Markus Fischbacher
Jeff GeerlingARM YOURSELF! The Compute Blade is here. Full review and testing with #RaspberryPi CM4 and clones from Radxa, Pine64, and more! https://www.youtube.com/watch?v=rKDGlpnP-vE
Shawn Powers@geerlingguy Love the YouTube thumbnail. :)
Stephen Greenham@geerlingguy Great video! Just wish the stock was free-flowing. Would love to play with this!
@solarisfire As do we all :( I hope it improves this year like @[email protected] has said!
Bret Towe@geerlingguy shame the #soquartz stuff is so hard to get going. I have seen in passing the nvme stuff works but requires several reboots for some before it behaves.
@brettowe Heh not so fun when you're trying to get it working in production!
Andreas Wachter 
Alan Sill@geerlingguy Awesome job, Jeff!
Douglas Eadline@geerlingguy
(also mirrored on the blue bird)
Great review! I'm still waiting for my dev version
(no time to play with it in any case)
Why clusters? If your app is scalable, Max memory BW of one blade times # of blades = scalable BW. On multi-core max memory BW per core tops out at (total cores)/(Motherboard memory BW).
A little over simplified, but the more cores pushed into the CPU, the more the memory BW per core decreases. This is fine for bursty cloud/web stuff, but not for HPC.