Re:Tinygrad AMD Linux Driver Crash Hang Tracker and Updates
by FloatArtifact on Wednesday April 03, 2024 @09:13PM
(#39925444)
It would be great to have the premise of open source but I wouldn't if I were a CEO give in to that attitude.
Re:Tinygrad AMD Linux Driver Crash Hang Tracker and Updates
by dagelf on Wednesday April 03, 2024 @10:02PM
(#39925746)
Its not about attitude, its about logic: bottlenecks. Why do you want to bottleneck your own sales? Makes no sense.
Re:Tinygrad AMD Linux Driver Crash Hang Tracker and Updates
by FloatArtifact on Wednesday April 03, 2024 @11:06PM
(#39926149)
Well explain how it would be a bottleneck?
Re:Tinygrad AMD Linux Driver Crash Hang Tracker and Updates
by imtringued on Thursday April 04, 2024 @02:32AM
(#39927191)
You need people who have positive experiences with your products. Isn't that obvious?
People's first impressions start with busted AMD GPUs and then they will switch to NVidia and you lost a customer, forever. It's that simple.
Re:Tinygrad AMD Linux Driver Crash Hang Tracker and Updates
by dralley on Wednesday April 03, 2024 @09:45PM
(#39925657)
And, 2 days later: https://twitter.com/__tinygrad__/status/1772139983731831051
If I were George Hotz I would not live-tweet my stream of conciousness to that degree.
[1] Spoke with @LisaSu. She was very polite, but said nothing of substance and indirectly rejected my offer.
Here’s a firmware crash, run ./loop.sh to trigger. Tested on tinybox and 1x7900XTX machine, ROCm 6.0.2 and 6.0.3 preview. Must reboot to bring back.
[2] No need to hire me, just open source the 7900XTX firmware+docs and remove the signature check. We'd treat it like bring up for our own chip, get builds in CI, HITL testing, fuzzing, etc...
Deadline, end of the week? Otherwise I'm not spending more time thinking about this.
Source https://twitter.com/realGeorgeHotz/status/177166635165378186...