Hi,

Charlie Burnes (charlie.burnes.delete@this.no-spam.com) on May 19, 2022 5:30 am wrote:
> Brendan (btrotter.delete@this.gmail.com) on May 19, 2022 4:42 am wrote:
> > I'm curious about whether they plan to support hybrid multi-socket - e.g. a 4-socket
> > motherboard with a mixture of "E-core only chips" and "P-core only chips".
>
> The E-core only (Sierra Forest) Xeon has the same socket as the P-core only (Granite Rapids) Xeon
> but the article I read says you can’t mix them in a multi-socket motherboard. I don’t know
> if that is something Intel said or something the article author inferred. If the two different
> types of Xeons have different instruction sets, they could only share data, not code, if on the
> same motherboard and special operating system support would be needed. Here is the article:
>
> nextplatform.com/2022/02/17/intel-unfolds-xeon-roadmap-with-more-cores-denser-transistors

Thanks.

I guess we'll need to wait to determine where/if the problem is - I can imagine a low level problem (e.g. incompatible APIC ID assignment methods) despite using the same socket.

To me, supporting dissimilar CPUs (for both "different ISA" and "same ISA with different instruction timings") is an interesting engineering challenge with multiple potential solutions (e.g. maybe the ability for programmers to specify which kind of CPU they want when spawning a new thread and the ability to inform compiler which CPU type to optimize for on a fine grained "per function" basis; with an "optimize for the common subset of all CPU types" option).

- Brendan