MacBook Pro 2012 Retina

TheFiddler · 2014-06-26 07:10:09

@JoshuaJolley: have you tried installing broadcom-wl-dkms. Unfortunately, the open-source broadcom drivers have not received much love for the chip inside the rMBP, so the closed source drivers work much better. It's either that, or using an external wifi receiver with e.g. a well-supported Atheros chip.

Does anyone know how to apply the grub GPU switching trick to refind?

TheFiddler · 2014-06-29 14:32:07

Ok, I have managed to chainload grub2 from refind (it's simply a matter of generating a standalone grub.efi with the correct configuration, according to the grub2 guide.)

However, I still cannot switch to the iGPU - the dGPU is always in use. I added the following lines to my menuentry, as suggested by the wiki:

outb 0x7c2 1
outb 0x7d4 0x28
outb 0x7c2 2
outb 0x7d4 0x10
outb 0x7c2 2
outb 0x7d4 0x40

When booting with these, I either get a black screen (25% chance) or output from the dGPU (75% chance.) According to the wiki, these lines should switch output to the iGPU instead.

My kernel boot parameters are: "noapic irqpoll". I have tried booting with none, one, or both of i915 and nouveau in my initramfs - no difference.

What am I doing wrong?

Elv13 · 2014-09-06 19:57:42

Hello,

Can someone here share his/her actual power usage? I just installed Linux again on the Mac (debian, this time, I had Gentoo and Arch before, but I use this laptop for work and sometime, it just has to work). I keep earing this laptop should be arround 12 in Intel GPU, but I have twice that

<code>powerstat -d 0
Running for 480 seconds (48 samples at 10 second intervals).
ACPI battery power measurements will start in 0 seconds time

Time User Nice Sys Idle IO Run Ctxt/s IRQ/s Fork Exec Exit Watts
15:45:43 8.6 0.0 0.9 90.5 0.0 3 3238 633 0 0 0 24.41
15:45:53 8.3 0.0 0.8 90.9 0.0 1 2725 573 0 0 0 25.91
15:46:03 8.1 0.0 0.7 91.2 0.0 1 2779 548 1 0 6 25.06
15:46:13 8.4 0.0 0.8 90.8 0.0 4 2829 591 0 0 0 25.49
15:46:23 8.1 0.0 0.8 91.1 0.0 2 2723 550 0 0 0 24.09
15:46:33 8.4 0.0 0.8 90.7 0.0 2 2783 654 0 0 0 25.25
15:46:43 8.2 0.0 0.8 90.9 0.0 1 2796 521 0 0 0 24.22
15:46:53 8.4 0.0 0.8 90.8 0.0 1 2769 537 0 0 2 24.25
15:47:03 8.6 0.0 0.9 90.5 0.0 1 3193 633 0 0 1 23.94
15:47:13 8.7 0.0 0.9 90.4 0.0 2 3078 637 0 0 1 26.09
15:47:23 7.8 0.0 0.6 91.6 0.0 3 2749 536 0 0 0 24.64
15:47:33 8.4 0.0 0.8 90.8 0.1 4 2971 673 2 0 0 25.26
15:47:44 8.2 0.0 0.8 91.0 0.0 2 2803 523 0 0 0 24.16
15:47:54 8.5 0.0 0.8 90.7 0.0 1 2788 620 0 0 1 24.14
15:48:04 8.2 0.0 0.8 91.0 0.0 2 2809 532 0 0 0 24.02
15:48:14 8.5 0.0 0.8 90.6 0.0 2 2845 689 0 0 0 24.56
15:48:24 8.2 0.0 0.8 91.0 0.0 2 2750 507 0 0 0 24.07
15:48:34 8.5 0.0 0.8 90.7 0.0 1 2793 637 0 0 1 24.34
15:48:44 8.1 0.0 0.8 91.1 0.0 1 2795 524 0 0 0 24.39
15:48:54 8.0 0.0 0.6 91.3 0.0 4 2762 676 0 0 0 25.53
15:49:04 7.8 0.0 0.6 91.6 0.0 2 2762 605 0 0 0 24.63
15:49:14 8.4 0.0 0.8 90.7 0.1 3 2808 644 0 0 0 24.65
15:49:24 8.2 0.0 0.8 91.0 0.0 2 2752 572 0 0 0 24.20
15:49:34 8.5 0.0 0.8 90.7 0.0 1 2783 590 0 0 0 24.91
15:49:44 8.4 0.0 0.8 90.8 0.0 2 2782 596 0 0 0 24.30
15:49:54 8.5 0.0 0.8 90.7 0.0 2 2815 662 0 0 0 24.74
15:50:04 8.1 0.0 0.7 91.2 0.0 2 2796 617 0 0 0 25.63
15:50:14 8.3 0.0 0.8 90.9 0.0 2 2825 613 0 0 0 24.04
15:50:24 8.4 0.0 0.8 90.8 0.0 2 2752 572 0 0 0 25.30
15:50:34 8.4 0.0 0.8 90.8 0.0 2 2804 627 0 0 0 23.90
15:50:44 8.4 0.0 0.8 90.7 0.0 1 2781 635 0 0 0 26.26
15:50:54 8.3 0.0 0.8 90.9 0.0 2 2789 622 0 0 0 24.42
</code>

My display brishness is at the lowest state, I use the ethernet adaptor instead of wifi and have nothing else connected. I also used the code from Arch wiki to turn off the dGPU and blacklisted nouveau. I ran PowerTop auto-tune but I it's still running hot and at 24watts. Any idea? PowerTop is still tuning itself, so I don't know where is the issue, but I would like to have some reference data on how it should actually consume power.

frigaut · 2014-09-07 00:08:59

Same, with backlight at a minimum (still can see), plus wifi off. No activity whatsoever.

129:~ $ powerstat -d 0
Running for 480 seconds (48 samples at 10 second intervals).
ACPI battery power measurements will start in 0 seconds time

Time User Nice Sys Idle IO Run Ctxt/s IRQ/s Watts
10:02:11 0.0 0.0 0.0 100.0 0.0 1 106 127 9.44
10:02:21 0.0 0.0 0.0 100.0 0.0 1 104 127 9.40
10:02:31 0.0 0.0 0.0 100.0 0.0 1 104 126 9.40
10:02:41 0.0 0.0 0.0 100.0 0.0 1 107 126 9.38
10:02:51 0.0 0.0 0.0 100.0 0.0 1 109 127 9.36
10:03:01 0.0 0.0 0.1 99.9 0.0 1 162 140 9.37
>> here I did a "xset dpms force off"
10:03:11 0.0 0.0 0.0 99.9 0.0 1 172 103 4.96
10:03:21 0.0 0.0 0.0 99.9 0.0 1 106 68 4.91
10:03:31 0.0 0.0 0.0 100.0 0.0 1 104 70 4.87
10:03:41 0.0 0.0 0.0 100.0 0.0 1 106 68 4.87
10:03:51 0.0 0.0 0.0 100.0 0.0 1 103 69 4.87
10:04:01 0.0 0.0 0.0 100.0 0.0 1 101 64 4.85
10:04:11 0.0 0.0 0.0 100.0 0.0 1 100 67 4.85
^C-------- ----- ----- ----- ----- ----- ---- ------ ------ ------
Average 0.0 0.0 0.0 100.0 0.0 1.0 114.1 98.6 6.96
StdDev 0.0 0.0 0.0 0.0 0.0 0.0 22.7 29.7 2.25
-------- ----- ----- ----- ----- ----- ---- ------ ------ ------
Minimum 0.0 0.0 0.0 99.9 0.0 1.0 100.3 63.9 4.85
Maximum 0.0 0.0 0.1 100.0 0.0 1.0 172.0 140.4 9.44
-------- ----- ----- ----- ----- ----- ---- ------ ------ ------
Summary:
6.96 Watts on Average with Standard Deviation 2.25
129:~ $ uptime
10:07:14 up 1 day, 18:58, 5 users, load average: 0.20, 0.13, 0.08
130:~ $

I think your nvidia card is not turned off. I get about 20-22W when mine is still on.
That said, I have some crashes sometimes because nouveau is confused.
I didn't filled a bug report because this is not my primary laptop anymore.

Edit: Sorry, just re-read your post and you mention that you did turn off the card. Suggest to use powertop to identify what process is using power? I know powertop is not super accurate in that respect, but that would be a start.

edit#2: This can also be handy:
130:~ $ uname -a
Linux poliahu 3.15.8-1-ARCH #1 SMP PREEMPT Fri Aug 1 08:51:42 CEST 2014 x86_64 GNU/Linux
131:~ $ cat /proc/cmdline
vmlinuz-linux.efi root=/dev/sda4 initrd=EFI\arch\initramfs-linux.img
132:~ $

Last edited by frigaut (2014-09-07 00:14:31)

Elv13 · 2014-09-07 01:28:27

  10.20 W    100.0%                      Device         USB device: Apple Internal Keyboard / Trackpad (Apple Inc.)

PowerTop finally came to life. Apparently, someone had the same issue on Arch in page 8. Kernel 3.16.1 (backported). This one is supposed to have the Thunderbold Ethernet patch, but I havn't tried it hard yet, so far so good.

EDIT: Apparently the additional 10W is only after a long sleep and resume, or at least it's not always there. I tried to turn on and on and on the USB power management for this device, but it keep draining 10W

Last edited by Elv13 (2014-09-07 01:48:09)

tomgg · 2014-09-08 22:54:06

What _is_ this device?

5.30 W    100.0%        USB device: Bluetooth USB Host Controller (Apple Inc.)

...and how can I kill it?

Edit: It turns out that powertop had a tunable for the 5W. That being said, I still don't know what this is.

Last edited by tomgg (2014-09-22 00:42:59)

tomgg · 2014-11-05 03:44:17

The gfxCardStatus hack, (set integrated only; restart)^2, is no longer working for me. I haven't updated anything on OSX for over 2 years, but I don't believe this could break for no reason.

A question for those with a working computer: after running gfxCardStatus (gCS), setting the flag, and rebooting, does the next invocation of gCS have the desired setting or does "Integrated Only" have to be selected again. Presently, "Dynamic Switching" is always selected on subsequent boots and launches.

frigaut · 2014-11-05 04:46:19

Every time you reboot in OsX, the state gets reinitialized to "Dynamic Switching". This is the expected behaviour, always done it for me.
I understand the state gets re-initialized during the osx boot process, not the shutdown process. That's how the trick works and sets the card for linux.

Now a linux boot won't touch this state, so the card selection survives boots.

hckr · 2015-05-13 22:25:22

Anybody else getting poor performance with the broadcom-wl module and kernel 4.0? On 4.0, the broadcom had constant drops. On the lts kernel (3.14.x), though, performance is nice and stable.

Arch Linux

#201 2014-06-26 07:10:09

Re: MacBook Pro 2012 Retina

#202 2014-06-29 14:32:07

Re: MacBook Pro 2012 Retina

#203 2014-09-06 19:57:42

Re: MacBook Pro 2012 Retina

#204 2014-09-07 00:08:59

Re: MacBook Pro 2012 Retina

#205 2014-09-07 01:28:27

Re: MacBook Pro 2012 Retina

#206 2014-09-08 22:54:06

Re: MacBook Pro 2012 Retina

#207 2014-11-05 03:44:17

Re: MacBook Pro 2012 Retina

#208 2014-11-05 04:46:19

Re: MacBook Pro 2012 Retina

#209 2015-05-13 22:25:22

Re: MacBook Pro 2012 Retina

Board footer