nvidia GPGPUs provide considerable compute performance, the challenge is feeding them with enough data from main memory. The idea of using a CPU to feed GPU cores wasn't viable until now, according to Zhou, because CPUs and GPUs were not on the same die.

The research will be presented at the 18th International Symposium on High Performance Computer Architecture on 27 February. µ

Share this:

Comments
Don't need to search for me

But they are not completes! Something is missing...

posted by : Guest, 13 February 2012 Complain about this comment
@Richard

@Richard

For integrated GPUs, memory bandwidth is the biggest bottleneck whether doing gfx or compute.
AMD must have tweaked their cpu ucode in such a way that when CPU sees instructions pointing to memory addresses which are part of "stolen memory for graphics", it might be prefetching the data to some small ondie buffer.

posted by : arjun, 12 February 2012 Complain about this comment
@Richard

They have removed the need for the communication to go from CPU GPU via RAM. Instead they're using on-die communication which isn't just 'a little more efficient', it's actually the greatest bottleneck.

posted by : Zagdul, 11 February 2012 Complain about this comment
Don't CPUs already do that?

Isn't that what CPUs already do: feed the GPU (along with doing some other sundry tasks)?

Ok, so they have made the intercommunication between the two a little more efficient, but ... the main problem is that most general purpose computations don't really scale well to GPU architectures.

posted by : Richard, 09 February 2012 Complain about this comment
aboutus
Advertisement
Subscribe to INQ newsletters
Advertisement
INQ Poll

Facebook starts selling shares

Will you buy Facebook shares?