Two bugs in MSI dragon ball 1.0.0.15

[val.zapod.v155402d6]

Erm, so why MSI Dragon Ball cannot read XMP on DDR5 correctly?

Aida 64 reports for my DDR5 Patriot 5 viper 8200 MT/s:

@ 4110 MHz 38-48-48-84 (CL-RCD-RP-RAS) / 121-1213-658-535-124 (RC-RFC1-RFC2-RFCSB-WR)


While Dragon ball reports (I will use AIDA 64 way of declaring it)

@ 4110 MHz 39-47-48-84 (CL-RCD-RP-RAS) / 121-1211-657-534-124 (RC-RFC1-RFC2-RFCSB-WR)


Same for 8000 XMP profile:
@ 4000 MHz38-48-48-84 (CL-RCD-RP-RAS) / 117-1180-640-520-120 (RC-RFC1-RFC2-RFCSB-WR)


While dragon ball reports

@ 4000 MHz38-48-48-84 (CL-RCD-RP-RAS) / 117-1177-639-519-120 (RC-RFC1-RFC2-RFCSB-WR)


Looks like tRFC1 is off-by 2 or 3 and tRFC2 and tRFCsb is off-by-one.

So called long training in BIOS also sets value lower than those in XMP, BTW, but that is a good thing. What is not good is that it does not set tRFC2 as 533 for A die, instead it sets the same value as for M die 2nd gen. Also it sets wrong value for tREFI when you change only tRFC2 to 533. Will open a second bug for this, probably in BIOS.

2nd bug is that it says tRFC2 value is tRFC1, while apps like HwInfo and AIDA 64 do not even see there is FGR enabled, they do not see tRFC2.

 

[val.zapod.v155402d6]

Actually AIDA64 also had a bug in this. Newer version reports this.
@ 4100 MHz 38-48-48-83 (CL-RCD-RP-RAS) / 120-1210-656-533-123 RC-RFC1-RFC2-RFCSB-WR)