Linux Developers Discussing Possible Kernel Driver For Intel CPU Undervolting

[phoronix]

Phoronix: Linux Developers Discussing Possible Kernel Driver For Intel CPU Undervolting

While the Intel Extreme Tuning Utility (XTU) on Windows allows for undervolting laptop processors, currently on Linux there isn't any Intel-endorsed way for undervolting your CPU should you be interested in better thermal/power efficiency and other factors. But a hypothetical Linux kernel driver could be coming for filling such void...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-Intel-Undervolt-Kernel

[brent]

It would be pretty dumb if a special sauce Intel undervolting driver is allowed, but on the other hand an AMD-specific CPPC2 driver is not. Few people need undervolting, but everyone with Zen 2+ CPUs will benefit from CPPC2.

[andrei_me]

Which Intel developer that come here we can ping about it? Kayden

[Michael]

It would be pretty dumb if a special sauce Intel undervolting driver is allowed, but on the other hand an AMD-specific CPPC2 driver is not. Few people need undervolting, but everyone with Zen 2+ CPUs will benefit from CPPC2.

It's not that the AMD CPPC driver isn't 'allowed', it was that issues were pointed out and ultimately AMD decided to abandon that effort for now to push their resources elsewhere.

[ms178]

It would be pretty dumb if a special sauce Intel undervolting driver is allowed, but on the other hand an AMD-specific CPPC2 driver is not. Few people need undervolting, but everyone with Zen 2+ CPUs will benefit from CPPC2.

What about a vendor-neutral driver for both, CPPC and undervolting? At least there is a use-case for undervolting in laptops but also for getting more juice out of TDP constraint desktop CPUs to achieve a more stable turbo frequency (even on non-overclockable CPUs).

[ix900]

Just undervolt it by removing it. Solves a lot of problems lol.

[wizard69]

Is this really needed in the Linux world. Honestly for most of us we have gone Linux for reliability. Undervolting is just a foolish endeavor for the tweak happy.

[brent]

It's not that the AMD CPPC driver isn't 'allowed', it was that issues were pointed out and ultimately AMD decided to abandon that effort for now to push their resources elsewhere.

I disagree. AMD's approach was NACK'd, but maintainers didn't provide a clear alternative path forward, only a vague plan to eventually move everything to schedutil and support CPPC2 from there. That's not a good way to foster development to say the least. And as far as schedutil is concerned, it still sucks and it's going nowhere.

[brent]

What about a vendor-neutral driver for both, CPPC and undervolting? At least there is a use-case for undervolting in laptops but also for getting more juice out of TDP constraint desktop CPUs to achieve a more stable turbo frequency (even on non-overclockable CPUs).

Sure, a vendor-neutral driver would be even better. The old patchset by AMD kind of laid the groundwork for that (by extending the existing generic CPPC support), the amd-cpufreq driver was basically an "empty driver" needed mostly to enable CPPC2 support on specific CPUs instead of the generic p-state driver.

I'd argue CPPC2 support is actually more useful, it allows for lower latency frequency adjustments and by that power and performance optimization, without out-of-spec tweaking.

[DanL]

Undervolting is just a foolish endeavor for the tweak happy.

What is foolish about wanting less power use, heat, and noise? Some CPU's have a decent amount of headroom for undervolting. And these days, we have laptops that thermal throttle out of the box.