hello, I am considering to buy nMP 2013 after the the price drop this April. The main use is to develop multithreaded CPU and industry standard OpenCL GPU applications. I am more concerned with server grade processor stability than single core performance like 7700k in iMac 2017. I plan to run some multi days simulations. My main concerns are : 1. Apple Metal API is not as open as OpenCL. Codes written in Metal is not running in Linux without major changes. 2. Apple OpenCL driver remains version 1.2 for the years given the latest OpenCL 2.2 C++. 3. After google and check various forum, it seems only one D700 can be used for GPU compute because the other must be used for display. I plan to use nMP as headless node to run dual D700 codes. The conventional wisdom is I should wait for the mMP in 2018 and beyond. But I like the small footprint of nMP and the Xeon class cpu is good enough for my usage. I also think mMP price will be sky high then. Would appreciate if anyone experience who have tried or consider nMP for simialr usages. Also, my development usage may involve lots of host to device traffics so the eGPU might not be viable option i.e, PCIe 3 x16 16GB/s vs TB3 40Gbs. thanks.