3 ms·

I believe these chips have Uniform Memory Access via a shared memory interface on the I/O die. Am I mistaken? Edit, just confirmed: >Thanks to this improved d

I believe these chips have Uniform Memory Access via a shared memory interface on the I/O die. Am I mistaken? Edit, just confirmed: >Thanks to this improved design, each chiplet can access the memory with equal latency. The multi-core beasts can support to 4TB of DDR4 memory per socket. https://www.tomshardware.com/news/amd-64-core-128-thread-7nm-rome-cpu,38032.html https://www.tomshardware.com/news/amd-64-core-128-thread-7nm...

ip26 7y ago
They exhibit NUMA because if chiplet0 wants a line of memory that is held by chiplet4, it has to go get it from chiplet4. So the degree of NUMA is improved from the previous generation, but it is still not UMA.
- paulmd 7y ago
  No memory is held by any chiplet, it's all held by the IO die and chiplets ask the IO die to access memory for it. So there is no longer "near" and "far". In a sense, it's all "far" now (but hopefully not too far). But it is all uniform now.
  - ip26 7y ago
    The chiplets have cache, which holds copies of memory. If a process has the line open in an exclusive state, e.g. locked, other chiplets cannot just get the line from memory, because it might be out of date. So they must go ask whoever holds the lock to flush & release. https://en.wikipedia.org/wiki/MESIF_protocol https://en.wikipedia.org/wiki/MESIF_protocol
    - wmf 7y ago
      When you're talking about cache it's NUCA, not NUMA.
- Symmetry 7y ago
  Generally caches aren't considered to be "memory" in this sense or otherwise every multi-core chip would be considered NUMA since they all have private caches. Instead you normally talk about an architecture being NUMA when cores have different access speeds to different parts of main memory, as when you need to get another socket to forward you information from a RAM bank. This is something that the OS generally has to be aware of in scheduling decisions, unlike caches which are automatically managed by hardware.

ip26 7y ago

They exhibit NUMA because if chiplet0 wants a line of memory that is held by chiplet4, it has to go get it from chiplet4. So the degree of NUMA is improved from the previous generation, but it is still not UMA.

paulmd 7y ago
No memory is held by any chiplet, it's all held by the IO die and chiplets ask the IO die to access memory for it. So there is no longer "near" and "far". In a sense, it's all "far" now (but hopefully not too far). But it is all uniform now.
- ip26 7y ago
  The chiplets have cache, which holds copies of memory. If a process has the line open in an exclusive state, e.g. locked, other chiplets cannot just get the line from memory, because it might be out of date. So they must go ask whoever holds the lock to flush & release. https://en.wikipedia.org/wiki/MESIF_protocol https://en.wikipedia.org/wiki/MESIF_protocol
  - wmf 7y ago
    When you're talking about cache it's NUCA, not NUMA.
Symmetry 7y ago
Generally caches aren't considered to be "memory" in this sense or otherwise every multi-core chip would be considered NUMA since they all have private caches. Instead you normally talk about an architecture being NUMA when cores have different access speeds to different parts of main memory, as when you need to get another socket to forward you information from a RAM bank. This is something that the OS generally has to be aware of in scheduling decisions, unlike caches which are automatically managed by hardware.