Yes, I tried setting them to 200MHz and they just stalled, producing only HW...
YMMV, I only have a sample of 10 devices (2 in my lancelot, 4 each in Blue's Ztex 1.15y and CM1). The lancelot ran at 200MHz, but with HW errors, so I run it at 195MHz. The ztex (modded for higher core voltage) and CM1 devices were all between 204 and 220 MHz, but with some HW errors. Just be sure to use the latest bitstreams (see the github for links to dropbox) as the earlier ones were slower, though nothing new has actually been built for a couple of months now (I got fed up of the the Xilinx toolchain, just way too frustrating).
PS the ztex 1.15x build does appear to be slower than the 1.15y (it's quoted as 180MHz, error free), but that one is not my particular baby (hal7 did the work on the port, many thanks). Not sure if the speeds are like-for-like comparable as I tend to tolerate a higher HW error rate.
BTW, is the Blakecoin wallet running ok on a Raspberry Pi? I mean the chain should be small enough still to not hog all the cpu... Can we optimise the client for lightweight use or are we too lazy?
I run it on raspi and it works fine (even with both a cgminer and python miner running on the same device). The huge caveat is that the network stack locks up on a regular basis (roughly once a day) requiring a reboot (which I've automated). I don't know if this is hardware or software related (though swapping for a different raspi made no difference). I am still running the original debian wheezy distribution though (its just far too much trouble to update it).
PS One other caveat on raspi is that you may need to use the unsupported version of berlkeydb (libdb5.3++-dev) rather than the recommended version 4.8. This has the effect of making the wallet.dat incompatible with the windows/ubuntu builds. I tend to sweep funds into just a few addresses that I have paper backups for, just to be safe, but it will need care with change addresses when I need to spend some coin.
