Anonymous
8/7/2025, 5:57:24 PM No.106176503
>0.7 tok/s on $8000 framework AI minipc cluster
lmao big retard
lmao big retard
Replies:
Thread 106176503 - /g/ [Archived: 297 hours ago]
Anonymous
8/7/2025, 5:59:20 PM No.106176527
>>106176503 (OP)
i don't really know much about these things, but why can macs run LLM and textbots so well but absolutely shit themselves for image generation?
i don't really know much about these things, but why can macs run LLM and textbots so well but absolutely shit themselves for image generation?
Anonymous
8/7/2025, 6:04:16 PM No.106176584
>>106176503 (OP)
Isn't lower better here? Aren't tokens like quarters for an arcade.
Isn't lower better here? Aren't tokens like quarters for an arcade.
Anonymous
8/7/2025, 6:06:00 PM No.106176604
>>106176503 (OP)
WTF is this labeling. I was looking for the t/s for the mac studio for way too long.
WTF is this labeling. I was looking for the t/s for the mac studio for way too long.
Anonymous
8/7/2025, 6:07:15 PM No.106176627
>>106176503 (OP)
>runs llama on the framework
>runs deepseek on the other devices
Wow, totally fair comparison.
>runs llama on the framework
>runs deepseek on the other devices
Wow, totally fair comparison.
Anonymous
8/7/2025, 6:26:31 PM No.106176831
>>106176503 (OP)
To be fair AMD still is fucking dogshit with compute drivers and the NPUs are still dead silicon
To be fair AMD still is fucking dogshit with compute drivers and the NPUs are still dead silicon
Anonymous
8/7/2025, 9:33:24 PM No.106180983
>>106176503 (OP)
why didn't they just run ollama on all of them am I missing something
why didn't they just run ollama on all of them am I missing something