[Solved] Intel "Chipset" Drivers?

eduncan911 · 2015-09-22 06:43:08

Does a bare Arch install not need any Intel "chipset-like" drivers to utilize full power savings, such as in Haswell?

I ask because I just installed Arch bare-metal on a server and i7z shows it idling in C6 state 99.6% of the time! Wow! (just configured the drive, not even started on setting up the machine yet).

https://wiki.archlinux.org/index.php/CP … cy_scaling <- this would indicate it's built into the kernel?

NOTE: I've already installed and configured /boot/intel-ucode.img for the Xeon C612 chipset of the Haswell E5-2690v3 in question.

TL;DR

Tried searching for clear answers and didn't see it. Wanting to know if there is anything I have to do to utilize full power savings for Intel Xeon C612 platform.

Coming from a Windows background in the past, there are always specific Intel chipset drivers to install. These do many things; but, I am focusing the power-saving features you get when installing them.

My primary reason for the question is I want to ensure full power-savings/Intel Speed-Stepping is working. So far, it looks like it is according to i7z?

I've got several Arch installs running around in VMs. This is my second bare-metal install, and on a large server.

Last edited by eduncan911 (2015-09-24 12:05:07)

graysky · 2015-09-22 10:15:25

Unlearn everything you know from windows the chipset drivers are included in your kernel. Haswell has a known bug against the patates driver. Search in our bug tracker against the kernel package for open tasks reported by me to get the link. Compile the Linux package yourself with a 1000 Hz tick rate to help. This isn't an Arch bug, it is with the kernel itself for these chips.

eduncan911 · 2015-09-23 22:01:59

Humm. I don't think I have that issue. Remember, I have an Haswell Xeon server CPU which doesn't have a GPU (I read your post about the kernel build findings with haswell).

This is from i7z-git after a bare install with the intel microcode bootstrap:

Cpu speed from cpuinfo 2394.00Mhz
cpuinfo might be wrong if cpufreq is enabled. To guess correctly try estimating via tsc
Linux's inbuilt cpu_khz code emulated now
True Frequency (without accounting Turbo) 2394 MHz
  CPU Multiplier 24x || Bus clock frequency (BCLK) 99.75 MHz

Socket [0] - [physical cores=12, logical cores=24, max online cores ever=12]
  TURBO ENABLED on 12 Cores, Hyper Threading ON
  Max Frequency without considering Turbo 2493.75 MHz (99.75 x [25])
  Max TURBO Multiplier (if Enabled) with 1/2/3/4/5/6 Cores is  30x/29x/28x/27x/27x/27x
  Real Current Frequency 1279.95 MHz [99.75 x 12.83] (Max of below)
        Core [core-id]  :Actual Freq (Mult.)      C0%   Halt(C1)%  C3 %   C6 %  Temp      VCore
        Core 1 [0]:       1247.79 (12.51x)         1    22.4       0    77.5    39      0.7761
        Core 2 [1]:       1197.99 (12.01x)         1    0.15       0    99.8    37      0.7739
        Core 3 [2]:       1197.64 (12.01x)         1    0.0721     0    99.9    37      0.7793
        Core 4 [3]:       1279.95 (12.83x)         1    0.368      0    99.4    39      0.7795
        Core 5 [4]:       1197.69 (12.01x)         1    0.147      0    99.8    38      0.7761
        Core 6 [5]:       1197.19 (12.00x)         1    0.048      0    99.9    39      0.7728
        Core 7 [6]:       1195.37 (11.98x)         1    0.0482     0    99.9    38      0.7775
        Core 8 [7]:       1196.32 (11.99x)         0    0.0487     0    99.9    39      0.7795
        Core 9 [8]:       1198.16 (12.01x)         1    0.0482     0    99.9    40      0.7764
        Core 10 [9]:      1196.30 (11.99x)         0    0.0486     0    99.9    39      0.7795
        Core 11 [10]:     1197.08 (12.00x)         1    0.0487     0    99.9    39      0.7791
        Core 12 [11]:     1196.45 (11.99x)         1    0.048      0    99.9    38      0.7764
[core-id] refers to core-id number in /proc/cpuinfo
'Garbage Values' message printed when garbage values are read

I don't seem to have the "C7" output as you mentioned in that other thread. Perhaps Xeons don't have it, and C7 is only for consumer CPUs? In my BIOS I only have up to C6 available as well.

Then again, this is an Engineering Sample CPU.

Last edited by eduncan911 (2015-09-23 22:22:30)

graysky · 2015-09-24 10:49:04

I suspect since the chip is a server CPU it may not have the C7 enabled. Ark doesn't tell. Since it's a prerelease chip, who knows about tech. specs. Your question about extra drivers is answered. Recommend you use the latest kernel version as upstream support for newer hardware is more likely in the fresh kernels. FYI, 4.2.1 is now officially a stable kernel per upstream: https://lkml.org

Arch Linux

#1 2015-09-22 06:43:08

[Solved] Intel "Chipset" Drivers?

#2 2015-09-22 10:15:25

Re: [Solved] Intel "Chipset" Drivers?

#3 2015-09-23 22:01:59

Re: [Solved] Intel "Chipset" Drivers?

#4 2015-09-24 10:49:04

Re: [Solved] Intel "Chipset" Drivers?

Board footer