News for hackers. Stuff that matters.

A comment on GPT-2: 6-Month Follow-Up. Parent comment

Re:GPT-2: 6-Month Follow-Up by the8472 on Tuesday August 20, 2019 @03:16PM (#20750852)
Possibly a stupid question, but does AMD lift such restrictions on models with its unified memory, by allowing the GPU to "page out" chunks of vram to system ram?
[ Parent | on HN ]
Re:GPT-2: 6-Month Follow-Up by p1esk on Tuesday August 20, 2019 @07:01PM (#20753120)
My guess is it would be much slower, because GPU processor would wait for data. Compare bandwidth - system RAM to GPU memory (PCIe): 16GBps vs GPU memory to GPU processor: 900GBps.
[ Parent | on HN ]
Re:GPT-2: 6-Month Follow-Up by minimaxir on Tuesday August 20, 2019 @03:26PM (#20750955)
No idea how modeling works on AMD. (most discussions are about NVidia/CUDA)
[ Parent | on HN ]
Re:GPT-2: 6-Month Follow-Up by vimy on Tuesday August 20, 2019 @07:17PM (#20753229)
Isn't that a macOS specific feature?
[ Parent | on HN ]