Anonymous
8/19/2025, 7:51:58 AM
No.103653402
>>103652805
I probably will stay for a bit with this thread and the next if you guys still have questions, I do have my extension's thread watcher on. Longer term, I'll try and come back sooner if I remember again, but might take months, sorry. I'm mostly in other places now, with Vtubers taking up less time.
If you are buying now, and you're extrapolating for future model sizes and etc. on a reasonable budget around the 128GB to 512GB range, I would say Mac given the inference is good enough and support comes by quite quickly now. The main holdoff on Mac really is speed is more important than any dollar you put in, if you want to run cutting edge models that aren't just LLMs, do model finetuning/LORA/training and/or if a new paradigm comes up again that will shake things up. Nvidia is still king in those scenarios. If you are just an above average tech user, getting a workstation/server at 128GB-512GB RAM with a GPU is a lot more painful than a Mac. In all other scenarios, a comprable PC/server is worth the hassle in spending because of the built in advantages at the low end and high end of the stack of Nvidia over Mac. At this time, even though I technically have non-Nvidia for the most part, I would not suggest it unless you are actually technically inclined who can install and tinker with Linux and etc. because that is what you need to do to keep it working in tip top shape for optimal speeds. Maybe in another year, Intel and AMD can be competitive here but not yet.
I probably will stay for a bit with this thread and the next if you guys still have questions, I do have my extension's thread watcher on. Longer term, I'll try and come back sooner if I remember again, but might take months, sorry. I'm mostly in other places now, with Vtubers taking up less time.
If you are buying now, and you're extrapolating for future model sizes and etc. on a reasonable budget around the 128GB to 512GB range, I would say Mac given the inference is good enough and support comes by quite quickly now. The main holdoff on Mac really is speed is more important than any dollar you put in, if you want to run cutting edge models that aren't just LLMs, do model finetuning/LORA/training and/or if a new paradigm comes up again that will shake things up. Nvidia is still king in those scenarios. If you are just an above average tech user, getting a workstation/server at 128GB-512GB RAM with a GPU is a lot more painful than a Mac. In all other scenarios, a comprable PC/server is worth the hassle in spending because of the built in advantages at the low end and high end of the stack of Nvidia over Mac. At this time, even though I technically have non-Nvidia for the most part, I would not suggest it unless you are actually technically inclined who can install and tinker with Linux and etc. because that is what you need to do to keep it working in tip top shape for optimal speeds. Maybe in another year, Intel and AMD can be competitive here but not yet.