upvote
If you're using a MoE model, then why do you care about the larger RAM offered by these devices? That's the main problem with low bandwidth devices: they limit the effective ram you can make use.

I do (and have historically done) quite a work with both local LLMs and local diffusion models. I have an M3 Max MBP at 400 GB/s and also a desktop with a RTX 4090 with 1,008 GB/s

While the M3 Max MBP can serve up MoE reasonably fast (~60 token/sec)the RTX 4090 is an entirely different experience (~170 token/sec). I also do a fair bit of experimentation and am currently running a custom decoder that requires expensive look-ahead, but I'm still able to get a usable 25 token/s on the RTX.

The raison d'etre for the DGX spark is not practical home inference, but rather offering the same fundamental architecture as data center cards for a affordable CUDA prototyping. If you want to build software to run on H100s, you probably can't justify buying (and running) a single card. The DGX spark solves this by having the same fundamental setup as what those cards have.

That makes these non-NVIDIA DGX-like devices confusing to me. The entire benefit of the DGX series is the NVIDIA architecture itself.

Anyone interested in home LLMs should decide whether a Mac or a dedicated GPU is the more sensible path based on their budget and other computer use. Each has their own benefits.

reply
I run DSv4 Flash at home on 2 DGX Sparks and am pretty sure there is no more cost effective way for me to do so. I'm not interested in running smaller models.
reply
What's the tokens/sec you're getting on that setup (genuinely curious because it's a setup I haven't actually run myself)?
reply
Using regular DeepSeek-V4-Flash, I see 2000tok/s prompt processing and from 40 to 50 tok/s generation. Performance doesn't drop much at long contexts, DSv4 is really nice for that.

With DeepSeek-V4-Flash-DSpark (Deepseek's new speculative decoding scheme), which is still barely supported anywhere, we're seeing a more steady 45-55 tok/s with bursts into the 60s.

reply
> from 40 to 50 tok/s generation

That's actually much better than I would have thought!

Thanks for the answer and it does make this approach make more sense as a budget solution to running larger models locally.

reply