Hi,
with the help of the Redhat documentation I found out that my Lenovo Yoga 6 13ALC6 is using simultaneous multithreading, which might be the cause for reducing the battery capacity too fast.
Unfortunately, this Lenovo offers no option of disabling it in the BIOS.
Can someone assist me with outlining the meaning of the following paragraph for CentOS Stream 10 with kernel 6.12.0.-233?
Regardless of release, in the event that the grub2-mkconfig command is used, the parameter will be overwritten. To ensure that the parameter will carry over a grub2-mkconfig, the nosmt parameter would need to be applied to the GRUB_CMDLINE_LINUX entry within the /etc/sysconfig/grub configuration file.
In my understanding, this recommends not to apply # grubby --args=nosmt --update-kernel=DEFAULT, which, however, is the only outlined command of that section.
Am I right, and what should I do to permanently disable smt?
The grubby line changes the command line for the current default kernel entry. Future updates (or separate grub2-mkconfig runs) may overwrite this. In order to apply the same command line to all future kernel updates, follow the outlined additional step.
The entire premise of the original post makes no sense to me. Disabling SMT doesn’t only cripple performance but cripples energy efficiency as well. On a modern Ryzen 5 processor there is also no practical security benefit as models as new as the Ryzen family are immune to MDS, L1TF, etc.
You’re throwing away between 18% and 70% performance gains (depending on Workload) for somewhere between 1% (and at the worst case:) to 2.5% additional power use.
Assuming you only ever get the typical worst case 18% performance uplift for the 2.5% worst-case power-draw hit, by disabling SMT you’re - in a typical best case - making your processor 7-9% less power-efficient and in a worst case where there is lots of threads which benefit from HT throwing away up to 69% of your processors performance and power efficiency.
simultaneous multithreading, which might be the cause for reducing the battery capacity too fast. … Am I right.
No. You’re actually making the problem far worse by turning off SMT. If your power draw is unexpectedly high for the workload on a cheap laptop, its more likely a crap or dying battery, too-high screen brightness, and/or bad cooling reducing the efficiency of components (assuming you’ve checked the obvious with top/iotop/etc. and the like for misbehaving processes - you can renice or kill off the naughty ones if they are not needed).
Thanks for outlining this. Now it makes sense that I cannot switch off SMT in the BIOS - which should be possible if the capacity is an optional and potentially risky one.