Kernel options for later Intel CPU's [SOLVED]

Tony0945 · Last edited by Tony0945 on Fri Jan 15, 2016 8:56 pm; edited 1 time in total

I'm running AMD k8 & K10 but considering Intel (to save power). I noticed that the kernel has only these two options:

Intel P4/Netburst (netburst?) Optimize for Intel Pentium 4, Pentium D and older Nocona/Dempsey │
│ Xeon CPUs with Intel 64bit which is compatible with x86-64. │
│ Note that the latest Xeons (Xeon 51xx and 53xx) are not based on the │
│ Netburst core and shouldn't use this option. You can distinguish them │
│ using the cpu family field │
│ in /proc/cpuinfo. Family 15 is an older Xeon, Family 6 a newer one.

And:
Core2/ Newer Xeon │
│ Select this for Intel Core 2 and newer Core 2 Xeons (Xeon 51xx and │
│ 53xx) CPUs. You can distinguish newer from older Xeons by the CPU │
│ family in /proc/cpuinfo. Newer ones have 6 and older ones 15 │
│ (not a typo)

This stuff sounds ancient. What about non-Xeon's? Haswell, Devil's canyon, Skylake? I've seen Xeons in industrial equipment but everyone else I know has a non-Xeon.

Should I stick with AMD for a pure Gentoo system despite their throwing heat like a blowtorch? Or are experimental options to use later features in the works?

NeddySeagoon · Posted: Tue Jan 12, 2016 8:27 pm Post subject:

Tony0945,

There is a patch, a patch set even. emerge gentoo-sources with the experimental USE flag.
I've not kept up with what that gives you, so you need to do some reading.

As usual, if it breaks, you can keep all the pieces.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

geki · Posted: Tue Jan 12, 2016 8:46 pm Post subject:

Just wondering.

Do you want to save power and therefore enable power saving options?
Or do you want to change the march flag for gcc?

The selection you mention does not save power, neither would a patch for other cpus to that selection.
_________________
hear hear

Tony0945 · Posted: Tue Jan 12, 2016 11:20 pm Post subject:

Thanks Neddy!

Geki: The Intel high performance CPU's take 91 watts, the AMD 220 watts. I have enough problems with ComEd complaining that my electricity usage is too high. Yes, I use the cpufreq module (I forget the name), it cuts my present CPU's power from 125W to 60 watts at idle at quarter speed.
What I'd really like is something like my old k6-3+ system that takes 40W TOTAL, but I realize that's impossible, especially since I have three hard drives.

I don't actually have an Intel system yet, I'm exploring options.

Tony0945 · Posted: Wed Jan 13, 2016 2:23 am Post subject:

geki · Posted: Wed Jan 13, 2016 6:50 am Post subject:

My Intel i5 4-core system has 45W TDP (MAX power). The Intel T-series have low power consumption. IIRC, U-series was even lower, I may be wrong.
Though, are not the fastest ones. However your needs are. Mine is power consumption.

toralf · Posted: Wed Jan 13, 2016 8:46 am Post subject:

Did you tried the ondemand governor of the kernel ? Works fine here (at Intel processors) since years.

Syl20 · l33t Joined: 04 Aug 2005 Posts: 621 Location: France

chithanh · Posted: Wed Jan 13, 2016 1:15 pm Post subject:

Nowadays for modern Intel CPUs it is mostly recommended to stay with the performance governor, and delegete power management to the intel_pstate driver instead.

1clue · Advocate Joined: 05 Feb 2006 Posts: 2569

I have a communications-oriented 8-core Atom c2758 that advertises 20w tdp, and while it's not nearly as fast at compilation as an old i7 920 it's actually faster at compression and a few other things. Not sure what it actually uses since I don't have a power meter.

I use core2 and the experimental flags, can't remember exactly what. Works fine. Use march=native.

Buffoon · Veteran Joined: 17 Jun 2015 Posts: 1369 Location: EU or US

Syl20 · l33t Joined: 04 Aug 2005 Posts: 621 Location: France

wrc1944 · Posted: Fri Jan 15, 2016 7:14 pm Post subject:

When you use the experimental gentoo-sources USE flag, among other things at the end of processor section there is a "native" option, which lets GCC detect and utilize the features your hardware (CPU) is capable of.

It's better to use native than select a specific processor option in most cases, as your current gcc compiler version knows best. After all, it is in charge of using your hardware to compile your binaries.

I would also recommend that one use -march=native in your /etc/make.conf file.

I've been doing this for years now with absolutely no problems.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.40-r5, gcc-14
kernel-6.12.8 USE=experimental python3_12.7-final-0