I'm running a Windows 7 SP1 x64 on a Dell Precision M4600 with i7-2720QM, 8 GB of RAM and 512 GB of Samsung 840 Pro SSD. Boot/shutdown times are quite fast.
However, hibernation and resume were always slow (30+ seconds) due to some misunderstanding between Microsoft and Dell, e.g., switching to low-power mode (1 CPU, reduced clock, etc.) before saving the hibernation file. Everything mentioned here was done with plugged-in power.
A few weeks ago my machine started to misbehave and resume now takes 4+ minutes. (Suspend is still about 30 seconds.) The hardware configuration hasn't changed, and I've run several test to check for hardware problems and to verify SSD read/write speeds.
I've already tried to disable and re-enable hibernation so that
hiberfil.sysgets deleted and re-created.
I've tried to make some boot traces - they only showed most of the resume time was spent reading from the disk.
powercfg -energyto check for problems. It listed "errors" like
The current power plan personality is High Performance when the system is plugged in.
just according to my intentions of preferring performance to battery life (plugged in?...).
The only useful piece information was
Platform Power Management Capabilities:PCI Express Active-State Power Management (ASPM) Disabled
PCI Express Active-State Power Management(ASPM) has been disabled due to a known incompatibility with the hardware in this computer.
but I cannot help it and I guess it was always disabled.
So does anyone have ideas to what to do/check?
I made some measurements based on Albin's idea. Hibernation and resume times (in seconds):
- clean, only Windows, 1.7 GB: 15/15
- default work apps running, 3 GB: 24/30
- SQL server service started, 3.2 GB: 24/30
- Firefox used for some time, 4.2 GB, 38/57
- working for a day, 6.7 GB, 54/193
- quitting Firefox, 2.8 GB, 32/29
- restoring Firefox, 5.8 GB, 49/181
#1: Hibernation time is directly proportional to the RAM used, as it should be, appr. 9 seconds for each GB.
#2: Resume time is not. It is much longer for a working state.
#3: It looks like if compressing the memory and writing it to the SSD would be faster than reading and decompressing. Which clearly should not be.
I haven't hit the several-minute resume yet (and kind of happy for the moment) but I'll check memory size in that case.