Linux 5.10 Will Be Able To Hibernate + Resume Much Faster

[polarathene]

Do any of those chinese SSDs have 500+GB?
And do you think that they will continue to work well in 7 years?

Those are two considerations that are important enough to deter me from buying those SSDs.
I do intend to buy one of those SSDs in the future, but only for lab tests that will require them, not for personal use.

Keep in mind you're discussing purchasing one from a chinese vendor for a lower price, that normally entails some tradeoffs.

I was in China in 2018 and had a desktop purchased off TaoBao there, arrived fairly quickly, definitely wasn't perfect but the 128GB SSD from Colorful I took back with me and that continues to work fairly well. I'm sure they sell 500GB+ models too, but I got the whole thing for 300-600USD or something, 16GB DDR4 with a GTX-1060 6GB GPU and i5-8400 6 core CPU (the marketing in china is kinda scary, there's "new" systems being sold with 1st gen i7's but relying on customers to only recognize the i7 branding).

Once I bought an Acer. There won't be a twice.
It was not for firmware reasons but for quality reasons. It was an expensive piece of crap (1500€) that had all sorts of physical issues within 3 years of very careful use, the worst being the hinges forcing on and crackling the body back cover after only 6 months.

When was that purchase? The hinge on my budget Acer is pretty decent, no issues so far. Although I'm not using it daily, the lid "lip" for gripping and raising the lid though is a bit of a hassle IIRC, needs a bit of force to open vs more premium models that can use a single finger.

A single bad experience is understandable for wanting to avoid the risk of going through it again in future, but it's not always going to remain the case. Did you do much research on the product before the purchase? What was the main reason you chose the Acer at the time? What did you end up going with afterwards?

[Sonadow]

Keep in mind you're discussing purchasing one from a chinese vendor for a lower price, that normally entails some tradeoffs.

I was in China in 2018 and had a desktop purchased off TaoBao there, arrived fairly quickly, definitely wasn't perfect but the 128GB SSD from Colorful I took back with me and that continues to work fairly well. I'm sure they sell 500GB+ models too, but I got the whole thing for 300-600USD or something, 16GB DDR4 with a GTX-1060 6GB GPU and i5-8400 6 core CPU (the marketing in china is kinda scary, there's "new" systems being sold with 1st gen i7's but relying on customers to only recognize the i7 branding).

In their defense, they always publish the processor model number clearly in the item listing, so the consumer has got to be really daft if they don't check the specs clearly.

When was that purchase? The hinge on my budget Acer is pretty decent, no issues so far. Although I'm not using it daily, the lid "lip" for gripping and raising the lid though is a bit of a hassle IIRC, needs a bit of force to open vs more premium models that can use a single finger.

A single bad experience is understandable for wanting to avoid the risk of going through it again in future, but it's not always going to remain the case. Did you do much research on the product before the purchase? What was the main reason you chose the Acer at the time? What did you end up going with afterwards?

Chiming in with my experience on Acers.

I had three Acers. Two were bought in 2008, and one bought in 2015. The build quality is really quite mediocre. Both units from 2008 were Aspire Gemstones and managed to crack at the palmrest area. I have no idea how that was even possible. They also developed severe heat issues after using them for about six years, and keep shutting down after some time. Replacing the thermal paste did not solve the problem.

The Aspire V3 I bought in 2015 had a hinge crack after two years of use and the battery swelled to a point where it was pushing out the trackpad. I had to replace the battery on my own dime. Finally in 2018 the bottom shell cracked and three of the screw sockets broke off. And mind you, this is a laptop I was very gentle with.

I doubt I will ever get another Acer again. They were good laptops for the budget conscious back then but heck, the small Chinese OEMs can do a much better job these days at a lower price. my $500 Chinese laptops with metal bodies have better build quality than the Acers.

[oiaohm]

No, I really didn't. You're welcome to show me a product within that budget I paid, that is comparable. Anything that was available at the time I purchased for the price it would be at that time, you'd find would be making notable sacrifices. Legit all I have to do is hibernate, which is what I have shutting the lid set to do, takes barely any time to hibernate or resume. The graphical issues was due to the freshness of the hardware at the time, had nothing to do with Acer, any laptop from any vendor was having similar problems until newer kernel updates/releases and drivers were available, some dependent on Mesa IIRC.

https://laptopunderbudget.com/best-laptops-under-400-dollars/#L6

Yes there are 2 dell systems in the 400 dollar price bracket these basically work with Linux from the get go both are on the Ubuntu certified list with all drivers in mainline kernels.

HP amd laptop is latter but they waited to release a product that was somewhere near solid.

https://www.bloovis.com/2020/05/27/mint-19-hp-notebook.html

Current amd hp turns out to have a secureboot caused issue with a non mainline driver. There is a reason not to be a very early adopter.

There is more to acers when you have been servicing them. Updating firmware on acer can change fuseable links meaning you cannot roll the firmware back.

I am pretty sure the Acer can do a reset to factory UEFI if needed.

This is false idea. Acer is the only laptop brand in the Windows/Linux laptop space where you cannot always reinstall the factory UEFI this is also why they put so many warning on updating your firmware at all because they don't promise you that updating firmware is reservable when all other vendors in the Windows/Linux laptop space do. People doing reviews don't both checking does maker provide sane firmware updating process this is how people end up with different lemon acer laptops as the reviews say they are good they do a firmware update and find it non reverseable so screwed.

Its quite a trade off being on a fairly new platform using a model from a company that does not have solid firmware replacement system allowing you to change to newer and older and even more more important back up what the laptop has on it before you change the firmware. Acer firmware policies can leave with a bricked device that Acer themselves cannot fix.

Remember just because something is a lemon design does not mean it will fail on you. Yes your idea of only updating firmware when you have to something you need to be very much more careful with a acer with as you may not be able to reverse the update to firmware. This bad policy is something reviewers really need to put pressure on Acer to change and fix. 10 years ago Acer did not have this current stupidity.

[Old Grouch]

I would just like to point out that not all Lenovo branded notepad PCs are easy to firmware upgrade with FLOSS software. Thinkpads are getting better, but the cheaper IdeaPads do not all offer UEFI or FreeDOS firmware updaters, and are not on LVFS/fwupd. Rather than reinstall Windows, I created a WinPE USB thumbdrive, which does the job. Finding out how to do that, I leave as a Student Exercise. I have an Acer notepad PC with similar constraints.

I suspect there is commercial pressure on the OEMs to make non-Windows alternatives difficult, if not impossible. The LVFS project needs as much support as possible

[Veerappan]

I'm in the same position as Mez' - my HTPC gets a kernel update MAYBE once a quarter at most, and goes through literally hundreds of suspend/resume cycles between each reboot, with no problems at all, and has done so for what must be close to 8 years now.

Hibernate is one of those things that I have almost no interest in. Despite the apparently popularity of it among DEs, many of which now not only make it the default but make it hard to turn the damn thing off, it's a pretty bad way to handle the pauses between sessions: it's slow, and will remain slow even with this patch; and for most users the vast majority of the data they're hibernating is cold to begin with.

On a laptop, there's almost no merit to using it over suspend. 2 seconds plus 0.001% battery drain, vs a minute or more plus a day's worth of writes to the drive? I'll take suspend every time, thanks. Same on a desktop, as long as you have a UPS.

IMO, hibernate really only has value when you have a seriously "busy" system (lots of apps / VMs / containers / whatever) AND that system is in an environment where the power can't be trusted. Other than that, it's just a really bad way of solving a problem that was already long-since solved.

Since the slowness of hibernate is one of its two big negatives though, this patch does represent a huge improvement. I still don't have any use for it, but it'll suck a whole lot less for the people who do, even if many of them would still probably be better off not using it in the first place.

I bought a new laptop this summer (HP Envy x360 13", Ryzen 4700U). Works wonderfully in Linux, except for the fact that the ACPI tables don't advertise S3 suspend-to-ram support. Supposedly the AMD devs are working platform-level connected-sleep for the Renoir platform, but it's not ready yet (HP Envy x360 w/ 4700U failing to suspend).

At this point, I can either shut down my laptop every time before closing its lid, with the side effect of it trying to suspend, failing because the ACPI table doesn't support it, and then the machine hanging if I leave the lid closed for more than a few minutes.... or I can enable hibernation. I did that.

At least now, I can just shut the lid, throw it in a bag (or carry it up/downstairs) and know that when I open it up and hit the power button again, I can at least continue with whatever I was working on.

[carewolf]

Can those be used to also speed up swapoff and recovery from swap usage? Because those are suspiciously slow as well, always felt something was off with the disk to memory pipeline of memory pages.

[arQon]

... or your laptop draws a lot of battery in suspend
... or you don't use the laptop very often between uses

Phoronix tends to be myopic when it comes to "reasons".

Or alternatively, Phoronix tends to be creative when it comes to them. Yours certainly are. (And I mean that in a positive way).

If your laptop "draws a lot of battery in suspend", it isn't suspended. Not saying that's YOUR fault, but suspend is by definition a (very, very) low-power state. A "high power draw" suspend is implicitly not suspend in the first place. That may well be a good reason to use hibernate, but, like Veerappan's situation, it's because something is broken. That's not a commentary on the relative value of the two options, it's "I HAVE to use X, because Y doesn't work".

"... or you don't use the laptop very often between uses" is just the same argument, restated. The power draw for my DESKTOP in suspend is literally immeasurably low for me: below the 10W granularity my UPS can show. (IDK where my multimeter is). When I had a work laptop, I'd suspend that on Friday evening and leave it alone until Monday morning, and the battery wouldn't even notice. If suspend isn't broken per your first point, no sane amount of time between uses has meaningful drain on the battery to begin with.

So yeah, maybe "don't use it often between uses" (confusing wording, but I know what you meant) is valid: but are you really suggesting that you would have enough ongoing processes that you "need" to use hibernate rather than just shutting the machine down, but somehow don't touch those processes for weeks at a time? That seems a bit of a stretch, but either way, artificial scenarios aren't really much of an argument against a general statement.
The goal here isn't actually a challenge to see who can contrive a scenario where hibernate does actually make sense, it's "What's the better option for 99% of use cases?", and hibernate is not it.

Semi-related, there are some interesting notes over here https://lifehacker.com/how-much-batt...-drain-5526542 where doing a full shutdown and restart can apparently burn MORE power on some laptops than just suspending, which I thought was interesting. I suspect mechanical drives are the killer there, but interesting nonetheless.

[kcrudup]

If your laptop "draws a lot of battery in suspend", it isn't suspended

Apparently 32GB of LPDDR4(x?) requires a fair amount of standby current, as if I leave it for 3 days I'm down to ~50% (Dell XPS 2-in-1 7390), but of course that implies the BIOS hasn't left something running in S3 (I manually block out XHCI and rfkill the radios on suspend, too).

[arQon]

Apparently 32GB of LPDDR4(x?) requires a fair amount of standby current, as if I leave it for 3 days I'm down to ~50% (Dell XPS 2-in-1 7390), but of course that implies the BIOS hasn't left something running in S3 (I manually block out XHCI and rfkill the radios on suspend, too).

Yeah, that would be my guess too: some piece isn't actually suspending. Funnily enough, I saw an ad yesterday for some of this "always connected" BS that Intel and MS are pushing now, and realised that of course there are going to be times that even the current IME might well just ignore suspend, because, hey, it's the IME and what YOU want doesn't matter. sigh.

I think your manual rfkill is an excellent idea, incidentally. I'm sure the IME will happily ignore it at will, but I can't think of a reason why it would try to drive the radios since it certainly shouldn't expect to be able to connect to any APs without user help. It WOULD obviously burn a ton of power if it supported WOWL etc, but it seems unlikely you'd have that enabled.

RAM isn't my thing, and standby numbers for LPDDR4 are strangely hard to find, but according to this validation document https://www.electronicdesign.com/tec...dr4-and-lpddr4 it actually draws NO current when not driven. Even without that aspect though, it's very hard to imagine it burning any meaningful amount of power in standby when even non-LP DDR3 doesn't.

So yeah, I think there's definitely something fishy going on with your laptop, and I think you've established what it is in a general sense: SOMETHING is ignoring the suspend and continuing to bleed power. Very annoying of it, and since you'll doubtless have already done all that CAN be done I guess it's likely to continue to do so forever. IME is certainly the most likely culprit, I'd imagine: especially "vPro" or whatever stupid marketing name it's carrying this week. But you have no control over that, so, bleh.

[Raka555]

I don't need it faster, I need it working.
On Kubuntu 20.04+ the whole Hibernate menu entry is missing and I tried once to jump through hoops to enable it, but I managed only to get the menu entry showing, but it was not hibernating.
I guess Ubuntu devs are too busy to change themes and wallpapers on every release instead of fixing important stuff like this.

I used suspend to RAM for years with various versions of Mint/Ubuntu and KDE neon.
It always was like a lucky packet to open the lid of my laptop.
I recently installed Kubuntu 20.04 and now I get a solid reboot every time I open the lid

So, I am also in the "I need it working first" class.