Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Alexander Couzens]

Hi Charlotte,

here are some description of the EC in your board. Don't know how much
you already know.

The Thinkpad has 2 EC controller.

1x Thinker
1x H8S (W530 has a SMSC chip).

The ectool dump only shows a buffer, which both side(host+ec) can use to
communicate.

The ectool usually talks to the H8S and not to the Thinker. There is a
way to community to the Thinker using different I/O address.

-
|Chipset|
-
   |
   | LPC
   |
---
| H8S |
---
   |
   | SPI
   |
-
|Thinker|
-

best,
lynxis
-- 
Alexander Couzens

mail: lyn...@fe80.eu
jabber: lyn...@fe80.eu
mobile: +4915123277221
gpg: 390D CF78 8BF9 AA50 4F8F  F1E2 C29E 9DA6 A0DF 8604


pgpRERBbonAzu.pgp
Description: OpenPGP digital signature
-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Charlotte Plusplus]

Quick update: I tried experimenting on the w530 to find which ACPI call
turns off the dGPU.

bbswitch reported it was: \_SB.PCI0.LPC.EC.PUBS._OFF however, in tests with
powertop it failed to show any improvement. A lot of information was found
on https://github.com/Lekensteyn/acpi-stuff/blob/master/notes.txt, with
further discussion on
https://github.com/Bumblebee-Project/bbswitch/issues/112

The following acpi call takes the power consumption of a W530 in the
default bios from 20W to 13W, and is reported to also work on the W520 by
http://hybrid-graphics-linux.tuxfamily.org/index.php?title=ACPI_calls

\_SB.PCI0.PEG.VID._DSM
{0xF8,0xD8,0x86,0xA4,0xDA,0x0B,0x1B,0x47,0xA7,0x2B,0x60,0x42,0xA6,0xB5,0xBE,0xE0}
0x100 0x1A {0x1,0x0,0x0,0x3}

\_SB.PCI0.PEG.VID._PS3

 _PS3 puts the card in D3 mode (what I was trying to do with setpci before)
but not in totally off model. The card is truly off when lspci shows rev
ff, and that is done by the _DSM.

powertop then reports about 12W, so _PS3 shaves 7W and _DSM shaves 1W which
is just what I may need to fix by extra consumption of 8W

Apparently with the default bios some fields in the DSDT also change when
Optimus is enabled, cf
http://www.insanelymac.com/forum/topic/273621-guide-os-x-lion-on-thinkpad-w520/page-3
:

OperationRegion (MNVS, SystemMemory, 0xBAF9D018, 0x1000) //
optimusOperationRegion (GNVS, SystemMemory, 0xBAF3FE18, 0x01A6) //
optimus
OperationRegion (MNVS, SystemMemory, 0xBF79D018, 0x1000) //
discreteOperationRegion (GNVS, SystemMemory, 0xBF73FE18, 0x01A6)  //
discrete

I could confirm that on the extracted DSDT. I have yet to make sense of the
content of MVNS and GNVS, and the consequences of the field changes.

Switching between \_SB.PCI0.PEG.VID._PS3 and \_SB.PCI0.PEG.VID._PS0 ( with
echo $i > /proc/acpi/call) changes the power consumption reliably by 7W, so
I think this controls the Nvidia power

I am now trying to replicate these results on coreboot, by studying the EC
dump with ectool -d before and after switching the power. No success so
far. I tried acpi_debug but the logs were not helpful.

In case anyone is interested, here is the relevant part of the ACPI tables.
Please read https://github.com/Lekensteyn/acpi-stuff/blob/master/notes.txt
for a possible interpretation:

Device(VID)
{
Name(_ADR, 0x00)
OperationRegion(VPCG, PCI_Config, 0x00, 0x0100)
Field(VPCG, DWordAcc, NoLock, Preserve)
{
Offset(0x2C),//Offset(44),
VSID, 32,
Offset(0x70),//Offset(112),
VPWR, 8,
}
Name(_S3D, 0x03)
Name(DGOS, 0x00)
Method(_INI, 0, NotSerialized)
{
\VUPS(0x02)
Store(\VCDL, VQDL)
Store(\VCDC, VQDC)
Store(\VCDT, VQD0)
Store(\VCDD, VQD1)
If(ISOP())
{
\VHYB(0x04, 0x01)
}
}
Method(_PS0, 0, NotSerialized)
{
If(ISOP())
{
If(DGOS)
{
\VHYB(0x02, 0x00)
Sleep(0x64)
\VHYB(0x00, 0x01)
Sleep(0x0A)
Store(0x01, \_SB.PCI0.LPC.PCRS)
Store(0x01, \_SB.PCI0.LPC.PCRQ)
Sleep(0x64)
\VHYB(0x02, 0x01)
Sleep(0x01)
\VHYB(0x08, 0x01)
Store(0x0A, Local0)
Store(0x32, Local1)
While(Local1)
{
Sleep(Local0)
If(\LCHK(0x01))
{
Break
}
Decrement(Local1)
}
Store(0x00, \_SB.PCI0.LPC.PCRQ)
\VHYB(0x04, 0x00)
\SWTT(0x01)
Store(Zero, DGOS)
}
Else
{
If(LNotEqual(VSID, 0x21D117AA))
{
\VHYB(0x04, 0x00)
}
}
\VHYB(0x09, \_SB.PCI0.PEG.VID.HDAS)
   

Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Charlotte Plusplus]

Quick update:

Just having "device pci 01.0 on end" in the devicetree results in the
following powertop measurements:
26W after boot, 21W with power savings applied, 20W at maximum power savings

So just declaring the nvidia makes things much worse, even without really
using the dGPU for anything. (there are no nvidia modules, to make sure it
would not interfere with the measurements)

Since it was in D0, I tried setting the Nvidia gpu in D3 state
setpci -s 01:00.0 CAP_PM+4.b
08
setpci -s 01:00.0 CAP_PM+4.b=0b

It didn't help with the power. To confirm whether disabling the Nvidia pcie
interface results in this +4W in maximal power savings, I completed
src/drivers/lenovo/hybrid_graphics.c with the PCI ID from my card. I could
return to 17W with all power savings applied.

So I'm back to trying to find a way to turn off the power completely with
the EC, and hoping this is where the extra 8W come from.

After some googling I found ec_access, but I don't know which register to
write to.

In case anyone is interested, here are the pci devices, of course ASPM is
already enabled (root: 0xB0 : 0x43, peripheral: 0x88 : 0x4B) and I am not
using the dGPU for anything at all.

$ cat /sys/devices/pci:00/:00:01.0/:01:00.0/power/runtime_status
suspended
$ cat /sys/devices/pci:00/:00:01.0/:01:00.1/power/runtime_status
active

$ lspci -s  00:01.0 -xxx

00:01.0 PCI bridge: Intel Corporation Xeon E3-1200 v2/3rd Gen Core
processor PCI Express Root Port (rev 09
)
00: 86 80 51 01 07 04 10 00 09 00 04 06 10 00 81 00
10: 00 00 00 00 00 00 00 00 00 01 01 00 20 20 00 20
20: 00 f0 00 f1 01 c0 f1 d1 00 00 00 00 00 00 00 00
30: 00 00 00 00 88 00 00 00 00 00 00 00 0b 01 03 00
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0a
80: 01 90 03 c8 08 00 00 00 0d 80 00 00 86 80 00 00
90: 05 a0 01 00 d8 02 e0 fe 00 00 00 00 00 00 00 00
a0: 10 00 42 01 01 80 00 00 00 00 00 00 03 ad 61 02
b0: 43 00 02 51 00 00 04 00 00 00 48 00 08 00 00 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 0e 00 00 00
d0: 03 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 00 01 00 00 00 00 00 01 00 10 00

#0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f

$ lspci -d 10de:0dfa -xxx
01:00.0 VGA compatible controller: NVIDIA Corporation GF108GLM [Quadro
1000M] (rev a1)
00: de 10 fa 0d 03 00 10 00 a1 00 00 03 10 00 80 00
10: 00 00 00 f0 0c 00 00 c0 00 00 00 00 0c 00 00 d0
20: 00 00 00 00 01 20 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 f1 60 00 00 00 00 00 00 00 0b 01 00 00
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50: 00 00 00 00 01 00 00 00 ce d6 23 00 00 00 00 00
60: 01 68 03 00 08 00 00 00 05 78 80 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 10 b4 02 00 a0 8d 00 00
80: 10 28 00 00 02 2d 05 00 4b 01 02 11 00 00 00 00
90: 00 00 00 00 00 00 00 00 00 00 00 00 10 00 00 00
a0: 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 00
b0: 00 00 00 00 09 00 14 01 00 00 00 00 00 00 00 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

#0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f

# 78 contains 10, on 88 : 4b

$ lspci -d 10de:0bea -xxx

01:00.1 Audio device: NVIDIA Corporation GF108 High Definition Audio
Controller (rev a1)
00: de 10 ea 0b 06 00 18 00 a1 00 03 04 10 00 80 00
10: 00 00 08 f1 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 60 00 00 00 00 00 00 00 0b 02 00 00
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50: 00 00 00 00 00 00 00 00 ce d6 23 00 00 00 00 00
60: 01 68 03 00 08 00 00 00 05 78 80 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 10 00 02 00 a0 8d 00 00
80: 10 28 00 00 02 2d 05 00 4b 01 02 11 00 00 00 00
90: 00 00 00 00 00 00 00 00 00 00 00 00 10 00 00 00
a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Charlotte


On Fri, Nov 18, 2016 at 4:04 PM, Charlotte Plusplus <
pluspluscharlo...@gmail.com> wrote:

> Hello
>
> Super interesting, I didn't know all that!
>
> Currently, I have only set in devicetree:
>
> device pci 00.0 on end # host bridge
> device pci 01.0 off end # NVidia
> device pci 02.0 on end # Intel
>
> and in my nvram options I have:
> hybrid_graphics_mode = Integrated Only
>
> I assumed that would be sufficient to turn off the power for the dGPU
> until I figure out a way to make it work. I will enable the NVidia and do
> more power tests.
>
> After reading hybrid_graphics.c, I understand a bit more, but I also have
> 

Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Charlotte Plusplus]

Hello

Super interesting, I didn't know all that!

Currently, I have only set in devicetree:

device pci 00.0 on end # host bridge
device pci 01.0 off end # NVidia
device pci 02.0 on end # Intel

and in my nvram options I have:
hybrid_graphics_mode = Integrated Only

I assumed that would be sufficient to turn off the power for the dGPU until
I figure out a way to make it work. I will enable the NVidia and do more
power tests.

After reading hybrid_graphics.c, I understand a bit more, but I also have
more questions:
 - can I add register "pcie_hotplug_map" = "{ 0, 1, 0}" to make the NVidia
removable?
(so that the operating system will not freak out when the NVidia disappears
from the PCI bus if I find a way to control the power by talking to the EC
and send
GFXCORE_ON_D + the other signal for the VRAM)

 - can you change the connection of the displayport? Based on the
specsheets, it is connected to the dGPU, while the internal display is
connected to the iGPU. If is it possible to control the muxes for the
internal display, I suppose it is possible for the other displays as well.
(I have just tested and I do not have any video on the displayport, and
xrandr does not detect anything)

 - there seem to be some missing IDs in pci_device_ids_nvidia : cf
http://envytools.readthedocs.io/en/latest/hw/pciid.html which agrees with
the W530 : 0x0ffb, so I will  propose a patch:
0x0dfa, /* Nvidia NVS Quadro 1000m Lenovo W520 */
0x0ffb, /* Nvidia NVS Quadro K1000m Lenovo W530
*/
0x0ffc, /* Nvidia NVS Quadro K2000m Lenovo W530 */

It may also be the reason why the Nvidia is still getting power, as Iru
noted that hybrid_graphics should turn off the power. I will test that
separately.

- until I can find a better solution, I am thinking of letting the Nvidia
show on on the PCIe bus and then sending commands to get in into advanced
sleep - like on https://wireless.wiki.kernel.org/en/users/documentation/aspm

It should be possible as the w530 lspci -v shows:
Capabilities: [60] Power Management version 3
Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA
PME(D0-,D1-,D2-,D3hot-,D3cold-)
Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=0 PME-

Would you have a better idea?

Thanks
Charlotte

On Fri, Nov 18, 2016 at 4:52 AM, Felix Held 
wrote:

> Hi!
>
> I don't know if Charlotte has added the ID of the dGPU to
>> src/drivers/lenovo/hybrid_graphics.c. Does the dGPU consume power after
>> hybrid_graphics.c disable the dGPU?
>>
> OPTIMUS_ENABLE is a PCH GPIO and controls the muxes that select if the
> internal display is connected to the iGPU or the dGPU; that's done in
> hybrid_graphics.c.
> GFXCORE_ON_D (and another signal that controls the power supply of the
> VRAM) are driven by the Thinker-1 chip; those switch on/off the power
> supply of the GPU. So to really disable the GPU you probably have to ask
> the EC to make that chip turn off the voltage for the GPU and VRAM.
>
>
> Regards
> Felix
>
> --
> coreboot mailing list: coreboot@coreboot.org
> https://www.coreboot.org/mailman/listinfo/coreboot
>
-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Felix Held]

Hi!

I don't know if Charlotte has added the ID of the dGPU to 
src/drivers/lenovo/hybrid_graphics.c. Does the dGPU consume power 
after hybrid_graphics.c disable the dGPU?
OPTIMUS_ENABLE is a PCH GPIO and controls the muxes that select if the 
internal display is connected to the iGPU or the dGPU; that's done in 
hybrid_graphics.c.
GFXCORE_ON_D (and another signal that controls the power supply of the 
VRAM) are driven by the Thinker-1 chip; those switch on/off the power 
supply of the GPU. So to really disable the GPU you probably have to ask 
the EC to make that chip turn off the voltage for the GPU and VRAM.


Regards
Felix

--
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Iru Cai]

Hi,

On Fri, Nov 18, 2016 at 4:43 PM, Felix Held 
wrote:

> Hi!
>
>
> BIOS: 11W Coreboot: 19W [...]
>>
> Can you also measure the power consumption with the GPU enabled under the
> vendor firmware?
>
> and no, the NVidia GPU was not enabled in corebooy. Both results were on
>> integrated GPU only
>>
> Have you only disabled the nVidia GPU in devicetree.cb (that will only
> disable the PCIe interface on the CPU side) or have you also disabled the
> GPU power supply via the GFXCORE_ON_D signal? Since that signal is driven
> by the power management companion chip (Thinker-1), you probably have to
> talk to the EC to do that.
>

I don't know if Charlotte has added the ID of the dGPU to
src/drivers/lenovo/hybrid_graphics.c. Does the dGPU consume power after
hybrid_graphics.c disable the dGPU?

Regards
> Felix
>
> --
> coreboot mailing list: coreboot@coreboot.org
> https://www.coreboot.org/mailman/listinfo/coreboot
>

Regards,
Iru

-- 
Please do not send me Microsoft Office/Apple iWork documents. Send
OpenDocument instead! http://fsf.org/campaigns/opendocument/
-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

Re: [coreboot] Powersavings: 8W of difference between bios and coreboot

[Felix Held]

Hi!



BIOS: 11W Coreboot: 19W [...]
Can you also measure the power consumption with the GPU enabled under 
the vendor firmware?


and no, the NVidia GPU was not enabled in corebooy. Both results were 
on integrated GPU only
Have you only disabled the nVidia GPU in devicetree.cb (that will only 
disable the PCIe interface on the CPU side) or have you also disabled 
the GPU power supply via the GFXCORE_ON_D signal? Since that signal is 
driven by the power management companion chip (Thinker-1), you probably 
have to talk to the EC to do that.


Regards
Felix

--
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

[coreboot] Powersavings: 8W of difference between bios and coreboot

[Charlotte Plusplus]

Hello

Since I started using coreboot, I have noticed my battery doesn't seem to
last as long. I have a very simple linux install on a separate partition
that I use to compare power consumption, so I decided to verify if my
perception was true. After some quick tests, I noticed the power profile
seem to have changed a whole lot, so I decided to check exactly how much by
running some tests.

I had saved results when running the thinkpad bios, so I decided to get
more measurements to compare the thinkpad bios to coreboot on the W520. The
test environment is X with just a xterm open to display powertop results.
Nothing else is running: there are just 130 wakeups per seconds (with 100
coming from tick_sched_timer), and according to powertop all the cores are
in mode C7-IVB between about 98 to 99% of the time. So it should be the
best environment to have accurate results without software interfering.

To reduce sources of error, I always start from a cold boot, with the
machine cold, no previous fan activity. With normal power saving options
(everything toggled in powertop), network stopped, and measurement repeated
every 5 seconds during 10 min and averaged, I get:
BIOS: 11W Coreboot: 19W

With the powersave governor
BIOS: 9W  Coreboot: 17W

With brightness reduced to 200
BIOS: 8W  Coreboot: 16W

The steps between each operation seem identical (activating powersave: -2W,
reducing brightness: -1W) however the baseline is off by about 8W. This is
on the exact same software environement, running the exact same linux
kernel with the exact same kernel options.  (and no, the NVidia GPU was not
enabled in corebooy. Both results were on integrated GPU only)

I found that weird, so to confirm these results I put the hard drive into a
borrowed W530. It has the exact same CPU and screen as my W520, the same
USB peripherals (except a colormeter) and a similar memory configuration (4
dimms): I get about the same as the bios results I got before: 11W on the
minimal install, going down to 8W at maximum power saving.

However, with coreboot I just can't do any better than 16W.

The difference of 8W is quite a lot, as with all powersavings applied,
running coreboot mean taking 2x as much power. The room temperature hasn't
changed (I have a thermostat), and I don't believe a single USB peripheral
can take 8W.

So I think there is only one conclusion: I have a lot of work to do on my
W520 port to get even close to the initial power savings

So I wonder if there is anything I can do to improve power savings? What
are some basic suggestions?? The full 8W difference means there should be
at least a few low hanging fruits!!

Thanks
Charlotte
-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot