Fix AMD Family 10h multiplier measurement

[namazso]

Fixes #32

[namazso]

@irusanov this all is completely untested and needs LHM-side modifications, but I believe should be enough to derive the same numbers.

[namazso]

I built a quick test program similar to Gigabyte EC to test the changes:

untitled6.zip

[irusanov]

I will test the module in LHM once I have access to my AM3 system, but can't really tell when that would be - hopefully next week. I am away from home.

[tellor1]

Hello Namazso. Can you give me a helping hand on this topic? I've this BSOD issue (MEMORY_MANAGEMENT 0x1A) and I really does not know how to manage it.

I've laptop AMD Ryzen 7 PRO 3700U w/ Radeon Vega Mobile Gfx. As non advanced user, if you explain me what to do to test something and give to you the results, it would be good.

I've checked that with V241 version I do not have the BSOD, but with the latest version yes.

What shall I do?

[namazso]

@tellor1 That CPU is not family 10h - 15h. Make sure what you’re experiencing isn’t what was fixed in #33

[irusanov]

Hey, @tellor1 The problem might be related to the issue @namazso linked and it has been fixed in newer pawnio modules, but Librehardwaremonitor still uses the old modules. Librehardwaremonitor is the underlying library used in FanControl. I've compiled a new DLL with the updated modules. Keep in mind this is unofficial version and I'm in no way related to LHM or FanControl.

Note: Generally it's a good practice not to trust "random" files from the internet, so it's up to you if you trust me or not
(I'm the owner of ZenTimings). If you're unsure, then maybe ask the developer of librehardwaremonitor to update them.

If you do, then exit FanControl, extract the archive and replace the 3 files in C:\Program Files (x86)\FanControl
It will ask for admin privileges.

If that works, then the problem is most probably from the modules and LHW would need to update them.

LHW-net472-test.zip

PS: This is unrelated to K10 module.

[Rem0o]

@irusanov youa re talking about the RyzenSMU module? Can it be safely integrated into LHM? Any signature change or gotchas?

[irusanov]

@Rem0o Yes, RyzenSMU was updated in the latest PawnIO modules release. On older Zen generations there was an error with the DRAM base address, so I suspect that could be the issue. In short - older code (coming from ryzen_smu gitlab project) was combining 2 base addresses (primary and secondary/alt) into one 64bit value which was then used to read the pm table from memory, but obviously that is incorrect as the consumer app would need to split it again and get the lower 32bit. In our case it was the RyzenSMU module.

It only affects certain APUs (Zen and Zen+) which Picasso (3700U) falls in.

Should be alright to update the module(s) from the latest release (0.2.1), it is working for me (ZenTimings).

Edit: Hmm, actually I might be wrong as this seems to be fixed in previous 0.2.0 release, which LHM already includes?

[Rem0o]

It's hard to say which one is in there right now, I suspect the first version of it as there is only 1 commit. Would be nice to have the release/version in the digital signature of the modules.

My git client thinks it's a different bin as it shows a diff when I put in the 0.2.1.

[irusanov]

It's not 0.2.1 for sure as that was released after the commit, but binaries are also different than 0.2.0 release. Hopefully replacing them with latest release will solve the issue.

PS: Binary compare matches RyzenSMU with 0.1.6 release, so it's not using the fixed module and that affects Picasso for sure.