MP 1,1-5,1 Classic Mac Pro, Hibernatemode 25, FileVault and GPUs without Bootscreen

[Macschrauber]

As the electricity bill gets higher and higher I decided to check out Hibernatemode 25 to lower standby wattage.

Hibernatemode 25 stores the ram content to /var/vm/sleepimage and sets the Mac Pro in a semi off state. Front LED is off, USB is off and one can wake the Mac only with the power button.

It takes a bit longer but it draws less current in that state. Also it should not be problematic when power cuts during sleep.


Some more details:

How to Sleep a Mac and Check Sleep Settings

Mac supports multiple methods of sleep. Some can extend battery life, while others can make entering and leaving sleep faster.

www.lifewire.com

Downsides:
- in my tests Hibernatemode 25 does not work with Big Sur and OpenCore. Kernel Panic into EFI (tested with OCLP 0.51 and BigSur 12.6)
- takes longer to get to sleep and get back on
- takes capacity of the system disk
- without a bootscreen gpu you can't see the status bar when it comes back

why I made the thread: If one has FileVault active (I have it running Mojave) you will _not_ see the FileVault dialog without a bootscreen GPU.

So the master question is: can one enter a FileVault password blindly without a GPU what runs the EFI GUI for FileVault?

 

[Macschrauber]

I tried it after a full backup.

This machine has not got to full standby. display was off, Front LED stayed lit and after waking it with the keyboard it restarted.

No kernel panic, but didn't worked. Am not sure why, I keep on investigating with other Mac Pros what I have on the bench. But as this is my productive machine not with this before I know more details.

My Mac Pro 3,1 with Dosdude Mojave did the full circle (standby, FileVault dialog, back on) in hibernatemode 25, but it has a bootscreen gpu, no NVMe blade and no USB 3 PCIe card.

 

[cdf]

So the master question is: can one enter a FileVault password blindly without a GPU what runs the EFI GUI for FileVault?

Unfortunately, if the FileVault login is anything like the native boot picker, then it apparently won’t even start without a Mac-EFI graphics card, making “blind” entry useless.

 

[Macschrauber]

Ok, got it.

Side note: of course I have a FileVault login without an EFI GPU (as the machine would not start without when using FileVault. I get it via RefindPlus.

But waking after Hibernatemode 25 is another EFI firmware thing.

 

[cdf]

Side note: of course I have a FileVault login without an EFI GPU (as the machine would not start without when using FileVault. I get it via RefindPlus.

But waking after Hibernatemode 25 is another EFI firmware thing.

Indeed. OpenCore is also able to display the FileVault login on boot without a Mac-EFI GPU (it should be able to do the same with the native boot picker, but currently can’t, so it seems that the FileVault login is less stubborn than the native boot picker!).

Note that OpenCore provides some support for hibernation, so it could be a matter of configuration to get Hibernatemode 25 to work…

 

[Macschrauber]

I will do some more tests with my test rig next week. If FileVault stops it I might think about another encryption strategy.

 

[arw]

I hope it‘s not too off-topic (as no FileFault is involved) but:
Has anyone recently tried to move the sleepimage with “sudo pmset -a hibernatefile“ to a non-root (SATA or USB) volume?
hibernatemode 25 already implies inactive pages are aggressively paged out but with lots of RAM the size is still considerable.
My question is purely motivated by SSD wear and not limited to OpenCore or the MP5,1.
The manual says the hibernatefile may only lie on the root volume but it is reported to work on an internal SD card-reader.

 

[Macschrauber]

I hope it‘s not too off-topic (as no FileFault is involved) but:
Has anyone recently tried to move the sleepimage with “sudo pmset -a hibernatefile“ to a non-root (SATA or USB) volume?
hibernatemode 25 already implies inactive pages are aggressively paged out but with lots of RAM the size is still considerable.
My question is purely motivated by SSD wear and not limited to OpenCore or the MP5,1.
The manual says the hibernatefile may only lie on the root volume but it is reported to work on an internal SD card-reader.

This is one of the things I have to test. As my main machine has 48GBs I needed to free some space on the boot nvme as well.

I'd configure the hibernatefile on my machine to a spinner as well if that works. No problem in such a box as it is filled with internal drives.

Just got an idea: my nvme reports to the system as external (running unpatched Mojave). Maybe a data point.

 

[Macschrauber]

I tested a few things with the location of the hibernatefile on Hibernatemode 25. I got limited sources this evening.

MP5.1 EL Capitan:
Hibernatefile to an USB Stick: Hibernatemode 0 instead of 25, wakeup ok but "normal" sleep with pulsing front LED.
Hibernatefile to another partition of the same physical boot drive: Hibernatemode 25 ok

MP3.1 Mojave:
Hibernatefile to another spinner on same onboard Sata bus: entered Hibernatemode 25 but restarted after waking up
Hibernatefile back to /var/vm/sleepimage: Hibernatemode25 ok

 

[Macschrauber]

Today I tested another thing with another machine I had on the bench.

Mac Pro 5,1
OCLP 0.51
Monterey 12.6.1
RX560 without EFI Bootscreen
Inateck USB 3 Card
Sata Startup disk
Hibernatemode 25

- set Mac to sleep, mac slept, LED off, can't get it to wake up by USB Keyboard (what to be expected)
- woke up by power button
- OpenCore loaded (what I not expected) and displayed the usual gui
- I selected Monterey
- Got status bar when hibernatefile loaded
- Mac was awake and all was like expected


---

Can't get my main Mac Pro 5,1 in that standby mode.
Mac Pro 5,1
RefindPlus (so boots "natively")
Mojave
RX560 without EFI Bootscreen
Inateck USB 3 Card
NVME startup disk
Hibernatemode 25

I turned FileVault off, now the machine does not restart, but does not really hibernate as well

- set Mac to sleep, mac slept, LED off, but can get it to wake up by USB Keyboard (what is not to be expected with 25)
- no hibernatefile status bar, Mac woke up immediately like with Hibernatemode 0

 

[Macschrauber]

did some more tests with the Mac Pro on the bench (what did standby correctly)

- added an Apple AHCI Blade with Monterey on it and booted it.
- tried Hibernatemode 25 and got the same behaviour as my main machine (acted like mode 0)

- booted Monterey on Sata SSD, 25 worked. AHCI Blade showed as external
- run OCLP again to get AHCI Blade internal

- booted Apple AHCI Blade again and now 25 worked on the Blade


conclusion: Seems I need to patch my Mojave system on my main machine to get the NVME Blade shown as internal. Hibernatemode 25 might not work on a boot volume that is shown as external.

What is logical, one can pull the drive with the sleep image on it. But they never thought of internal drives shown as external.



edit:

added Innie (1.3) kext on native botting Mojave on AHCI Blade and it appeared as internal - and Hibernate 25 worked as expected.

So Innie helps for Hibernatemode 25 on bootvolumes what appear(ed) as external as well.

 

[Macschrauber]

Side note:

set up a Mac Pro 5.1 with Montery 12.6.1, OCLP 0.52 and RX560 without EFI Bootscreen

System was on a Sata SSD on Southbridge bus (usual Sata port on an adapter sled)


System went to full standy, woke up by power button, got the status bar for loading the hibernatefile.


from OCLP 0.5.2 changelog: "Hide OpenCore Boot Picker when waking from hibernation"

so perfect behaviour, exactly same as with an EFI Bootscreen GPU.

 

[mattspace]

Hibernatemode 25 stores the ram content to /var/vm/sleepimage and sets the Mac Pro in a semi off state. Front LED is off, USB is off and one can wake the Mac only with the power button.

You know, I'm wondering if that would solve crash on wake / wake failures that seem to be triggered by usb devices...

 

[Macschrauber]

You know, I'm wondering if that would solve crash on wake / wake failures that seem to be triggered by usb devices...

Try it…

 

[mattspace]

hmm, not a viable solution - machine comes back up with warnings about improper unmounting for all my external drives

 

[Macschrauber]

hmm, not a viable solution - machine comes back up with warnings about improper unmounting for all my external drives

Is this with Hibernatemode 0 the same?

You know the silly usb3 wake bug about unmounting is quite normal after sleep?

 

[mattspace]

Is this with Hibernatemode 0 the same?

You know the silly usb3 wake bug about unmounting is quite normal after sleep?

No, it never does it with hibernate 0. I'm going to try hibernate 3 and see what happens, but the drives are all on a powered usb hub.

 

[Macschrauber]

No, it never does it with hibernate 0. I'm going to try hibernate 3 and see what happens, but the drives are all on a powered usb hub.

The Hub is maybe the cause. In full standby by hm25 the Mac is completely off. Thats the cause for wanting hm25. Way less power draw when set to sleep.

 

[Macschrauber]

I tested something and found an interesting detail:

My Mac Pro 3.1 (what does correct Hibernatemode 25, goes to full power off standby) does not power off when I have my 2,5" USB WD Elements Drive in an USB Port (tested front port). It leaves USB on and does not hibernate.

Easy test: when it wakes up by USB (Mouse or Keyboard) it is not in full Hibernate power off mode.

Even if I had the Drive in, eject and physically disconnect it that prevents Hibernation. Had to restart to get full hibernation back.

 

[Macschrauber]

Some Data Points (measured with a -not-so-accurate-power-meter)

the test box I have here on the bench:

Mac Pro 5,1
2 Spinners
USB3 Card
GT640
4x4GB 10600R Ram
X5660 (2.8 6core)


Power off (16 Watts, turns on just with power button)
Hibernatemode 25 (16 Watts, same as power off, front LED off, wake up just with power button)
Hiibernatemode 0 (18 Watts, sleep, front LED "snorching", wake up with USB)

2 watts difference is not too much but as it depends on the power bill it could be a pizza / year.


Guess with more Ram (banks) it is way more. Also my power meter is a cheap consumer product what have its problems with pulse current what a smps draws in standby.


side note: this box does not go to hm25 with Monterey and OCLP 0.52, (even if other test boxes did), but with Mavericks. Don't tested other systems as I had those two handy.

 

[startergo]

@Macschrauber Are you using hibernationfixup kext and have you enabled it in NVRAM or boot arguments?

 

[mattspace]

2 watts difference is not too much but as it depends on the power bill it could be a pizza / year.

To put that in perspective, an electric kettle is around 2000w (per hour) minimum, and around 4 mins to boil when full, there's 133w alone.

 

[Macschrauber]

To put that in perspective, an electric kettle is around 2000w (per hour) minimum, and around 4 mins to boil when full, there's 133w alone.

It depends how you see it. The Mac Pro is not really a power saving device :-] but if you can save a few watts with just one setting (often it works) that will make a small profit.

1 Watt 24/7 is about 5$ a year in my area - and the prices are rising. I also bet that the difference is more on a Dual System with full Ram and more stuff in the machine. Also my power meter is not accurate for the pulsing current smps draw in standby.

 

[Macschrauber]

@Macschrauber Are you using hibernationfixup kext and have you enabled it in NVRAM or boot arguments?

On the test box with Monterey I used OCLP standard settings with little modifications for chainloading. Will check if they add hibernationfixup.kext.

On my main machine I use Mojave. NVMe fixup looks promising. Will try that - and the kext. NVMe against an Apple AHCI blade are the main differences between my machine and the one I tested against.