Help Needed (PM Table)

added Help label

changed the description

ryzen 7 4800u (renoir) here. cant execute the included script because of the "unsupported" error. Just removing this check does not work. What should i do?

Can you provide a partial kernel log for the driver?

# dmesg | grep "ryzen_smu"

I've already tested that a 4900H had a previously unknown version which I had to add in commit 9f14b8f0. I suspect this is also the case here.

yes it was PM Table Version: 0x00370003. I just pulled and compiled again and run the pm table dump script. pm_dumps.tar.gz

@leogx9r I emailed you a pm table from my 3900x a few weeks ago, along with some more info, as I was unsure if the pm table was something that is smart to send publically, although now it looks like there shouldn't be any problem with that. Not sure if you noticed the email tho, so, pointing it out here just in case

3700U

Code Name: Picasso
PM Table Version: 0x00000000
Tester: "pigz"

I can use RyzenAdj to change what I'd assume to be some of the values reflected in the SMU data if that'd be useful. pm_dumps.tar.gz

Only relevant thing in dmesg: [ 9.071464] ryzen_smu: SMU v30.42.0

Hi,

attached pm dumps of a Ryzen 7 Pro - 4750U

Code Name: Renoir
PM Table Version: 0x00370003
Tester: "pigz"

ryzen_smu: SMU v55.65.0

Thank you, mopm_dumps_4750U.tar.gz

Greetings,

here are the dumps from a Ryzen 7 4800H, system is a Dell G5 15 SE 5505.

Code Name: Renoir
PM Table Version: 0x00370003
Tester: "gzip"

From demsg:

ryzen_smu: CPUID: family 0x17, model 0x60, stepping 0x1, package 0x0
ryzen_smu: SMU v55.57.0

With best wishes,
Tobias

pm_dumps.tar.zst

Hi,

Attached is table dump form Dell Inspiron 5405, 4700U

[13973.909197] ryzen_smu: CPUID: family 0x17, model 0x60, stepping 0x1, package 0x0 [13973.909270] ryzen_smu: SMU v55.67.0

pm_dumps.tar.gz

Acer Nitro5 (AN515-43), R7 3750H


ryzen_smu: CPUID: family 0x17, model 0x18, stepping 0x1, package 0x0
ryzen_smu: SMU v4.30.57.0

pm_dumps-R7-3750H.tar.xz

a Ryzen 3600XT

ryzen_smu: CPUID: family 0x17, model 0x71, stepping 0x0, package 0x2
ryzen_smu: SMU v46.64.0

pmdump.tar.gz

model name : AMD Ryzen 7 PRO 4750U with Radeon Graphics (HP EliteBook 845 G7)

[ 4950.389561] ryzen_smu: CPUID: family 0x17, model 0x60, stepping 0x1, package 0x0
[ 4950.390312] ryzen_smu: SMU v55.69.0

pm_dumps.tar.gz

Hi, I have a Raven Ridge but the monitor tool says the PM table version is not supported. I am attaching some information just in case.

Model: AMD Ryzen 5 PRO 2500U w/ Radeon Vega Mobile Gfx Code Name: Raven Ridge PM Table Version: 0x001E0004

pm_dump_001e0004.zip

The monitoring tool is only intended for a 3700X/3800X. For other processors you'll need to reverse-engineer the PM table via dumps and adjust the tool's source code so it can display the relevant information correctly.

Did anyone manage to reverse engineer PM table version 0x00370003? I saw that it was already uploaded a few times, but I also made a dump: model name: AMD Ryzen 3 4300U, code name: Renoir pm_dumps_00370003.tar.xz

@leogx9r Do you have any tips on how to do the reverse engineering?

Maybe this with help:

https://github.com/moson-mo/rpms/blob/main/tables.go

Those are for the 8 core models, but if you remove the core specific values to end up with 4, it should work out i guess.

Thanks a lot, I will have a look

It works like a charm! It is exactly the same table, 4 core CPUs only fill the values for cores 3-7.

Hi there,

after upgrading the BIOS to the newest (ASRock X570 Pro4 BIOS 4.00, AGESA 1.2.0.0), and and also pulling/installing the newest ryzen_smu driver (0.1.2), ryzen_smu says this PM Table (00380905) is not supported yet. I have created the dumps and attached them. Hope this helps. pm_dumps.tar.gz

Linux b0rg 5.11.0-16-generic #17 (closed)~20.04.2 SMP Fri Apr 23 18:06:13 CEST 2021 x86_64 x86_64 x86_64 GNU/Linux

with best regards

Jens

ryzen_monitor is a different project. You'll need to contact their own developer to add support for that PM table version. The driver itself already has support added in commit 71153d9e.

CPU0: AMD Athlon 300U with Radeon Vega Mobile Gfx (family: 0x17, model: 0x18, stepping: 0x1) ryzen_smu: CPUID: family 0x17, model 0x18, stepping 0x1, package 0x0

ryzen_smu: SMU v37.27.0

pm_dumps.tar.gz

mentioned in issue #2

CPU: Ryzen Renoir 4750U

$ dmesg | grep "ryzen_smu"
[20367.658081] ryzen_smu: CPUID: family 0x17, model 0x60, stepping 0x1, package 0x0
[20367.658360] ryzen_smu: SMU v55.71.0

pm_dumps.tar.gz

I'm working on a tool based on ryzen_monitor with more features and more PM tables support.

https://github.com/mann1x/ryzen_monitor_ng

Please check if it's working on your CPU and if you find bugs (there are for sure). You need ryzen_smu from my fork to use it: https://gitlab.com/mann1x/ryzen_smu

I have added Lucienne and Picasso support.

I've just tried to clone https://gitlab.com/mann1x/ryzen_smu and to use https://github.com/mann1x/ryzen_monitor_ng, loading new module seems to work well, but I must do something wrong, cause when launching ryzen_monitor, I got "Error accessing SMU: SMU Driver Version Incompatible With Library Version"

`python test.py Retrieved SMU Version: v37.27.0 Processor Code Name: Picasso PM Table: Supported SMN Offset[0x50200]: 0x00000124

Everything seems to be working properly!`

@pierreTr

That's because I forgot to commit 0.1.4 version of ryzen_smu.

Can you please try now?

@pierreTr

Made some fixes with your dump, version 2.0.1 is live

Let me know if you see or not the 2 disabled cores (without using the command line switch to show them).

Would be nice if you can gather another set of dumps with the new script. Doesn't seem like it was able to full load your cpu.

works way better! here is the new set of dumppm_dumps.tar.gz

if I run ryzen_monitor, I see 2 sets of results : one with core 0 and 2 sleeping, core 1 and 3 working, and a another set with core 1 working, 0 2 and 3 sleeping... if I run rysen_monitor -d, I see 2 sets too, same first one, second one with core 3 working, all others sleeping

Good, thanks for the info. It's not detecting the disabled cores. I'll add the hack I used for Cezanne to create a fake fuse. Please send me the output of:

dmesg | grep ryzen_smu

@pierreTr

Try version 2.0.2

ryzen_smu: CPUID: family 0x17, model 0x18, stepping 0x1, package 0x0

it miss the line with SMU version!?!

I've just git-clone ryzen_monitor. but even if I can see you have uploaded a new version, when I run it, it still pretend to be "Version: v2.0.1 (NextGeneration - ManniX fork)"?

Weird that is missing the line... Yes I forgot to update the version Does it mask the sleeping cores?

by "masking the sleeping cores", you mean in the output of ryzen_monitor, the raw between "Electrical & Thermal Constraints " and "Core Statistics (Calculated) "? if so, then no, nothing is masked, I still have 4 cores, some sleeping, some working

Yes, you should not see them. Let me check.

Should be actually above Core statistics and below system information.

Instead of this:

╭─────────┬────────────┬──────────┬─────────┬──────────┬─────────────┬─────────────┬─────────────╮ │ Core 0 │ Sleeping | 0.000 W | 0.000 V | 0.00 C | C0: 0.0 % | C1: 0.0 % | C6: 0.0 % │ │ Core 1 │ Sleeping | 0.095 W | 0.240 V | 53.63 C | C0: 3.8 % | C1: 3.6 % | C6: 92.6 % │ │ Core 2 │ Sleeping | 0.000 W | 0.000 V | 0.00 C | C0: 0.0 % | C1: 0.0 % | C6: 0.0 % │ │ Core 3 │ 3300 MHz | 4.490 W | 1.195 V | 61.86 C | C0: 100.0 % | C1: 0.0 % | C6: 0.0 % │ ╰─────────┴────────────┴──────────┴─────────┴──────────┴─────────────┴─────────────┴─────────────╯

You should see:

╭─────────┬────────────┬──────────┬─────────┬──────────┬─────────────┬─────────────┬─────────────╮ │ Core 0 │ Sleeping | 0.095 W | 0.240 V | 53.63 C | C0: 3.8 % | C1: 3.6 % | C6: 92.6 % │ │ Core 1 │ 3300 MHz | 4.490 W | 1.195 V | 61.86 C | C0: 100.0 % | C1: 0.0 % | C6: 0.0 % │ ╰─────────┴────────────┴──────────┴─────────┴──────────┴─────────────┴─────────────┴─────────────╯

my mistake, it looks like system information, except, the label "system information" is not display... and no, I see the first set, with sleeping cores (0W, 0V, 0C...)

I guess if you want the sleeping core not to be display, you should comment line 179 to 185 in ryzen_monitor.c

From what I see in ryzen_monitor.c, a core is either disabled, or working when C0>=6, or sleeping?

@pierreTr Made a stupid mistake, tried to check the family and model before they were populated. Try 2.0.3 please

works. I've noticed that "Average Core Voltage" and "Average Core CC6" are inf. probably means that sysinfo->enabled_cores_count is 0...

I'll check it out, I didn't notice it

@pierreTr

That's weird, I've enabled the hack also for when reading the dump and they are both calculated properly for me.

Do you see in the Core0 and Core1 lines the voltage and CC6?

I have fixed the enabled_cores but it doesn't explain why you would get inf. The calculation should have been done in any case, just with wrong values. I've updated to 2.0.4, let me know.

here is what I see: ╭─────────┬────────────┬──────────┬─────────┬──────────┬─────────────┬─────────────┬─────────────╮ │ Core 0 │ Sleeping | 0.075 W | 0.231 V | 44.50 C | C0: 3.7 % | C1: 6.3 % | C6: 90.0 % │ │ Core 1 │ Sleeping | 0.045 W | 0.222 V | 44.21 C | C0: 2.4 % | C1: 4.6 % | C6: 93.0 % │ ╰─────────┴────────────┴──────────┴─────────┴──────────┴─────────────┴─────────────┴─────────────╯ ╭── Core Statistics (Calculated) ───────────────┬────────────────────────────────────────────────╮ │ Highest Effective Core Frequency │ 92 MHz │ │ Highest Core Temperature │ 44.50 C │ │ Highest Core Voltage │ 0.231 V │ │ Average Core Voltage │ inf V │ │ Average Core CC6 │ inf % │ │ Total Core Power Sum │ 0.120 W │

if I stress cpu, C6 fold to 0 and Average Core CC6 becomes NaN. as I've suggested, average core voltage and core CC6 are calculated by dividing total_core_voltage and total_core_CC6 by sysinfo->enabled_cores_count (line 212). if sysinfo->enabled_cores_count is 0, then inf is found. and if total_core_CC6 is 0 too, 0/0=NaN.

@pierreTr

Yes it shouldn't have happened but I used the wrong variable.

Can you please try latest 2.0.5 and send me the output of:

ryzen_monitor --init-debug

Family: 0x17 Model: 0x18 ccds: 1 ccxs: 2 cores: 2 cores_per_ccx: 8 enabled: 8 core_disable_map: 0x0 tmp: 0xFFFFFFFF pmt: 0x5 core_disable_addr: 0x30081A38 ccds_present: 0x150078 ccds_down: 0x0 ccd_enable_map: 0x1 ccd_disable_map: 0x0

cores marked as disabled and not shown in previous version, are now marked as sleeping and displayed. averages seem to be calculated for 8 cores? ─────────┬────────────┬──────────┬─────────┬──────────┬─────────────┬─────────────┬─────────────╮ │ Core 0 │ Sleeping | 0.000 W | 0.000 V | 0.00 C | C0: 0.0 % | C1: 0.0 % | C6: 0.0 % │ │ Core 1 │ Sleeping | 0.036 W | 0.226 V | 45.13 C | C0: 2.6 % | C1: 6.0 % | C6: 91.4 % │ │ Core 2 │ Sleeping | 0.000 W | 0.000 V | 0.00 C | C0: 0.0 % | C1: 0.0 % | C6: 0.0 % │ │ Core 3 │ Sleeping | 0.037 W | 0.215 V | 44.51 C | C0: 1.9 % | C1: 3.3 % | C6: 94.9 % │ ╰─────────┴────────────┴──────────┴─────────┴──────────┴─────────────┴─────────────┴─────────────╯ ╭── Core Statistics (Calculated) ───────────────┬────────────────────────────────────────────────╮ │ Highest Effective Core Frequency │ 57 MHz │ │ Highest Core Temperature │ 45.13 C │ │ Highest Core Voltage │ 0.226 V │ │ Average Core Voltage │ 0.055 V │ │ Average Core CC6 │ 23.29 % │ │ Total Core Power Sum │ 0.073 W │

Yes, I just needed the info. Thanks. Now I'll simulate your fuses to try to get the proper coremap.

@pierreTr

I'm rewriting the whole fused core detection function. But it takes time, a lot of disassembly to go through.

Now I'm going to a fair for the weekend so I'll ask you to test again next week.

Maybe I've found the reason why the PM table for your CPU is messed up, hopefully it'll be properly populated and you can get the same info like the other CPUs.

Help Needed (PM Table)

Designs

Child items ...

Activity