nitro, a new Intel CPU is out, Knights Landing. With its 64 cores and 256 threads, it's capable of 6 teraflops (single precision). I thought you may be interested.
http://www.nextplatform.com/2016/06/20/intel-knights-landing-yields-big-bang-buck-jump/


















Thanks! Single Precision I believe won't work on many option models. Still, that is the way nVidia started out with GPUs, and eventually offered DP. I suspect this is the path Intel will take as well.
I am really waiting for the Xeon+FPGA right on chip![]()
Ah I see re: double precision.Knights Landing supports double precision, too. The throughput would be 3+ teraflops. To me, the huge advantage of Knights Landing over GPU and FPGA is that it's x86 binary compatible. That is, I don't have to do anything esoteric (like OpenCL, Verilog) to take advantage of the massive computational power of Knights Landing. Any conventional program in a high level language (C++, Java, C#) will run (without even recompilation), and scale up as long as your algo is "embarrassingly parallel".
Just something to consider, before you embark on your 1000+ Raspberry PI project.
Ah I see re: double precision.
Do you know if you can have more than one Knights Landing in a computer. So can I have a quad socket KL?