possible i915kms regressions (moved from questions@)

[Kevin Oberman]

On Sat, Apr 25, 2015 at 9:26 AM, Patrick Powell <papowell at astart.com> wrote:

> On 04/24/15 08:54, Joseph Mingrone wrote:
>
>> Henry Hu <henry.hu.sh at gmail.com> writes:
>>
>>> Have you tried setting
>>> drm.i915.enable_rc6=7
>>> in /boot/loader.conf?
>>>
>> I just did after your message.
>>
>>  If you have it enabled, you would see
>>> info: [drm] Enabling RC6 states: RC6 on, RC6p on, RC6pp on
>>> in messages. It helped me reducing the GPU temperature.
>>>
>> Before today I saw, e.g.,
>>
>> Mar 31 20:07:07 phe kernel: info: [drm] Enabling RC6 states: RC6 off,
>> RC6p off, RC6pp off
>>
>> now I see
>>
>> Apr 24 12:34:14 phe kernel: info: [drm] Enabling RC6 states: RC6 on,
>> RC6p on, RC6pp on.
>>
>> Now the temperatures, like before the upgrade, are in a normal range.
>> Thank you for the tip Henry, much appreciated.
>>
>> Joseph
>>
> Are there any details about this?
>

It's mostly undocumented. RC6 states are power management states in the
Intel APUs. By enabling them, you get a substantially cooler system.

What is needed is a bit better dissemination of how to turn it on. To do
so, add "drm.i915.enable_rc6=7" to /boot/loader.conf. There have been
reports that RC6p may cause GPU hangs on Sandy Bridge APUs. Setting the
value to "5" will disable RC6p while leaving RC6 and TC6pp. It is rumored
that the code already disables RC6p on Sandy Bridge APUs, but I'm paranoid,
so I don't enable it.

My thanks to Jan Kokemueller who provided most of this information in
answer to my question about the enabling of RC6 back in February. It
requires fairly recent i915 code. Works on 10 and current. I assume it
works on 9, but I have no 9 systems.
--
Kevin Oberman, Network Engineer, Retired
E-mail: rkoberman at gmail.com