Open JackMyers001 opened 5 years ago
I was expecting to get somewhere in the neighbourhood of 1GHz at 0.5 - 0.7w running Mojave (seen several people online get around this figure.)
Where have you seen this? Cause my cpu idles at 1,4-1,5w at 1ghz.
@Mateo1234454545 this original guide for Sierra says that OP gets "1.0GHz @ 0.5W idle <-> 3.5GHz @ 24W+ under load" (under the "Results to Expect" section).
The same happen here. My XPS 13 9360 have i5-7200, 8G RAM, HP EX920 SSD, FHD display with Mojave installed . The CPU idled between 0.5 and 0.8GHz, at 1.3 - 1.8w.
Note that the original guide is using a model with i7-7500U, the model I use is the i7-8550U, as such I cannot test how i7-7500U behaves.
As a test, I would:
I have the suspicion that AppleALC uses more power than strictly required.
With regards to the actual CPU power package, I have no other ideas besides the steps you did CPUFriend or HWP. According to the ioreg they are working as expected and PikerAlpha's kext shows all relevant CPU power states are reached.
I found the SSDT-TYPC.aml was using extra power. Removing it dropped power consumption for me.
I found the SSDT-TYPC.aml was using extra power. Removing it dropped power consumption for me.
Removing it also works for me.
Works for me too. But also disable Type C port
Can confirm that disabling this ACPI patch cuts the CPU pkg in half (for me from 1,52 to 0,70 while idle). However, I noticed that the frequency is now stuck at 1,50 GHz (I5 7200u). Anyone noticed the same behaviour? (After enabling the ACPI patch an rebooting CPU throttles back down to 0,6 GHz as its supposed to)
@Richylitschy I noticed this too.
Doesn't seem to affect battery life though, as I'm getting around 1 1/2 hrs more with this fixed applied
Give this updated SSDT-TYPC.aml
a try (place it in /EFI/CLOVER/ACPI/patched
) and see if it reduces the power usage of the CPU package.
@the-darkvoid with the updated SSDT-TYPC.aml
, the package stays the same as without it (0.85-0.9w idle), and the clock speed also remains the same (1.6-1.7 GHz idle, randomly dipping to as low as 1.3 GHz for a fraction of a second). However, USB-C now works (it didn't work without the SSDT before).
If you figure out a way to get the clocks down to where they were with the old SSDT (possibly lowering the package even further) then let us know, however I think this patch fixes the power draw and keeps USB-C.
Thanks @the-darkvoid !
@the-darkvoid with the updated
SSDT-TYPC.aml
, the package stays the same as without it (0.85-0.9w idle), and the clock speed also remains the same (1.6-1.7 GHz idle, randomly dipping to as low as 1.3 GHz for a fraction of a second). However, USB-C now works (it didn't work without the SSDT before).If you figure out a way to get the clocks down to where they were with the old SSDT (possibly lowering the package even further) then let us know, however I think this patch fixes the power draw and keeps USB-C.
Thanks @the-darkvoid ! Hi, you can try run the following command in terminal: sh -c "$(curl -fsSL https://raw.githubusercontent.com/daliansky/XiaoMi-Pro/master/one-key-cpufriend/one-key-cpufriend.sh)" to make you own cpufriend.kext, but my cpu is 8250u.
If you figure out a way to get the clocks down to where they were with the old SSDT (possibly lowering the package even further) then let us know, however I think this patch fixes the power draw and keeps USB-C.
I expected only the power draw to drop, I am surprised the change affected the clock speed as well. The modification simply turns off USB-C more aggressively when its not used.
If you figure out a way to get the clocks down to where they were with the old SSDT (possibly lowering the package even further) then let us know, however I think this patch fixes the power draw and keeps USB-C.
I expected only the power draw to drop, I am surprised the change affected the clock speed as well. The modification simply turns off USB-C more aggressively when its not used.
Fix worked for the power draw! Thanks a lot! The clock speed is still hovering around 1,5 GHz though.
@the-darkvoid with the updated
SSDT-TYPC.aml
, the package stays the same as without it (0.85-0.9w idle), and the clock speed also remains the same (1.6-1.7 GHz idle, randomly dipping to as low as 1.3 GHz for a fraction of a second). However, USB-C now works (it didn't work without the SSDT before).If you figure out a way to get the clocks down to where they were with the old SSDT (possibly lowering the package even further) then let us know, however I think this patch fixes the power draw and keeps USB-C.
Thanks @the-darkvoid !
Does this change persist for you after sleep? For me the power draw stays at 1,30 (instead of the 0,7 ) after waking up from sleep. Still an improvement, but not as good as without the .aml
@Richylitschy, try the attached SSDT-TYPC
.
Also modify your config.plist and remove all dropped tables under the "DropTables" key. This should make the XPS 13 behave as "vanilla" as possible.
@Richylitschy, try the attached
SSDT-TYPC
.Also modify your config.plist and remove all dropped tables under the "DropTables" key. This should make the XPS 13 behave as "vanilla" as possible.
This patch reduced my power consumption by 50% across the board and USB-C still works.
Now idles around 1W, with a bump up to 4W if USB-C is plugged in. I used a USB-C -> HDMI adapter in the below screenshot.
The latest 10.14.6 update added an update with regards to NVMe support.
In the System Report
screen under NVMe Express
, it now shows Generic SSD Controller
.
I believe this might be positive for power management. Can others corroborate this?
I believe this might be positive for power management. Can others corroborate this?
definitely also seeing this in my system report. cannot yet tell whether it makes any difference in terms of power management though
I have a 9360 with the i7-7560U, 8GB RAM, SK Hynix 256GB SSD, FHD display
I used this repo to setup Mojave 10.14.5 and then updated to 10.14.6. CPU idles at 800Mhz at 2.5 to 3W package power (SD Card not disabled as I use it in Windows Dual Boot).
Biggest problem is that when I try to run XCode/iOS Simulator the systems freezes after few seconds because CPU boosts to 3.5+Ghz and when temperatures reach about 98C the system just freezes and reboots.
I made my own CPUFriend.kext/CPUFriendDataProvider.kext using https://github.com/stevezhengshiqi/one-key-cpufriend (choosing 800Mhz for idle) and also changed the SMBios to MacPro 14,1 (as that is a closer match for the i7 7560U than MacPro 15,2)
Would anyone know if the problem of CPU boosting and not thermal throttling properly is related to the initial idle package power (where I get 2.5+W in Intel Power Gadget)?
Biggest problem is that when I try to run XCode/iOS Simulator the systems freezes after few seconds because CPU boosts to 3.5+Ghz and when temperatures reach about 98C the system just freezes and reboots.
This sounds more like some messed up ACPI patching to me, very strange behavior... not sure though
This sounds more like some messed up ACPI patching to me, very strange behavior... not sure though
Thanks for the hint, I had used --compile-dsdt
from this repo and copied the created files into EFI/CLOVER/ACPI/patched
However the comment about ACPI led me to find this repo https://github.com/hoanX/xps13-9360-i7-7560u (same CPU as me) and I deleted the whole EFI/CLOVER folder and replaced with the one from that repo and this has fixed my CPU Temperature/Freeze problem under high load.
@sbaier1 do you see any difference now? Is it worth updating?
@rhaenni i had exactly the same issue, when I ran a simulator, osx freezed and reboot. With the repository you linked I can now re-use simulators.
Can someone investigate on that issue ? Do you need that we create a new issue ?
Thank you so much !
HI, I have too an XPS 9360 with an i5 7200u. I can't go below 1.79W measuring with the Intel Power Gadget.
I have tried disabling the audio and removing its kext. I've tried the new SSDT-TYPC.aml of this thread and also have tried to remove it but nothing. I'm always stuck at around 1.79-2W... which makes me wondering if my laptop is even using correctly CPUFriend as with or without I have the same result. Anything else I could check ? I have BIOS2.10 and have used the BIOS settings of the guide. Could it be the Hynix SSD sucking all that extra watt ?
Have you tried https://github.com/acidanthera/NVMeFix ?
This reduced my power consumption quite remarkably.
Have you tried https://github.com/acidanthera/NVMeFix ?
This reduced my power consumption quite remarkably.
What NVME do you have ? (model)
WDS100T2X0C-00L350
NVMeFix is now part of the repo, anyone seeing a positive change?
in general yes, it drastically lowers power consumption and the fan runs way less often.
in general yes, it drastically lowers power consumption and the fan runs way less often.
I've tried but no luck on my Hynix NVME...what NVME do you have ? Have you just copied NVMEFix.kext and changed boot args ? Or something else ? Did you measure a change with Intel Power Gadget ?
@dropbrick, can you post a screenshot of hackintool pci tab?
in general yes, it drastically lowers power consumption and the fan runs way less often.
I've tried but no luck on my Hynix NVME...what NVME do you have ? Have you just copied NVMEFix.kext and changed boot args ? Or something else ? Did you measure a change with Intel Power Gadget ?
i just installed the kext and checked the power states with smartctl
and it behaved like the README states it should (additional power states added).
i am using a WDS100T2X0C formatted to 4KB sectors.
i did not check the power consumption before and after, and i don't think power gadget is a good indicator for this as it only tracks CPU package consumption AFAIK(?).
coconutbattery does report a much lower consumption in general though (about 6 watts at "normal" brightness, streaming video and playing audio through bluetooth, and no fans spinning up at all either)
@dropbrick, can you post a screenshot of hackintool pci tab?
@Mateo1234454545 Any thoughts ?
@Mateo1234454545 Any thoughts ?
No ,same with mine.
@Mateo1234454545 Any thoughts ?
No ,same with mine.
Can I ask if you can measure a difference in the power used with and without NVMEFIX kext enabled ? Intel Power Gadget should be able to show you.
I have a ADATA XPG SX8200 pro 512gb and power consumption is very low , as with a standard ssd , with or without nvmefix.kext
I purchased an SSD at the end. Coconut battery went from 6.5W to around ~3.2W. Nvmefix did help a little when I tried it but it needs to be more aggressive I feel.
I also had the very high power consumption. Today I added alcid bootflag (=11) and now everything is perfect. AppleALC.kext was spinning kernel_task like crazy (got only 3h on battery):
Now power drain is ok:
@the-darkvoid: would it makes sense to add the bootflag to config.plist? Or is a different layout-id for different machine configuration required?
Update: I detected another battery intensive behavior: If i just touch my touchpad with one finger (doing nothing, just leave it there). CPU consumption fires to 15%... So I guess there is potential to optimise the touchpad drivers, too....
I also had the very high power consumption. Today I added alcid bootflag (=11) and now everything is perfect. AppleALC.kext was spinning kernel_task like crazy (got only 3h on battery):
Now power drain is ok:
@the-darkvoid: would it makes sense to add the bootflag to config.plist? Or is a different layout-id for different machine configuration required?
Update: I detected another battery intensive behavior: If i just touch my touchpad with one finger (doing nothing, just leave it there). CPU consumption fires to 15%... So I guess there is potential to optimise the touchpad drivers, too....
holy crap it was driving me nuts how fast it drain the batteryI thought my CPUfriend setup wasn't working right...this totally did the trick. Thank you 1000x .
@TaiPhamD: I'll have to test this layout in terms of compatibility with headset and HDMI audio. Have you encountered issues with either?
@TaiPhamD: I'll have to test this layout in terms of compatibility with headset and HDMI audio. Have you encountered issues with either?
i tried the headphone jack but it sounds like crap. I haven't install combo jack setup so not sure if that's an issue or not. I will report back later when i get the chance to try out the combo jack. Just using AirPods for now for headphones.
By the way how many watts are your guys computer draining on light browsing (using safari for example). Mine is about 10 W. I am getting an average of 5 hours battery life right now on 80% brightness.
I have a 9360 with the i7-7560U, 16GB RAM, SK Hynix 512GB SSD, QHD+ display.
I was expecting to get somewhere in the neighbourhood of 1GHz at 0.5 - 0.7w running Mojave (seen several people online get around this figure.)
Installed Mojave, using your Clover config. Checked power management with Intel Power Gadget; CPU idled between 1 and 1.3GHz, at 2.3 - 2.7w (nothing open except Power Gadget).
I generated my own CPUFriend kext, which appeared to do basically nothing. Tried turning off HWP, deleting CPUFriend etc., nothing improved the situation.
Did more research, suspected it might be my NVME drive. Swapped the SK Hynix drive out for a Samsung 960 Evo 250GB, reinstalled. Also applied ComboJack fix for audio. Now with my CPUFriend Kext, I idle at 0.9 - 1.1 GHz, at 1.7 - 1.8w. A minor improvement, but still a long way off of 0.5w.
Is there something obvious I'm missing? Any help much appreciated.