There's a bit more to fibre channel then meets the eye, the more recent (better) cards all operate at a 10Gb speed which requires your switch to run at 10Gb as well. If you need to interface to a 1Gb network this can get expensive fast.
We are mounting the Kaby Lake i7-7700K. This may strike you as a bit of an odd choice so I'll elaborate on why we chose to go with this model. We have been running this model in one of our prototype servers for probably close to 18 months. We were never planning to actually use it though and were holding for newer cpu releases hunting single pass xtr (polysinc extreme) DSD512 up sampling capability. Alas all newer releases turned out to come with a net negative sound quality, even compared to the Sky Lake i7-6700K we were using in the original SGM2015. Everything capable of doing single pass DSD512 sounded (significantly) worse then dual pass DSD512 on the 7700K, or even single pass DSD256, Intel X series, Intel i9, AMD threadripper (by far the worst but they are doing a great job driving down Intel pricing). At this point I should probably mention we do have a solution that can do this and sound good at the same time but this would almost double the SGM's retail and significantly increase its size and weight.
The reality is:
-A significant percentage of our users use PCM dacs, therefor maximum quality PCM playback, be it bit perfect or up sampling, is a priority
-The better (higher pricerange) DSD dac's only support DSD128 and/or DSD256, providing these customers with maximum sound quality is a priority
-The few owning DSD512 capable dacs, these are usually lower pricerange dacs and a level below the dac's mentioned above, have a choice of running single pass xtr-DSD256 or dual pass xtr-DSD512, this group of clients usually have the best value/performance equation, however better sound while using single pass xtr-DSD512 undermines this value/performance equation, so they're better off with our current choice aswell.
I hope this clarifies our CPU choice, being relatively affordable also allows us to cherry pick the best performing ones which again makes a larger difference then you would expect.