How the PicoCray Was Made

Re:How the PicoCray Was Made by meepmorp on Friday July 28, 2023 @09:42AM (#36906649)

I wonder what fraction of a real Cray-1’s computing power is in that thing.

[ Parent | on HN ]

Re:How the PicoCray Was Made by jameskilton on Friday July 28, 2023 @10:26AM (#36907259)

I think you need to reverse this statement actually. How many CRAY-1 would you need to match the processing power of PicoCray?

CRAY-1: 80MHz, 160 MFLOPS (https://en.wikipedia.org/wiki/Cray-1)

PicoCray: 133MHz * 8 (https://www.raspberrypi.com/products/raspberry-pi-pico/)

[ Parent | on HN ]

Re:How the PicoCray Was Made by mechagodzilla on Friday July 28, 2023 @11:02AM (#36907795)

Can a RasPi Pico do a double-precision floating point op every cycle?

[ Parent | on HN ]

Re:How the PicoCray Was Made by primis on Friday July 28, 2023 @10:24AM (#36907233)

The Cray-1 had 160 MFLOPS, and a single Pico has about 133 KFLOPS, so 8 of them would be just over 1 megaflop (not accounting for any overhead). So about 1/160th!

[ Parent | on HN ]

Re:How the PicoCray Was Made by meepmorp on Friday July 28, 2023 @11:22AM (#36908196)

Thanks!

[ Parent | on HN ]

Re:How the PicoCray Was Made by kjs3 on Friday July 28, 2023 @01:38PM (#36910378)

That's only part of the story. Cray spent a lot of time architecting an I/O system that could keep the CPU fed. So peak and sustained performance were closer than a lot of systems.

[ Parent | on HN ]