Nvidia's Project Digits is a 'personal AI supercomputer'

$3k for a 128GB standalone is quite favorable pricing considering the next best option at home is going to be a 32GB 5090 at $2k for the card alone, so probably $3k when you’re done building a rig around it.

The press-release says "up to 128GB" while the price is a single figure of $3,000. So it won't be out of the real of possibility that the 128GB version would cost quite a bit more.

The memory bandwidth has not been announced for this device. It's probably going to be more appropriate to compare vs a 128GB M4 Max (410-546GB/s MBW) or an AMD Ryzen AI Max+ 395 (yes, that's its real name) at 256GB/s of MBW.

The 5090 has 1.8TB/s of MBW and is in a whole different class performance-wise.

The real question is how big of a model will you actually want to run based on how slowly tokens generate.

From what I've seen the general consensus is that the 128GB of memory is standard across all models, and that the price would vary for different storage and networking configurations. Their marketing materials say that "Each Project DIGITS features 128GB of unified, coherent memory and up to 4TB of NVMe storage."

https://nvidianews.nvidia.com/news/nvidia-puts-grace-blackwe...

Indeed!

Well obviously it has to be low otherwise they would cannibalize their high end GPUs.

Agreed. I care more about LLM size than tokens/sec so the GB10 or Strix Halo with 128GB are my leading choices. Both look to be cheaper than a similar mac studio with 128GB (minimum $4,800 currently). Will have to wait on final config, pricing, and performance.