Essentially, the Grace CPU is a memory and IO expander that happens to have a bunch of ARM CPU cores filling in the interior of the die, while the perimeter is all PHYs for LPDDR5 and NVLink and PCIe.

> have a bunch of ARM CPU cores filling in the interior of the die

The main OS needs to run somewhere. At least for now.

Sure, but 72x Neoverse V3 (approximately Cortex X3) is a choice that seems more driven by convenience than by any real need for an AI server to have tons of somewhat slow CPU cores.

there are uses cases where those cores are used for aux processing. there is more to these boxes than AI :-)

If someone gave me one for free, I'd totally make it my daily driver. I don't do much AI, but I always wanted to have a machine with lots of puny cores since the Xeon Phi appeared.

The justification is that processors cores aren't getting much faster, but what they are is getting more numerous - entry-level machines have between 4 and 8 cores - and adapting code to run across multiple cores is important if we want to utilise all those cores.

anybody writing single core code in 2025 professionally isn't very professional..

the core count doesn't matter. a top of the line Turin system has less than 1.4TB/s memory for the whole dual CPU system. A 2020 era A100 has 2TB/s.

fully agree!

with MGX and CX8 we see PCIe root moving to the NIC, which is very exciting.