If anyone needs to see what 62 tokens/second looks like: <a href="https://token.qusai.pro/?speed=62" rel="nofollow">https://token.qusai.pro/?speed=62</a><p>Which seems sufficient for mortals who are not running gas town, but want a local model.
60 toks/sec may be ok for pure chat or token generation, but not for token preprocessing nor any sort of agentic work.
Good luck when you need to prefill context..
I wonder when we will start seeing small commercial accelerators integrated with NVMes; could potentially use far more bandwidth than typical NVMe drive can push thru the four lanes.<p>Given that you'd have probably a lot more capacity just to get to required bandwidth it could be used to switch very quickly between models (as all stored there would essentially be "loaded")
would this work on a virtual disk, say w s3 or dropbox
There are many of these stream-experts-from-SDDD projects now. What's needed is a guide to all of them and maybe for someone to take the best features of each of them and merge them into one uber project.