News:

Willkommen im Notebookcheck.com Forum! Hier können Sie über alle unsere Artikel und allgemein über notebookrelevante Dinge diskutieren. Viel Spass!

Main Menu

Asus launches PC with 748 GB RAM and Nvidia GB300 for $99,999

Started by Redaktion, Yesterday at 21:09:26

Previous topic - Next topic

Redaktion

With the ExpertCenter Pro ET900N G3, Asus aims to offer a "desktop AI supercomputer." The workstation features an Nvidia processor with 72 ARM cores and an Nvidia Grace Blackwell Ultra GPU with a memory bandwidth of 7.1 TB/s.

https://www.notebookcheck.net/Asus-launches-PC-with-748-GB-RAM-and-Nvidia-GB300-for-99-999.1322290.0.html

razvanrux

This is the type of devices Jensen Huang is expecting the "one bilion people" to have in their homes.

captainobvious

Sounds cool until you realize it requires specialized software. No thanks, I'll stick with AMD/Intel.

CuriousDragon


100k,but only 396GB/s RAM

QuoteThe processor is paired with 496 GB of LPDDR5X RAM with a bandwidth of 396 GB/s, while the graphics chip can access 252 GB of HBM3e VRAM with an impressive bandwidth of 7.1 TB/s.
A 100k price tag and only 396 GB/s!? Also, only 496 GB RAM?

  • A AMD Strix Halo has 256 GB/s (= 256-bit * 8000 MT/s / 1000 / 8).
  • An APPLE Mac Studio M3 Ultra has 819.3 GB/s (= 1024-bit * 6400 MT/s / 1000 / 8) (en.wikipedia.org/wiki/Apple_M3) (yes: "up to 512 GB until March 4, 2026;[30]
    256 GB until May 5, 2026[31]"
  • A 500 bucks AMD Epyc Turin 9015 CPU has 614 GB/s (amd.com/en/products/processors/server/epyc/9005-series/amd-epyc-9015.html)
  • A RTX 4090 has 1008 GB/s
  • A RTX 5090 has 1.8 TB/s

396 GB/s sound like a 512-bit memory bus width chip (e.g. en.wikipedia.org/wiki/Apple_M4, en.wikipedia.org/wiki/Apple_M5).

The 7.1 TB/s of the GPU is very fast tho. But how about less very slow RAM and more, actually impressive, much faster VRAM? Ty.

Adamski

I think author is looking at this from yesterday's perspective.

Why is everyone obsessed with ports? The long-term trend is clearly toward wireless connectivity, cloud storage, high-speed networking, and fewer physical interfaces. For AI workloads and large datasets, I'd rather have ultra-fast network throughput than a dozen legacy ports.

The future isn't more cables—it's fewer.

Quick Reply

Name:
Email:
Verification:
Please leave this box empty:
Shortcuts: ALT+S post or ALT+P preview