Framework mentioned that they worked with AMD extensively to try and find a way for upgradeable memory to be on this board. The fact that they went through iterations of simulations suggests to me that the basis for the motherboard was already largely designed in this stage.

As you mentioned, framework has a finite amount of resources. Once they determined that upgradeable ram is not a valid path forward without sacrificing half the compute capabilities of the APU, what should be their next step? Just can the project that already has a motherboard design? Like folks mentioned, there’s nothing really fancy outside of the motherboard. So if that works was already done, it seems financially responsible to me to provide that as an offering. AFAIK they’re the only desktop form factor for this apu suitable for a cluster, so I can see why it’s worth a gamble after extensive R&D was already invested into designing the hardest part of the product.

I’m actually really excited about this. I’m hoping that an affordable hardware platform like this might be able to kickstart interest in ROCm, and hopefully will help break the stranglehold Nvidia has on gpu compute.

It’s always been a catch-22 for ROCm, hopefully by providing large amounts of vram at a fraction of the price (a 40gb a100 gpu by itself is like 8k), maybe we can break out of the the carch-22.

Afaik all of the other computers that use this apu simply suck for compute - ones a tablet and the other is a laptop.

But being able to cluster these, and have over 500gb of vram for less than 10k? That’s really exciting for me, not from a personal perspective, but for research and companies.

I meant if the 13" had a touch screen.

I am also loathe to think about the screen wobbling complaints given that weak hinges was such a concern.