MSI to unveil desktop AI supercomputer at Computex 2025, powered by Nvidia DGX

News
By published

MSI is also expanding its deskstop system and unveiling a new motherboard

Nvidia DGX
(Image credit: Nvidia)

MSI has confirmed it will take the covers off a host of exciting new products at Computex 2025 later this month, including a brand new desktop AI supercomputer powered by Nvidia's DGX Spark platform.

The company confirmed in a press release that it will unveil its EdgeXpert MS-C931, a new desktop AI supercomputer built on the Nvidia DG Spark platform. The MS-C931 is powered by Nvidia's GB10 Grace Blackwell Superchip and is capable of 1,000 AI TOPS FP4 performance. The GB10 SoC features Nvidia Blackwell GPU architecture and fifth-generation Tensor Cores, as well as NVLink-C2C connection to Nvidia's Grace CPU, an Arm architecture core featuring 20 power-efficient chips.

Follow 3DTested on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Stephen Warwick
Stephen Warwick
News Editor
7 Comments Comment from the forums
  • DS426
    Because it has an nVidia "superchip," that makes this a desktop "supercomputer?" Wow, no wonder nVidia is always winning the mindshare race.

    Kind of a slap in the face to true supercomuting IMO. This is an "AI workstation."
    Reply
  • bit_user
    By definition, nothing you can fit on a desktop qualifies as a supercomputer. When you quote Nvidia's claims to that effect, you need to put the word in quotes.

    The article said:
    The MS-C931 is powered by Nvidia's GB10 Grace Blackwell Superchip and is capable of 1,000 AI TOPS FP4 performance.
    This is not even as fast as a RTX 5090. So, by that definition, we already have desktop supercomputers!

    The only advantage it has is more RAM than any single GPU and higher bandwidth than non-server/workstation CPUs. Still, with only 128 GB and the AI model having to share that with the OS and everything else, you're probably not far off from that amount with a 96 GB card, like the RTX Pro 6000 Blackwell. Sure, there's a bit of a price-disparity, but the dGPU is also faster.
    Reply
  • Rob1C
    Should be called a hand-held supercomputer, because it fits in your hand. While at the same time, it shouldn't be called a supercomputer because it fits in your hand; well, at least not for another 50 years or so.

    Not much different from the ASUS version shown with a great image of the motherboard.
    Reply
  • sseemaku
    So after 'AI PC's we have desktop super computers now! I guess next we will see quantum computer PCs.
    Reply
  • kealii123
    isn't 1,000 TOPS at FP4 only about as strong as a 4060ti?
    Reply
  • bit_user
    sseemaku said:
    So after 'AI PC's we have desktop super computers now!
    Well, it fits into a long pattern of Nvidia trying to play up their embedded SoC's as "supercomputers":
    https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-xavier-nx/
    https://blogs.nvidia.com/blog/jetson-generative-ai-supercomputer/
    Someone in their marketing department must be incredibly lazy or stuck in a time warp, back 2-3 decades, in order to think of these tiny things as supercomputers.
    Reply
  • bit_user
    kealii123 said:
    isn't 1,000 TOPS at FP4 only about as strong as a 4060ti?
    RTX 4000 didn't do fp4 - that was one of the few new capabilities introduced in the RTX 5000 generation.

    The RTX 5070 has a sparse fp4 rate of 998 TOPS. So, that's probably the equivalent of what they're measuring in the GB102 - because Nvidia always headlines the biggest number possible. If you want dense TOPS, the RTX 5080 hits 900 and the RTX 5090 hits 1676.
    Reply