Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
pbhjpbhj
on May 14, 2023
|
parent
|
context
|
favorite
| on:
Run Llama 13B with a 6GB graphics card
There's a type of DMA for GPUs to access NVMe on the motherboard, IIRC. Perhaps that is a better solution here?
https://developer.nvidia.com/blog/gpudirect-storage/
boppo1
on May 14, 2023
[–]
Isn't pci-e latency dramatically higher than onboard vram?
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
https://developer.nvidia.com/blog/gpudirect-storage/