| ▲ | Readerium 4 hours ago | |
LLMs are memory bandwidth bound not compute bound. | ||
| ▲ | ondra 3 hours ago | parent | next [-] | |
This is incorrect, prompt processing is compute bound. | ||
| ▲ | AntiUSAbah an hour ago | parent | prev | next [-] | |
LLMs are bound by both and depends on the hardware which factor is higher. | ||
| ▲ | icelancer 2 hours ago | parent | prev [-] | |
This is only true for some parts of the time cost function. | ||