is there any chance to have a cpu-board with the AMD RYZEN AI 9 HX 370 cpu.
As you are already offering an AMD type I hope this might be a not too difficult task.
Framework doesn’t really reveal their roadmap or timeline for new versions of the motherboard.
So when / if those will be available, it will announced here and the blog and social media.
1 Like
Siggi,
Out of curiosity, would you be willing to expand on your first post and explain why you are interested in that particular processor? I had a look at the specification on AMD’s web site and I agree it looks interesting - 12C24T, base clock at 2GHz boosting to 5.1, default TDP of 28W, but configurable between 15-54W and Radeon 890M, giving it some tasty display capabilities.
But I was thinking that if I were a Framework engineer reading your post, you’re not giving me anything by way of an explanation as to why you think this CPU should be on the roadmap. Given that Framework have already got both Ryzen 7 and Ryzen 9 silicon in the F16 model, the obvious question to ask would be: why this CPU.
The most obvious reason I could think of from reviewing the AMD spec was that it claims “up to 80 TOPS” of performance, listed under “AI Engine Capabilities” … But if you want to work with AI, where the critical performance features are: high core counts, high clock speeds, large cache, high memory bandwidth and then compatibility with the more popular software libraries [TensorFlow, PyTorch, Keras]… In short, nice as the AI9HX370 undoubtedly is, there are many CPUs much better for AI than this one… so it isn’t clear why you’d want this specific CPU.
Maybe if you set out your use case[s] and explain how you came to identify this CPU as the best fit for your requirements, Framework will listen. After all, if your need is popular they would be crazy to ignore you. It’s just that right now, you haven’t given them any means of figuring that out.
Meanwhile, if you’re willing to look at this slightly differently, what about considering suggesting that someone develop an Expansion Bay Module containing the CPU you’re interested in, to operate in a “second processor” configuration… OK, so I’m suggesting a 40-year-old design principle… but the idea of having a daughterboard version of “your processor of choice” that could fit neatly inside an existing machine would give you direct access to the functionality but maybe with less development cost. Especially if there are already “development boards” out there and you just need to find one with the right form factor…
If you are looking AI related stuff, then it is a major issue because the motherboard uses DDR and not the soldered LPDDR.
AI doesn’t bottleneck by CPU, but mostly by RAM speeds.
For now, our best approach is GPU, and we get lot better performance speeds