i have a dell precision m6800 laptop.
i have set sudo cpupower frequency-set -d 2000mhz
but every now and then, the system still allows it to go slower. as low as 600MHZ!
i even get analyzing CPU 0:
driver: intel_cpufreq
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 20.0 us
hardware limits: 800 MHz - 3.70 GHz
available cpufreq governors: ondemand performance schedutil
current policy: frequency should be within 2.00 GHz and 3.70 GHz.
The governor "performance" may decide which speed to use
within this range.
current CPU frequency: Unable to call hardware
current CPU frequency: 698 MHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
but the slowdowns still happen!! how do make Linux obey my commands here?
UPDATE: been doing research. this is apperently a common bug with haswell boards. ive reached out to Dell, they refuse to make a bios update to fix.
somebody at my university told me that AVX causes power spikes in some haswell systems. and that disabling avx might work.
so ive added noxsave to my boot parameters. this disables avx, but not avx2. hopefully its enough. i will run some experiments with this configuration.
while i do still suspect avx to be the cause of this issue, linux 6.8 just will not obey the parameter to disable it. unless im doing something wrong.
my current boot parameters are
and avx is clearly still enabled because my dmesg has
[ 0.132933] x86/fpu: Supporting XSAVE feature 0x004: 'AVX registers'
[ 0.142281] raid6: skipped pq benchmark and selected avx2x4
[ 0.142281] raid6: using avx2x2 recovery algorithm
[ 0.245916] xor: automatically using best checksumming function avx
[ 5.250089] AVX2 version of gcm_enc/dec engaged.
does anybody here have a working method to disable avx1 and 2?
currently building a new kernel by
CFLAGS=“-fno-tree-vectorize -mno-avx” make -jnproc bindeb-pkg.
hopefully this should result in a kernel that doesnt do avx at all.
update: this diddnt work. the kernel still uses avx. i guess the kernel just doesnt take gcc flags?
now im out of ideas as to how i might disable avx system-wide.
So, i replaced the cooling system, it now never goes over 70C, i disabled boost, sucessfully locks to 2.7GHz when idle. I used a non-avx load while capped at 2.7GHz, and this STILL happened.
Not a power issue. The laptop is rated to input a max of 300w, i measured under 100w usage when under load, just before it dropped to 600MHz.
Please tell me somebody has seen this before, and has found a solution. I know this is almost certainly a bios bug, amd i can find no modded(bugfixed) bios for this laptop anywhere on the internet.
After a decade of this laptop being iut, nobody has found an exploit/workaround to fix dell’s inabikity/refusal to program?
does it go to 600mhz over all the cores or just some of them? is this also during an all core workload? intel cpu’s will put some cores into a c3 or c6 state to allow the other cores to boost higher.
I’ve had the same issue with a thinkpad p52 from somewhere in 2017/2018 even with the cpu cool. It kept throttling the cpu to 800mhz.
anybody know of a way to implement this while stuck with dell’s bios? as this system is unsupported by coreboot.
im in need of a haswell wizard here.
i managed to compile iotools and attempt this, but i am unable to write to any of the mchbar registers.
I know its not wat you want to hear but, do you have the same problem in windows? You can try throttlestop on windows to see the reason behind the throttling.
i do not currently have the means to setup a windows environment on this laptop, nor do i have enough familiarity with windows to test for this failure.
im in university right now, such tests are beyond what i can do in my dorm. if this is still unsolved by the time i return home for winter, i will then be able to setup a windows 7 ultimate test environment. i doubt i will be able to get windows 10 to behave at all on this machine.
what could cause registers to be unwritable? silicon degradation?