Hibernate and Sleep not working with an L1 cache for system-boot NVME drive
Posted: Mon Dec 07, 2020 12:00 am
OK. It's "been awhile". When I last visited here as a forum member, I had just built my "super-system" and finished up with the Win 10 update (1703?) by mid-year 2017.
The System
Sabertooth Z170 motherboard
2x NVME drives: 1TB boot-system-disk; 256GB L2 caching disk; a 1TB SATA SSD and a 2TB media disk -- plus a 2TB backup disk for use with Macrium Reflect
RAM: 32 GB (8GB x 4) DDR4-3200 14-14-14-34 2T
NVidia (Gigabyte) GTX 1070 ("short-board")
That should do it for now. Primo-Cache was an intended early, integral part of this equation. I think I added half the RAM later -- to use 8GB caching the NVME boot disk, and another 8GB in two-tiered caching for the SSD (including the L2 256GB NVME drive).
I had been keen in building this system to assure that hibernation and sleep worked properly. Even so, the size of the hiberfil.sys is big, even if trimmed to minimal 50% of non-caching RAM. But I haven't used sleep and hibernate much for at least a year or more -- maybe two.
Few nights ago, I tried to test hibernate. It goes into hibernate, but must reboot (failing hibernate) when you wake it up. I identified the bug-check error code as 0x000000a0. Quick web-search seemed to show promise that this was a software or driver conflict. I'm pretty sure it's not a hardware failure. Testing the sleep state, the very same thing happens: computer goes to sleep, then reboots itself waking up.
It didn't take long for me to consider Primo-Cache as related to the problem. BlueScreenView seems to indicate storport.sys microsoft driver, and (maybe) a timing problem. I have yet to update that driver, if it's at all possible.
The Caching
Again -- 8GB RAM for L1 of the 1TB NVME drive; 8GB RAM for L1 of the SSD. L2 is implemented for the SSD -- so it's a two-tier setup. I decided to delete the L1 cache for the boot-system NVME.
That solved the problem, at least for sleep. I can test hibernate later, but at the moment it's disabled.
Can anyone tell me how this happens? Or whether others have noticed it? Or whether my original configuration makes it impossible for hibernate/sleep to work properly? Or whether this is something that arose with more recent major Windows updates?
Truth be told, I don't need to cache the main NVME drive, since it's very fast. I don't need a benchmark which shows 19,000 MB/s throughput as opposed to the drive's touted 2,000-plus. But the caching works great as long as I only shut the machine down and start it up again. It apparently doesn't work for sleep and hibernate.
The System
Sabertooth Z170 motherboard
2x NVME drives: 1TB boot-system-disk; 256GB L2 caching disk; a 1TB SATA SSD and a 2TB media disk -- plus a 2TB backup disk for use with Macrium Reflect
RAM: 32 GB (8GB x 4) DDR4-3200 14-14-14-34 2T
NVidia (Gigabyte) GTX 1070 ("short-board")
That should do it for now. Primo-Cache was an intended early, integral part of this equation. I think I added half the RAM later -- to use 8GB caching the NVME boot disk, and another 8GB in two-tiered caching for the SSD (including the L2 256GB NVME drive).
I had been keen in building this system to assure that hibernation and sleep worked properly. Even so, the size of the hiberfil.sys is big, even if trimmed to minimal 50% of non-caching RAM. But I haven't used sleep and hibernate much for at least a year or more -- maybe two.
Few nights ago, I tried to test hibernate. It goes into hibernate, but must reboot (failing hibernate) when you wake it up. I identified the bug-check error code as 0x000000a0. Quick web-search seemed to show promise that this was a software or driver conflict. I'm pretty sure it's not a hardware failure. Testing the sleep state, the very same thing happens: computer goes to sleep, then reboots itself waking up.
It didn't take long for me to consider Primo-Cache as related to the problem. BlueScreenView seems to indicate storport.sys microsoft driver, and (maybe) a timing problem. I have yet to update that driver, if it's at all possible.
The Caching
Again -- 8GB RAM for L1 of the 1TB NVME drive; 8GB RAM for L1 of the SSD. L2 is implemented for the SSD -- so it's a two-tier setup. I decided to delete the L1 cache for the boot-system NVME.
That solved the problem, at least for sleep. I can test hibernate later, but at the moment it's disabled.
Can anyone tell me how this happens? Or whether others have noticed it? Or whether my original configuration makes it impossible for hibernate/sleep to work properly? Or whether this is something that arose with more recent major Windows updates?
Truth be told, I don't need to cache the main NVME drive, since it's very fast. I don't need a benchmark which shows 19,000 MB/s throughput as opposed to the drive's touted 2,000-plus. But the caching works great as long as I only shut the machine down and start it up again. It apparently doesn't work for sleep and hibernate.