Linux's Bedtime Routine
Linux's Bedtime Routine
How does Linux move from an awake machine to a hibernating one? How does it then manage to restore all state? These questions led me to read way too much C in trying to figure out how this particular hardware/software boundary is navigated.
How does Linux move from an awake machine to a hibernating one? How does it then manage to restore all state? These questions led me to read way too much C in trying to figure out how this particular hardware/software boundary is navigated.
You're viewing a single thread.
This is certainly an interesting feature, though my one use case has become much less relevant now that systems boot so quickly.
Perhaps if you have long running jobs and no implementation of state saving it could find applications.
Boot times on AM5 are soooo slow due to some memory training feature of DDR-5, even after following many suggestions for settings. It appears to be a general issue with the platform,
so hibernation is very much back on the menu for me.Duh, it won't matter since the delay is before POST.
On Asus motherboards you can enable 'Memory Context Restore', and it'll remember the training. Unfortunately it seems rapid changes in the weather make my system unstable with it on.
I have an MSI motherboard. Memory Context Restore shaves significant time off of boots, but it is still extremely slow. Just a hang before I see POST complete.
hibernation is very much back on the menu for me
Uh, does hibernation on whatever you're using not actually power the system completely off?
I mean, my build does, so it's not really saving any "boot time", where by boot I mean the minute or two you spend POSTing.
....yeah, I'm an idiot. I hadn't thought very carefully about it yet. Won't help me since the delay is before POST.
I was hoping you had figured out a way around that, lol.
My AM5 platform takes so long to boot (and other weirdnesses) that the only nice thing I have to say about it is at least it's not Intel 13th/14th gen and literally melting.
Unfortunately not.
My AM5 has been pretty good, the boot issue notwithstanding. It has been quite stable at least. For me it's a 7600x.
Yeah I'm on my 2nd motherboard due to the first having issues where it just plain wouldn't post 75% of the time, and this 7700x is just.... sporadically unreliable.
It won't fail ANY kind of stress test I throw at it, but will then have notepad crashing repeatedly. I'm confused because it makes SO no sense and tired of dealing with it and it's on it's way to replacement as soon as there's benchmarks out for the Core Ultra stuff, and whatever Tim Apple does with the M4 Macs.
(Not really asking for diagnostic advice, as I've spent most of the last year diagnosing and replacing pieces and not being able to firmly blame the CPU or the PSU or the Motherboard or whatever since the failures are sporadic, and what doesn't work seems to change occasionally.)
Maybe you have just ended up with a lemon CPU. Though for random crashes like that, I'd almost always look to RAM first.
I did have some stability issues early on when trying to enable Expo. Never quite got that working right so it is currently disabled. I keep my 7600x in Eco mode since it is air cooled and the performance difference is not that great anyway, so I haven't noticed any major differences with Expo off.
The Expo issues were also with a very early MSI BIOS. I haven't tried it again after upgrading, but I probably should.
Yeah the goofy thing is it pass any length of memtest I care to toss at it, and will happily run prime95 forever with zero issues.
And then immediately have apps crash or freeze or otherwise misbehave.
Like, something is wrong, but nothing is actively broken, which is just.... annoying, heh.
The MSI board has been a source of less than enjoyable usage, but it's almost exclusively tied to the super super long POST times and the fact that, sometimes, it just... doesn't. Hard to know if the 90+ second wait is the normal 90 second wait, or if it's actually not going to turn on for some reason.
It's fine other that little quirk, at least as far as I can tell.
It volts up under load, maybe the problem is too little voltage at light loads.
Well, shit. This may have led me down a useful path.
So not voltage exactly, but load line calibration adjustments looks like it very much MAY have resolved the issue.
Or at least, I've been whacking at it with all the workloads that were unstable and crashing and so far it hasn't misbehaved at all.