MemTest86 Reports Correct RAM GB / Doesnt List All DIMMs [Did I MemTest All DIMMs???]

[svtkobra7]

I have a bit of a mismatch between the amount of reported memory (191.9 GB which is comprised of 24 x 8 DIMMs - see figure 1), and the memory that MemTest86 provides detail on from the menu (i) and in the report provided upon test completion (see figure 2). Using (i) from the menu and also reviewing the report shows that information is only provided on 16 of the 24 DIMMs. That being said even though SPD is missing for 8 DIMMs, the result summary notes: Memory Range Tested 0x0 - 3080000000 (198656MB) and that amount would be all 24 DIMMs.

I suppose that either I'm missing something, this is a bug, or a limitation of MemTest86.

High level specifications include:
Board: Supermicro X9DRi-LN4+
CPUs: 2 x Intel Xeon E5-2680 v2 @ 2.80GHz
Memory: 24 x 8 GB - Micron 36KDYS1G72PZ-1G4M1
Test parameters: Parallel | ECC Polling Enabled | 12 of 12 tests passed.
Version: MemTest86 V7.5 Free (64-bit)

I would like to confirm that I have performed 1 full pass on all 24 DIMMs but I am unsure if that is the case.

I'm not sure how actively monitored the help@passmark.com e-mail address is, but I have provided the full log and report to that email address.

Thanks in advance for your thoughts - this has me stumped!

Figure 1


Figure 2

[David (PassMark)]

Yes, the available free RAM on all 24 DIMM would have been tested.

The SPD data for each stick can sometimes be problematic to read. But that doesn't effect the actual test, just the reporting of the part numbers & timings for each stick. The SPD data is read across the SMBus. But there can be multiple SMBus's in new machines. Especially in servers with multiple CPUs. So it can be hard to locate all the different buses. Certain vendors deliberately hide them as well.

Send us a debug log if you want
https://www.memtest86.com/support/index.htm

[svtkobra7]

Yes, the available free RAM on all 24 DIMM would have been tested.

Thanks for the confirm on that.

The SPD data for each stick can sometimes be problematic to read. But that doesn't effect the actual test, just the reporting of the part numbers & timings for each stick. The SPD data is read across the SMBus. But there can be multiple SMBus's in new machines. Especially in servers with multiple CPUs. So it can be hard to locate all the different buses. Certain vendors deliberately hide them as well.

I think there are 6 SMBus in that particular board and 2 CPUs, but to me, not knowing anything about this stuff, I'd expect to see 12 / 24 or something. I suppose missing 8 could make sense maybe because there are 4 memory channels and 1 is being excluded.

Send us a debug log if you want
https://www.memtest86.com/support/index.htm

I sent over the report and log this morning @ 9:14 AM to help@passmark.com. Can you kindly confirm receipt and that is what you are looking for, please?

Thanks for your help!

[keith]

Update on the SPD issue:

We found a bug that fails to report SPD modules greater than the maximum of 16 modules in MemTest86 V7.5. This has been fixed for the next public release. There will be no hard limit in the next release.

Note that this does not affect the memory test coverage as it always run on the full system memory regardless of what is being reported in the SPD.