This has been an interesting and very eye opening discussion for me. Joe, I'm aware of SGI's woes. I was using SGI RASC as an example since I don't have much experience in hardware accelerators in general aside from a brief fling with Paracel 5 years ago. I do like their interconnect and SSI technology but their long term viability isn't great. In general my only concern is whether or not it makes sense to get a hardware accelerator. The cluster in question is in the late planning stages today and isn't large (hundreds of CPUs max). Man power wise, there are some programmers in the group but there are very few postdocs actually doing advanced algorithms that will truly take advantage of the FPGA accelerator. I've been asked to see if the accelerators can be used with general purpose apps first and for custom algorithms when the group needs it. TIA again