-1

Is it possible to have a multi-CPU server where the CPUs do not necessarily have the same number of core counts?

For example, a group of 4 at dual core and a group of four at quad core?

This question is independent of any OS.

EDIT

David answered that this is possible. So I'd like to extend my question and ask if it's possible to mix hyperthreading and non-hyperthreading CPUS.

3
  • Why was this downvoted?
    – IamIC
    Mar 22 '13 at 16:53
  • Not the downvoter, but how is this a) practical, and b) relevant to SF? Mar 22 '13 at 19:21
  • Practical: I'm writing a threading framework. If not SF, then where?
    – IamIC
    Mar 22 '13 at 21:51
3

There are subtle changes among CPU packages. I've yet to see such a configuration supported. Last time we wanted to upgrade a machine there was no way of getting the same chip as the one installed, and even changing the existing one and adding a twin wasn't possible as support for the motherboard was discontinued.

It is probably cheaper in the end if you just either get the same CPU (might have to rummage on ebay or such, can have a hefty premium), or move to a new/larger machine.

BTW, in Linux they are just starting work on supporting the ARM BIG.little (I believe 2 fast + 2 slow cores, can switch from one pair to the other) decently, all current systems assume the processors are identical (except for "hyperthreading" shenanigans, which are popular for a while now).

1

Absolutely. There's absolutely no reason this wouldn't be possible.

4
  • Thanks. Are there any restrictions the CPUs within a group?
    – IamIC
    Mar 22 '13 at 15:51
  • @IanC: Particular CPUs or operating systems might have such restrictions, but conceptually, no. Mar 22 '13 at 15:55
  • @DavidSchwartz - I removed my comment because I haven't ever tried it, however, just for discussion sake: intel.com/content/www/us/en/processors/xeon/… - looking at that datasheet on page 26 it states: Processors must share symmetry across physical packages with respect to the number of logical processors per package, number of cores per package (but not necessarily the same subset of cores within the packages). --I'm not doubting you, just curiousity at this point. (I understand this is just 1 proc family I'm basing this on.
    – TheCleaner
    Mar 22 '13 at 16:21
  • @TheCleaner: That's a design choice Intel chose to make in that series of CPUs. Nothing compelled that design choice. Mar 22 '13 at 17:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.