Hacker News new | ask | show | jobs
by huac 754 days ago
25% MFU :( maybe because of the P2P nerf?
2 comments
anthonix1 754 days ago
Maybe get a 7900 XTX. 122 TFLOPS of BF16/FP16 for less than $1k and I'm getting 55.4% MFU
link
sabareesh 753 days ago
These are not apples to apple comparison, as this is running across GPU and much bigger model
link
sabareesh 753 days ago
This much bigger model (500M), P2P is enabled via Mailbox. It is expected because of memory to compute ratio
link
huac 753 days ago
can you elaborate?
link