Macs with unified memory vs dedicated AI boxes which is better for local inference
Questo potrebbe essere il dispositivo che fa per te perché non devi noleggiare qualcosa nel cloud. This device can be repaid over time. If you rent a GPU in the cloud for $30, it won't give you the same cloud performance, but it can do the same things you can do in the cloud, while before it wasn't possible. So having this and being able to simply connect it to your laptop through the network is so small and small that if it's there, it's pretty interesting. So if I have to make some adjustments every day, I would think about taking it. But if I'm running Olam Open Web UI, Confi UI, running very high inference activities, I want more speed. Terry wins again, but I can't wait for someone... No matter who gives it to us, give us a device like this one able to perform the largest and worst models at the speed of the cloud. Or at least give me half of that speed. Just give me something. Now, though, I'm curious and I haven't tried it yet. I haven't tried it yet. I wonder how it will fit against a Mac. Speaking of similar things, Macs have a unified memory. Now, Mi hai visto raggruppare cinque Mac insieme. In questo momento li ho collegati a delle cose. Questo è un Mac che mi ha inviato Apple. È un Mac Studio M3 al massimo delle sue potenzialità. Ha 512 GB di memoria unificata. Mi chiedo come se la caverà contro questo tizio.