Fujitsu, the makers of the most advanced CPUs for supercomputers and SPARC servers, has decided to give Intel, Nvidia and AMD some competition in the AI and deep learning department. https://www.top500.org/news/fujitsu-will-pursue-ai-with-custom-built-processor/ The DLU(deep learning unit) coprocessor consists of 16 DPE(deep learning processing elements), each of which contains eight SIMD execution units, specifically tailored for deep learning workloads. It includes on package HBM2 and on die 6 dimensional Tofu interconnect, which will let it scale massively to tens of thousands of nodes, like PrimeHPC does. They are targeting 10x the performance of "the competetion", and since its release is schedueled for FY2018, that means the latest Nvidia and Intel deep learning chips. Given their history with K Computer, and the fact that its still at the #1 spot in HPCG and Graph500, I am inclined to believe that they will hit 10x their competitors performance.
How does 10x Perf./Watt statement they made compare to this? http://www.research.ibm.com/cognitive-computing/neurosynaptic-chips.shtml
I am familiar with those neuromorphic chips, but those are a bit different. They function differently than traditional "computer chips". Neuromorphic chips process data more like stochastic biological neurons. Its another REALLY cool technology and its actually been around for a long time now. They are going to start using those neuromorphic chips to monitor the health of and predict failures in traditional supercomputers, among other things.
I wasn't aware Fujitsu made SPARC CPUs. Apparently SPARC is still owned by Oracle, though I'm a bit surprised Fujitsu hasn't just bought the rights from them at this point.
http://www.marketwatch.com/story/nv...-artificial-intelligence-processor-2017-07-18 Financial analysts have noticed.
