You are not logged in.
Hi,
In Windows in Cinebench R23 I can boost to over 3GHz and that is an extremely heavy load - the laptop fans are going very fast.
In Linux, if I do a kernel compile (make -j32) the frequency of all cores drops down to 1400-1800MHz (but mostly 1500MHz) and that is a lighter load than Cinebench. The laptop fan barely does anything; only barely audible.
I have tried Linux v5.14 and v5.15-rc1 and both have the problem. I did the thing where I set the governor of all cores to "performance" and "sudo smbios-thermal-ctl --set-thermal-mode=Performance" and nothing helps.
Has anybody else had this issue?
Maybe I should start an Arch wiki page for this model - only the XPS 17 9700 has a page.
Thanks
Offline
https://wiki.archlinux.org/title/Improv … rmance#CPU
Interesting line here:
Many Intel chips will not correctly report their clock frequency to acpi_cpufreq and most other utilities , This will result in excessive messages in dmesg, which can be avoided by unloading and blacklisting the kernel module acpi_cpufreq. To read their clock speed use i7z from the i7z package. To check for correct operation of an overclocked CPU, it is recommended to do stress testing.
which may explain something.. then it goes on about stress testing:
https://wiki.archlinux.org/title/Stress … g_software
Offline
Hi, thanks for replying. After recompiling the kernel to make acpi_cpufreq not built-in and making sure it is no longer loaded, nothing has changed. Besides, that is under the overclocking heading and I am not overclocking.
Offline
i can see similar things. ironically, on single core performance it hits the fan and maxes at 4.2GHz.
Offline
I assumed it to be ok until now though. I would suggest you to try actually not comparing a compile task with cinebench. But rather try compiling linux on windows and then compare both the time needed to compile from both system + the clock speed to make sure that there really is an issue.
Offline
So I did a real validation of your claim, with 7z b.
The results 4100 avg on windows and on linux 2000 avg. So yes there is a real problem. Fans are not spinning up and the clock speed gets reduced quite quickly, and throttled. Some cores report spikes too 100° in between that might be the reason for the throttle. But not exactly sure yet what is wrong.
Offline
sudo smbios-thermal-ctl --set-thermal-mode=cool-bottom gets the fans up to live forcefully. But that has only little impact on the clock speed, it still throttles, but clocks a little higher at least.
Offline
i8k can't read the temperature, so this might be a deeper problem with that.
Offline
so the boost doesn't work, but it can handle the load up to the cpu spec (2.3ghz) if the fans are running. there is some issue with dell-smm-hwmon as well, so cpu doesn't get cooled and steps down too quickly
edit:
seems like this helps for whatever reason, can't tell you why though.
sudo rmmod intel_rapl_msr
sudo rmmod processor_thermal_device_pci_legacy
sudo rmmod processor_thermal_device
sudo rmmod processor_thermal_rapl
sudo rmmod intel_rapl_common
sudo rmmod intel_powerclamp
sudo modprobe intel_powerclamp
sudo modprobe intel_rapl_common
sudo modprobe processor_thermal_rapl
sudo modprobe processor_thermal_device
sudo modprobe intel_rapl_msr
Last edited by wzrdtales (2021-09-23 14:32:12)
Offline
I started the wiki page for this model right now https://wiki.archlinux.org/title/Dell_XPS_17_(9710)
Offline
so the boost doesn't work, but it can handle the load up to the cpu spec (2.3ghz) if the fans are running. there is some issue with dell-smm-hwmon as well, so cpu doesn't get cooled and steps down too quickly
edit:
seems like this helps for whatever reason, can't tell you why though.
sudo rmmod intel_rapl_msr sudo rmmod processor_thermal_device_pci_legacy sudo rmmod processor_thermal_device sudo rmmod processor_thermal_rapl sudo rmmod intel_rapl_common sudo rmmod intel_powerclamp sudo modprobe intel_powerclamp sudo modprobe intel_rapl_common sudo modprobe processor_thermal_rapl sudo modprobe processor_thermal_device sudo modprobe intel_rapl_msr
Thanks, that did the trick, kernel compile in 18m 50.538s, pretty good.
What is in your kernel tree you put on the wiki? Are you going to try to mainline them?
Offline
the audio driver fix + the input driver fix. If you're on wayland the keyboard and touchpad will stop randomly working. The former is landing in main probably in 5.15, the latter is not scheduled yet.
Offline
I started the wiki page for this model right now https://wiki.archlinux.org/title/Dell_XPS_17_(9710)
Hi, thanks for a nice solution for the CPU power state issue!
I did it the following way:
Create a file like /bin/cpufix
➜ ~ cat /bin/cpufix
#!/bin/bash
sudo rmmod intel_rapl_msr
sudo rmmod processor_thermal_device_pci_legacy
sudo rmmod processor_thermal_device
sudo rmmod processor_thermal_rapl
sudo rmmod intel_rapl_common
sudo rmmod intel_powerclamp
sudo modprobe intel_powerclamp
sudo modprobe intel_rapl_common
sudo modprobe processor_thermal_rapl
sudo modprobe processor_thermal_device
sudo modprobe intel_rapl_msr
Then create a systemd service
➜ ~ cat /etc/systemd/system/cpufix.service
[Unit]
Description=Fix CPU power states
[Service]
ExecStart=/bin/cpufix
[Install]
WantedBy=multi-user.target
and enable it
sudo systemctl enable cpufix
Also, it is mentioned that fingerprint works on 9710. However, I am facing the following issues:
https://gitlab.freedesktop.org/libfprin … te_1087598
are there any specific steps that need to be completed to get the fingerprint scanner to work?
Last edited by Silverthorne (2021-10-02 20:39:18)
Offline
i just have fprintd and libfprint installed with latest gnome as an interface. everything works perfectly fine here, so not sure what the issue on your side is. You could check if you got a different fingerprint reader mine is 27c6:63ac
Offline
i just have fprintd and libfprint installed with latest gnome as an interface. everything works perfectly fine here, so not sure what the issue on your side is. You could check if you got a different fingerprint reader mine is 27c6:63ac
yep, it's the same as mine. Doesn't work with either version of libfprint (1.94 or git) though, as described in the issue...
but sorry for the offtopic.
Offline
Hey all, does anybody know what would be involved to fix this properly? Do I need to hit up one of the kernel maintainers at Intel? I have tried Linux v5.16-rc4 and the problem still persists. Cheers!
Offline