[Bug 217589] New: Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[Bug 217589] New: Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

            Bug ID: 217589
           Summary: Intel CPU maximum boost frequency reduced after resume
                    from suspend to RAM
           Product: Power Management
           Version: 2.5
          Hardware: All
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P3
         Component: cpufreq
          Assignee: linux-pm@vger.kernel.org
          Reporter: bjackson0971@gmail.com
        Regression: No

Hardware: MSI MAG Tomahawk Z790 Wifi motherboard, 7D91vH6 BIOS, with Intel
i7-13700K, p-cores overclocked to 5.5GHz, e-cores overclocked to 4.4GHz
Kernel: Linux fedora 6.3.9-200.fc38.x86_64

While running: sysbench cpu run --num-threads=24, then run cat
/sys/devices/system/cpu/cpu*/cpufreq/scaling_cur_freq | sort -rn

Correct output after reboot:
5500054
5500054
5500000
5500000
5500000
5500000
5500000
5500000
5500000
5500000
5499945
5499945
5499945
5499945
5499945
5499945
4400096
4400044
4400017
4400006
4400004
4400002
4400000
4399960

After suspend to RAM and resume, CPU frequencies are at hardware default
without overclock:
5399941
5399941
5300057
5300057
5300000
5300000
5300000
5300000
5300000
5300000
5300000
5300000
5300000
5300000
4500000
4499951
4200050
4200038
4200036
4200004
4199991
4199985
4199963
4199897

CPU boost clock should retain overclocked speeds after resume from suspend.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

Srinivas Pandruvada (srinivas.pandruvada@linux.intel.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |srinivas.pandruvada@linux.i
                   |                            |ntel.com

--- Comment #1 from Srinivas Pandruvada (srinivas.pandruvada@linux.intel.com) ---
How did you overclock?

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #2 from Brad  Jackson (bjackson0971@gmail.com) ---
The CPU was overclocked by setting the CPU ratio in the BIOS
(https://youtu.be/X26WU900IAQ?t=2632). This is the only way I'm aware of to
overclock on Linux. There is no Intel XTU for Linux and I don't know of a way
to increase maximum frequency by writing to /sys/devices.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #3 from Srinivas Pandruvada (srinivas.pandruvada@linux.intel.com) ---
I suggest to dump
rdmsr -a 0x774
before and after suspend to RAM.

They should be same.

The problem is voltage settings done via BIOS. OS doesn't have support to read
and restore. That must be done by the BIOS.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #4 from Brad  Jackson (bjackson0971@gmail.com) ---
Output of rdmsr -a 0x774 before suspend:
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
4646
2c2c
2c2c
2c2c
2c2c
2c2c
2c2c
2c2c
2c2c

After resume:
4646
4444
4444
4444
4444
4444
4444
4444
4545
4545
4545
4545
4444
4444
4444
4444
2a2a
2a2a
2a2a
2a2a
2a2a
2a2a
2a2a
2a2a

So this is a BIOS bug to report to MSI?

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #5 from Srinivas Pandruvada (srinivas.pandruvada@linux.intel.com) ---
Capture one log with:
echo "file intel_pstate.c +p" >/sys/kernel/debug/dynamic_debug/control 
echo 0 > /sys/devices/system/cpu/cpu1/online 
echo 1 > /sys/devices/system/cpu/cpu1/online 

Another with 
echo "file intel_pstate.c +p" >/sys/kernel/debug/dynamic_debug/control 
suspend to RAM and resume

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #6 from Brad  Jackson (bjackson0971@gmail.com) ---
Had to disable secure boot for debug output. First log:

intel_pstate: CPU 1 going offline
smpboot: CPU 1 is now offline
x86: Booting SMP configuration:
smpboot: Booting Node 0 Processor 1 APIC 0x1
intel_pstate: CPU 1 going online
intel_pstate: set_policy cpuinfo.max 5500000 policy->max 5500000
intel_pstate: cpu:1 min_policy_perf:11 max_policy_perf:70
intel_pstate: cpu:1 global_min:11 global_max:70
intel_pstate: cpu:1 max_perf_ratio:70 min_perf_ratio:11

Second log attached below.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #7 from Brad  Jackson (bjackson0971@gmail.com) ---
Created attachment 304475
  --> https://bugzilla.kernel.org/attachment.cgi?id=304475&action=edit
pstate debug log

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #8 from Srinivas Pandruvada (srinivas.pandruvada@linux.intel.com) ---
It is indicating BIOS issue.
Please dump:
rdmsr -a 0x771
before and after suspend to RAM.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #9 from Brad  Jackson (bjackson0971@gmail.com) ---
Before:
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
1112c46
10d1a2c
10d1a2c
10d1a2c
10d1a2c
10d1a2c
10d1a2c
10d1a2c
10d1a2c

After:
10b2c44
10b2c44
10b2c44
10b2c44
10b2c44
10b2c44
10b2c44
10b2c44
10b2c45
10b2c45
10b2c45
10b2c45
10b2c44
10b2c44
10b2c44
10b2c44
1081a2a
1081a2a
1081a2a
1081a2a
1081a2a
1081a2a
1081a2a
1081a2a

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #10 from Srinivas Pandruvada (srinivas.pandruvada@linux.intel.com) ---
The BIOS didn't restore max after suspend to RAM. So this is not OS issue.
Report to MSI.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

[Bug 217589] Intel CPU maximum boost frequency reduced after resume from suspend to RAM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=217589

--- Comment #11 from Brad  Jackson (bjackson0971@gmail.com) ---
MSI has ignored my support ticket for weeks, but after chatting with a support
person, they don't provide support for Linux or seemingly overclocking either.
I don't have a convenient way to test Windows on my PC to prove it's a BIOS
bug.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.