I am currently looking to upgrade my laptop. as my work involves use of GPU especially for ML projects. I am considering framework 13 pro but I might have to purchase something like an egpu or something like a DGX spark. It there a plan to have RTX spark powered motherboard?
I’d personally doubt it. I’d say that only Framework Desktop will be optimised for energy-intensive inference workloads, and for now, FW are leaning into AMD’s Strix Halo, with perhaps some dedicated GPU on the side. Another CPU type might spread their technical teams too thinly.
The other thing I’d note is that FW don’t tend to mention what they are up to outside of official teasers and official marketing. So it’s fine to ask the question, but only FW know the answer, but they won’t be hurried on a Yes. They might give you an immediate No, perhaps.
1 Like
Is RTX Spark intended for portable inference tasks, which would imply energy efficiency? Furthermore, while AMD is excellent for gaming and capable of running large language models, the complete ecosystem of readily available tools still largely belongs to Nvidia, particularly for high-fidelity robotics simulations.
No. LLMs are energy hungry, there’s no way around that.
They are power hungry but RTX spark architecture allegedly is designed to run such models locally. But we still need to see them in a benchmark when launched.
I think you’re getting confused between local running (where the model operates) and portability (whether something can be done on battery power). Yes, Spark machines are intended for local/on-device running, but not on batteries.
There are some very simple inference workloads that will run on a laptop—Ollama will run without a GPU—but it will run batteries down very quickly.
What I meant by portability pertains to physical ease of carrying a device around, not the battery drain. As I have a DGXspark on loan and when I heard that same chip js integrated for a laptop form factor, I was excited for the possibility of performing all my ML workload in a single device.