61–70 of 100 posts
You guys that continue to compare DGX Spark to the Mac Studios, please remember two things: 1. Virtually every model that you'd run was developed on Nvidia gear and will run on Spark. 2. Spark has fast-as-hell interconnects. The sort of interconnects that one would want to use in an actual AI DC, so you can use more than one Spark at the same time, and RDMA, and actually start to figure out how things work the way th…
Earlier quoted context omitted.
A warning to any home consumer throwing money at hardware for AI (fair enough if you have other use cases)... Things are changing rapidly and there is a non insignificant chance that it'll seem like a big waste of money within 12 months.
For this form factor it will be likely ~2 years for the next one based on Vera CPU and whatever GPU. The 50W CPU will probably improve power efficiency. If SOCAMM2 is used it will still probably be at most near the range of 512/768 GB/s bandwidth, unless LPDDR6X / LPDDR7X or SOCAMM2 is that much better, SOCAMM on the DGX Station is just 384 GB/s w/ LPDDR5X. Form factor will be neutered for the near future, but will p…
Maybe a company is working on something totally different in secret that we cant even imagine. The amount of £ thrown into this space at the moment is enormous.
DGX Spark
pp - 1723.07/s
tg - 38.55/s
Ryzen AI Max+ 395
pp - 711.67/s
tg - 40.25/s
Is it worth the money?
Earlier quoted context omitted.
Something is wrong with your numbers: gpt-oss-20b and gpt-oss-120b should be much much faster than what you are seeing. I would suggest you familiarize yourself with llama-bench instead of ollama. Running gpt-oss-120b with a rtx 5090 and 2/3 of the experts offloaded to system RAM (less than half of the memory bandwidth of this thing), my machine gets ~4100tps prefill and ~40tps decode. Your spreadsheet shows the spar…
Your system RAM is probably 1/20th the VRAM bandwidth of the 5090 (way way less than half) unless you're running a workstation board with quad or 8 channel RAM, then it's only about 1/10th or 1/5th respectively.
You guys that continue to compare DGX Spark to the Mac Studios, please remember two things: 1. Virtually every model that you'd run was developed on Nvidia gear and will run on Spark. 2. Spark has fast-as-hell interconnects. The sort of interconnects that one would want to use in an actual AI DC, so you can use more than one Spark at the same time, and RDMA, and actually start to figure out how things work the way th…
> ollama gpt-oss 120b mxfp4 1 94.67 11.66 This is insanely slow given its 200+GB/s memory bandwidth. As a comparison, I've tested GPT OSS 120B on Strix Halo and it obtains 420tps prefill and >40tps decode.
Earlier quoted context omitted.
Hi, author here. I crowd-sourced the devices for benchmarking from my friends. It just happened that one of my friend has this device.
FYI you should have used llama.cpp to do the benchmarks. It performs almost 20x faster than ollama for the gpt-oss-120b model. Here are some samples results on my spark: ggml_cuda_init: found 1 CUDA devices: Device 0: NVIDIA GB10, compute capability 12.1, VMM: yes | model | size | params | backend | ngl | n_ubatch | fa | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | --: | ----…
It isn't that good for local LLM inferencing. It's not designed to be as such. It's designed to be a local dev machine for Nvidia server products. It has the same software and hardware stack as enterprise Nvidia hardware. That's what it is designed for. Wait for M5 series Macs for good value local inferencing. I think the M5 Pro/Max are going to be very good values.