0
\$\begingroup\$

This question is about how multiple DDR5 DIMMs in the same memory channel are wired to the processor. This is mostly an electrical engineering question, and the goal to answer this question:

Will re-seating a processor increase the likelihood that the second DIMM per channel will function, specifically in the case when both memory channels are known to function populated with only one DIMM each (ie, both memory channels are functional)? All memory modules are known good.

I looked around at DDR5 documentation and there is quite a bit related to performance and ranks, but none that I could find about how they are physically wired to their memory bus in the layout.

My thinking is this: if the channels are each independent, and a series of memory modules on the same channel is wired as a parallel bus (with some kind of chip- or address-selector), then reseating the processor is unlikely to solve the problem when both channels are known to function. Is this true?

I have an AM5 system that sees the memory in one DIMM slot of each memory channel for the processor, but when 2x DIMMs per channel are installed, the system will not POST.

I could just reseat the processor and find out, but I would like to understand the technical details at the implementation level (at least a little better) before just tearing things apart and crossing my fingers.

\$\endgroup\$
11
  • \$\begingroup\$ This is highly processor dependent, we need a block diagram of the system or schematic \$\endgroup\$ Commented Dec 18, 2024 at 21:43
  • \$\begingroup\$ Bus loading is also dependent on things like the module's rank and whether your modules are buffered/registered, etc... the modules you've linked are not buffered/registered, the rank is unspecified, and there's no detailled datasheet (yay). \$\endgroup\$ Commented Dec 18, 2024 at 22:05
  • \$\begingroup\$ @Attie wow, ok, so maybe I just try the reseat. I was expecting that the dimms on the same channel were on a bus with chipselect pins or low-bit-address-permutation or something, but that the IO lines were basically the same for the same channel. \$\endgroup\$ Commented Dec 18, 2024 at 22:07
  • \$\begingroup\$ It is of course possible to reseat both CPU and DIMMs and get improved signal integrity (either by chance/luck or with cleaning), and this may even get memory to train that previously failed... but unless the cleaning was visibly required, I doubt the result would be stable. If you were designing a mainboard, then I think this question would fit (with a refocus on SI and validation), but as you're discussing consumer equipment, I think it's off-topic. Sorry. \$\endgroup\$ Commented Dec 18, 2024 at 22:07
  • 1
    \$\begingroup\$ A single DDR5 UDIMM has 2x 32-bit sub-channels, each sub-channel has 2x Chip-Select signals, each UDIMM socket therefore gets a separate set of CS signals. Your mainboard has 4x UDIMM sockets, two for each memory channel. If you've confirmed that all of your modules work, and have confirmed that both of the sockets on each memory channel are operational, then why is the wiring in question? \$\endgroup\$ Commented Dec 18, 2024 at 22:19

0

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.