Samuel Moore writes an article for IEEE Spectrum magazine (Nov 2008) discussing how adding cores slows data intensive applications.
“With no other way to improve the performance of processors further, chip makers have staked their future on putting more and more processor cores on the same chip. Engineers at Sandia National Laboratories, in New Mexico, have simulated future high-performance computers containing the 8-core, 16‑core, and 32-core microprocessors that chip makers say are the future of the industry. The results are distressing. Because of limited memory bandwidth and memory-management schemes that are poorly suited to supercomputers, the performance of these machines would level off or even decline with more cores. The performance is especially bad for informatics applications—data-intensive programs that are increasingly crucial to the labs’ national security function.”
It is true that many applications are not scalable with more number of cores and memory bandwidth is a bottleneck. There is a lot of research going on in this area. Multicore processors are a challenge for supercomputing, but not an unsolvable one. This article briefly talks about one way of solving the problem by using memory stacked on processor. Worth a read.
6 responses so far ↓
1 Is more than 16 cores pointless? // Dec 7, 2008 at 6:31 pm
[...] Jon Stokes writes an article for ArsTechnica saying that developing multicore processors with more than 16 cores may well be pointless due to the severity of the memory wall problem. His analysis is inspired by last month’s IEEE Spectrum article titled, “Is Multicore Bad News For Supercomputers?“. [...]
2 The “Nulticore” effect // Dec 8, 2008 at 5:02 am
[...] Posts Is Multicore Bad News For Supercomputers? Is more than 16 cores [...]
3 Intel and AMD Say That We Can Compete In HPC // Dec 10, 2008 at 6:36 am
[...] the Institute for Electrical and Electronics Engineers (IEEE) publication IEEE Spectrum found that multi-core processors are not the best solution in some HPC scenarios. The problem, uncovered by engineers at Sandia National Laboratories in New Mexico, home [...]
4 Understanding and Avoiding Memory Issues with Multi-core Processors // Dec 11, 2008 at 12:04 pm
[...] Posts The “Nulticore” effect Is Multicore Bad News For Supercomputers? Is more than 16 cores pointless? Is Multicore The Way To Go? Multi-Core a Drag on Some HPCs The [...]
5 Multicore Review: Best Multicore Posts of 2008 // Dec 23, 2008 at 6:19 am
[...] Khronos Group Releases OpenCL 1.0 Specification 7. ‘101 Parallel Languages’ by Greg Pfister 8. Is Multicore Bad News For Supercomputers? [The “Nulticore” effect] 9. Data Prefetching in the Era of Multicore Processors 10. Amdahl’s [...]
6 More chip cores can mean slower supercomputing, Sandia simulation shows // Jan 14, 2009 at 6:35 pm
[...] flood of articles similar to IEEE Spectrum article “Is Multicore Bad News For Supercomputers?” are hitting the [...]