EEVblog Electronics Community Forum
Products => Computers => Topic started by: SiliconWizard on June 28, 2021, 06:01:07 pm
-
Just a general question here... Do you ever put your desktop computers in standby mode? Do you do this often, and if so, is that flawless or do you occasionally get problems when resuming?
For many years, I used never to use the standby mode for desktop computers (only for laptops.)
Standby was often poorly supported on desktop motherboards/some additional cards/etc.
Things have evolved though. Now in order to avoid wasting a few minutes every day powering on my workstations (POST/boot/open session/launch all programs that I'm currently using/opening documents/etc...), I have taken the habit of putting them in standby. And only reboot or power-cycle them every once in a while.
Whereas that has mostly worked fine, I have encountered an issue: processor cooling would occasionally NOT start upon resuming, leading to overheating. You then need to close your apps pretty fast and shut the computer down, because while overheating, the CPU throttles down to a point the OS becomes almost unresponsive...
So anyway, any experience with this? What are your thoughts?
-
Always on, just use DPMS for displays. In particular, I ensure that the HDDs never spin down.
-
Haven’t had any problems with this on desktops for a few years on windows to be honest. I did have one issue which turned out to be a Radeon GPU issue though which would stop it posting when it rebooted.
Linux on the other hand is a complete ball bag randomly.
-
I've had a lot more success with Linux than Windows when it comes to power management and reliability. I suspect it boils down to specific hardware configurations and the quality of the drivers.
-
As I said, it has mostly worked well for me, but the only problem I've encountered was with processor cooling on one workstation (watercooling.)
Upon resume, when the issue arises, the pump just doesn't restart. But it's only occasional. Most resume events go OK, so this is not a systematic issue. I suppose the power to the pump does occasionally remain switched off when this happens. Since it's random, I can't pinpoint the issue - I guess it's a BIOS-specific problem.
I could of course power the pump through a power rail that is always ON, but I don't like the idea of the pump constantly on when the computer is in standby.
-
No issues here and i have a liquid cooled CPU. I however mostly use hibernation than sleep.
-
In the early PC days standby was tricky on desktops, but it should now work Ok on most systems. I use it quite a lot and no problem so far.
There could be difficulties with some bias / graphics combinations.
An occasional near standstill happens when windows is downloading an update - it does not need an hardware fault or standby for this and it also effects notebooks.
-
No issues here and i have a liquid cooled CPU. I however mostly use hibernation than sleep.
The issue I have is clearly due to a resume from standby. Hibernation should IMO not exhibit this issue. I just don't use this because I've had mixed experiences with it. And, it requires a pretty large file (I have 64GB of RAM), something I try to avoid because space on SSDs is still kind of at a premium...
-
Invariably for me, standby mode has always resulted in breaking things requiring a restart, so I disable it. Hibernation has been even worse.
What I do do now though is use the power button to put the monitors in standby.
-
I sometimes put my desktop to sleep or hibernate overnight if I want all of my work to still be open the next morning. It's quieter and probably saves me a little bit of electricity.
Suspend support:
Laptops (Linux): seem to work perfectly. I've only ever used laptops with Intel graphics (no discrete GPUs).
Desktops (Linux): hit and miss. Used to work fine, but the last time I tried I got a black screen on resume (Radeon RX570).
I use hibernate on my desktop now, that works much more reliably. It is (after all) essentially a normal cold boot, but then the kernel goes "oh hey there's an image of RAM and other stuff sitting in the swap area, let's use this!" rather than continuing with a normal boot. Probably still requires some extra driver magic for initialising stuff like peripherals.
-
Just a general question here... Do you ever put your desktop computers in standby mode? Do you do this often, and if so, is that flawless or do you occasionally get problems when resuming?
Yes, I am using standby all the time, in order to preserve for the next session exactly were I left last time. Always the STR mode (Suspend To RAM) not hibernate. It's been about 10 years since the desktops are doing standby and wakups correctly, just like the laptops. It might depend from one hardware to another, drivers, OS, but I didn't have any problems since about the Win98/ XP era.
Now I am using Kubuntu 20.04 LTS/i7, but it worked just fine some years ago with Win10, too.
-
I use standby all the time with with Windoze 10. Never had any issues with it.
-
I never use any of that stuff, just leave the machines on all the time. I do set spinning HDDs to turn off after 20 mins or so and monitors to power down after an hour.
-
usually hibernate , not standby,
due to standby\sleep mode still depend on a power, if it off , you lost unsaved process
-
I do set spinning HDDs to turn off after 20 mins or so and monitors to power down after an hour.
My monitors turn off after about 10-15 minutes (depends on the system) and have been working fine for a decade or longer.
On the other hand, the HDDs never spin down. Their life spans are a lot more variable, ranging from about 4 years to about 15 years. HDD lifespan seems to be model specific, but 3.5" HDDs failures mostly coincide with power downs.
-
AFAIK nothing is lost at suspend to RAM, or any other type of standby. If it were to be so, then wakeup wouldn't work.
Hibernate also works fine in desktops, just that it is unpractical for the nowadays sizes of RAM. To write on a HDD, say 32 GB of RAM, like I have now on the desktop, will take a while. To write 32 GB on a SSD at each standby, will wear out the SSD pretty fast.
-
Yes, I was using standby and hibernate on my work PC for years, with windows 7 and 10. No issues, except sometimes update would restart it in the middle of the night, and unsaved data was gone,but honestly, this is my mistake.
I do that because then it wakes up in 3 seconds instead of 40. That's like 2 hours in a year. Or get a coffe.
-
AFAIK nothing is lost at suspend to RAM, or any other type of standby. If it were to be so, then wakeup wouldn't work.
I believe Gigajoe is referring to when your battery dies when you are in suspend. I've had that happen a few times (plugged the charger in but forgot to turn it on!).
Hibernate also works fine in desktops, just that it is unpractical for the nowadays sizes of RAM. To write on a HDD, say 32 GB of RAM, like I have now on the desktop, will take a while. To write 32 GB on a SSD at each standby, will wear out the SSD pretty fast.
The size of your RAM doesn't matter, it's how much RAM you are using at the time of hibernation. Often I want to hibernate when I only have Kicad + Firefox open (less than a few GB), this would take the same amount of memory & time even if I had only 4GB of RAM in my machine.
If I'm running VMs then my memory usage is way up, but I don't tend to hibernate with those open.
I know that hibernation compression is an option for Linux, I wonder if it's by default and whether or not Windows does it. It might make sense: if you're going to try and write GB of data from RAM to disk then you have an eternity for the CPU to sit around and do very little other than buffer-shuffle data. Almost all programs keep big zeroed chunks of memory around, may as well use the CPU to do some simple compression.
-
Yes, it depends of the usage, though modern OS's take advantage of the available RAM as much as possible.
I'm using most of it. Right now, after only a couple of hours from a cold boot is in use about half of it, and I only browse with Firefox, have a spreadsheet open, a texteditor and watch a YouTube talk in the background.
Apart from the OS, ZFS takes advantage of the RAM, too, when it's available.
-
Check how much of that is disk cache. Disk cache should get flushed (writes) or discarded (reads) before hibernating, as it's useless writing disk cache temporarily to disk.
$ free -m
total used free shared buff/cache available
Mem: 11962 1401 7005 41 3555 10235
Swap: 6143 0 6143
-
AFAIK nothing is lost at suspend to RAM, or any other type of standby. If it were to be so, then wakeup wouldn't work.
So what is wrong when wakeup fails? State was lost.
Typically network drives are not reattached, network connections are completely lost, or USB devices vanish, but I have seen just about everything fail at one time or another.
The size of your RAM doesn't matter, it's how much RAM you are using at the time of hibernation. Often I want to hibernate when I only have Kicad + Firefox open (less than a few GB), this would take the same amount of memory & time even if I had only 4GB of RAM in my machine.
It matters when waking up takes longer than a cold start.
-
Mine looks like this
free -m
total used free shared buff/cache available
Mem: 32002 7422 19417 1001 5161 23172
Swap: 2047 0 2047
Another thing why I don't like about hibernation (apart from taking too long to load/unload the RAM) is that it needs hibernation space as big as the entire RAM. My Kubuntu 20.04 LTS lives on an 160 GB.
lsblk /dev/sdd
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sdd 8:48 0 149.1G 0 disk
├─sdd1 8:49 0 2M 0 part
├─sdd2 8:50 0 128M 0 part /boot/efi
├─sdd3 8:51 0 512M 0 part
├─sdd4 8:52 0 34.3G 0 part
└─sdd5 8:53 0 114.1G 0 part /
The 34.3G partition is for a Gentoo install, the swap is the small 512M one. The only place where I wouldn't care about locking out of use 32GB of disk would be on another disk. I want to keep that one as a ZFS storage place only.
But most of all I didn't like the time it takes to go in and out of hibernation. I've tested once a couple of years ago and found that to be very slow, don't know the numbers but in the range of couple of minutes, while STR standby is almost instant, faster than it takes to spin up the HDDs or than the monitor could wake up, so I disabled hibernation.
Maybe on a SSD hibernation would be faster, but still not as STR, and writing on a SSD about 50GB for 2-3 hibernation daily would mean to wear it out probably 10 to 100 times faster than it will wear without hibernate.
To me, hibernation is just too slow.
OTOH, the main drawback for STR would be that the RAM remains powered and still has to be continuously refreshed even during suspend, therefore some current consumption over night, but this is insignificant when compared with, let's say, a refrigerator or with other white goods.
Another drawback for STR is that there is a risk of power outage during suspend to RAM. It happened a few times over the years, but both ext4 and ZFS I use are very good at coping with that, they have some sort of journaling and they always recovered by themselves after power outages no matter if it was an outage while working or while in standby.
I'll take these drawbacks for a faster in/out of standby.
-
Invariably for me, standby mode has always resulted in breaking things requiring a restart, so I disable it. Hibernation has been even worse.
What I do do now though is use the power button to put the monitors in standby.
Looks like I'm not having it too bad, then. Mostly... because I haven't had much issues other than this occasional cooling problem (that I'm sure I can fix powering the pump from another source, but since the issue is not systematic, I don't want to waste time trying things in hopes it will never happen again...)
Of course one final fix would be to get rid of watercooling. The fans always spin up, so a purely air-cooled system would be 100% fine. I've installed watercooling on this machine because it's overclocked and I couldn't really find a proper air-cooled system at the time that would be efficient enough AND fit on the motherboard! (Big ventirads tend to not have enough room due to RAM sticks... anyway.)
Although I could leave the computer on as some of you do, I still prefer standby if at all possible: this machine draws about 140W on idle. This isn't tiny.
-
Absolutely no problems with hibernation nor sleep mode, doing it every day on my main computer for the sleep mode, another one is doing hibernation for long periods, they all backed with ups just in case.
-
To write 32 GB on a SSD at each standby, will wear out the SSD pretty fast.
This is theoretical. I hibernate several times a day over the last 5 years which was when i installed the SSDs, no any ill effect observed on SSD.
-
I've got a 256Gb Samsung 840 Pro here which has spent about 6 years being hibernated on to daily and SMART says it has used about 4% of its lifespan. This is with an 8Gb machine. The hibernation file on windows is compressed and is only the active memory pages as well. So unless you're ramming that up to 32Gb all the time with already compressed data then you're not going to get anywhere near the point the SSD starts dying.
Also to note my new Mac has an SSD soldered to the motherboard and actually actively uses the SSD as a pagefile as well. This is smartctl output:
Available Spare: 100%
Available Spare Threshold: 99%
Percentage Used: 0%
Data Units Read: 5,714,067 [2.92 TB]
Data Units Written: 2,599,798 [1.33 TB]
Over the lifespan of the device (6 years) I expect to write around 50TiB to the SSD on this basis. Drive is rated to 250TBW. :-//. Meh don't give a shit.
-
That's a surprisingly low number. You made me curious so I checked mine.
My SSD was used with Windows for about 2-3 years. Never used hibernate with it. Since I migrated to Linux the SSD is still plugged in the computer, but it remained unused.
sudo smartctl -a /dev/sda
...
Device Model: Samsung SSD 850 PRO 512GB
...
User Capacity: 512,110,190,592 bytes [512 GB]
Sector Size: 512 bytes logical/physical
...
12 Power_Cycle_Count 0x0032 092 092 000 Old_age Always - 7646
177 Wear_Leveling_Count 0x0013 098 098 000 Pre-fail Always - 106
...
241 Total_LBAs_Written 0x0032 099 099 000 Old_age Always - 41378694293
41378694293 * 512 = 21.19 TB (about 7% lifespan is gone, and never used hibernation)
TBW for Samsung SSD 850 PRO 512GB (https://downloadcenter.samsung.com/content/UM/201711/20171115103115156/Samsung_SSD_850_PRO_Data_Sheet_Rev_3.pdf) is 300 TB.
I guess the writing habbits varies a lot from one user to another.
Yours was ~0.2 TB/year with hibernate on, mine was ~8.5 TB/year, about 40 times more and that's without hibernate.
:-//
-
As always it depends. Had DB server at last outfit with 10x 6.4TB PCIe cards in it. They last about a year each :scared:
-
I haven't powered up my desktop in about 6 months but when I was using it frequently I had it go into standby after a couple of hours, almost never shut it down or rebooted, maybe a 2-3 times a year. Never really had any problems with standby, that was solved by the mid 2000's.
-
On a second thought, I guess my numbers are strange.
Assuming 2.5 years is about 1 000 days, that means I was writing each and every day 20 GB. :o
20 GB writes/day seems way too much. Maybe it was a virus I didn't know about, or a bug, IDK.
It's hard for me to believe I was writing that much. :-//
Anybody else can post the TB written/year please?
-
Probably windows search indexer :-DD
My old work desktop used to get that sort of writes. Each compile run of the product I was working would kick out 500 megs of binaries.
-
As I said, it has mostly worked well for me, but the only problem I've encountered was with processor cooling on one workstation (watercooling.)
Upon resume, when the issue arises, the pump just doesn't restart. But it's only occasional. Most resume events go OK, so this is not a systematic issue. I suppose the power to the pump does occasionally remain switched off when this happens. Since it's random, I can't pinpoint the issue - I guess it's a BIOS-specific problem.
I could of course power the pump through a power rail that is always ON, but I don't like the idea of the pump constantly on when the computer is in standby.
Assuming you have the water pump connected to a fan port, that sounds like the fan controller doesn't like the pump startup current.
You could try disabling BIOS speed control for that fan port, or connect the pump to a different fan port, or connect it to the regular 12V rail.
-
Assuming you have the water pump connected to a fan port, that sounds like the fan controller doesn't like the pump startup current.
You could try disabling BIOS speed control for that fan port, or connect the pump to a different fan port, or connect it to the regular 12V rail.
I'll have to check, but IIRC, I connected it to a fan port that is not speed controlled. I used the speed controlled ones for the fans.
But, the pump startup current could be an idea here. Didn't think of that. Yes maybe (probably) this port still has a protection, and it occasionally kicks in. It never happened upon a cold power-on, but maybe this is a BIOS thing - the port protection could behave differently upon resume maybe due to a small BIOS bug. Or maybe it's due to how the power supply reacts to a resume condition.
I didn't connect to a regular rail, as I said, to avoid the pump running when the computer is in standby. But that could be an option otherwise indeed.
-
Another thing why I don't like about hibernation (apart from taking too long to load/unload the RAM) is that it needs hibernation space as big as the entire RAM.
It does not. Actually it is not even directly dependent on RAM capacity. And it’s a probability distribution, not a fixed value.
Suspend-to-disk is probabilistic in nature, with 100% certainity requiring infinitely large disk. For any success rate R%, a good rule of thumb is: approximately a bit over the amount of memory used in R% of cases. I will stress that: memory, not RAM alone. So, taking Linux as an example, if your processes and tmpfs use 2GB of memory 99% of time, 2.5GB is fine to have suspend-to-disk work 99% of time.
RAM capacity itself is not directly affecting that value: most of RAM is not used to store anonymous pages, which is the primary thing written to disk during suspend-to-disk.
The actual amount of swap needed for hibernation is often even less than the estimate given above, because the hibernation image is being compressed.
-
This is with an 8Gb machine. The hibernation file on windows is compressed and is only the active memory pages as well. So
The actual amount of swap needed for hibernation is often even less than the estimate given above, because the hibernation image is being compressed.
What a swap size you would use if a whole system memory, lets say, 256GB? ::)
What about reactivation time from that state? :-DMM
-
I didn't connect to a regular rail, as I said, to avoid the pump running when the computer is in standby. But that could be an option otherwise indeed.
Only the 5VSB line is active in standby - the 12V supplies that you would use for the pump are turned off.
-
For my "sleeper" 64gb machine, the pagefile.sys is just 10 gig loll
Since i'm on a pcie 4x nvme drive, just take around 3 secs to be ready ....
-
I didn't connect to a regular rail, as I said, to avoid the pump running when the computer is in standby. But that could be an option otherwise indeed.
Only the 5VSB line is active in standby - the 12V supplies that you would use for the pump are turned off.
Ah. I'll look into this.
-
I've only ever used suspend to RAM / S3 on my desktop. Hibernate is much slower since it saves everything to disk and shuts the PC off--but it is better for battery life on a laptop. For a snappy "wake up" on a desktop machine S3 is the better way to go. I generally put my desktop to "sleep" once or twice a day until some event requires me to reboot the machine, which is rare. I believe I had my desktop reliably going in and out of sleep for almost a year at one point. I think the power went out for a couple of hours, at which point my UPS gave up the ghost and it finally went down.
To extend the life of my SSDs I turn off search indexer, hibernate and page file. I also use a RAM disk for the temp folders and as a scratch disk for downloads and other activity that can live in my 32GiB of RAM until it actually needs to be saved to disk. After two years of rigorous use I only have 3.4TBW to my 1TB Samsung 970 Pro and 2TBW on my 2TB Micron 1100. This is with Windows 8.1. Windows 10 is known to cause excessive writes to SSDs not only from constant updates but the file history updating constantly:
https://superuser.com/questions/1229362/windows-10-averaging-over-50gb-of-writes-day-to-ssd-over-9months
On a side note, I noticed that "sleep mode" on my Windows 10 laptop is a lie. While playing music or even a video before putting the machine to sleep it continues to play... with the screen off, lol. I don't know if this is a bug in LTSC 2019 or what. Even when "off" I notice that the battery drains at an alarming rate compared to my past laptops. This is a new Lenovo Yoga "ultra book"... so I don't really know what's going on here. I just turn it off to save the battery as best I can while I'm traveling. It boots so fast that it's not much of a deak breaker for me.
-
I've only ever used suspend to RAM / S3 on my desktop. Hibernate is much slower since it saves everything to disk and shuts the PC off--but it is better for battery life on a laptop.
Yeah. Well, I've never used hibernate even on laptops. I usually put my laptop in standby (suspend to RAM). It draws very little power. Even when it's not plugged to mains while suspended, it can last several days on a single battery charge in standby. So that's fine with me.
-
On a second thought, I guess my numbers are strange.
...
Anybody else can post the TB written/year please?
This is a Linux workstation, always on. It's a fairly recent build, and has seen about half a year of use so far.
Model Number: ADATA SX8200PNP
Namespace 1 Size/Capacity: 2,048,408,248,320 [2.04 TB]
Namespace 1 Utilization: 1,350,480,986,112 [1.35 TB]
Namespace 1 Formatted LBA Size: 512
Available Spare: 100%
Available Spare Threshold: 10%
Percentage Used: 1%
Data Units Read: 13,606,991 [6.96 TB]
Data Units Written: 20,517,355 [10.5 TB]
Host Read Commands: 105,944,790
Host Write Commands: 249,362,510
Controller Busy Time: 6,730
Power Cycles: 57
Power On Hours: 5,123
The drive has a 5 year warranty and is rated for 1280 TBW.
The machine has 64GB RAM, no swap and it sees moderate use. As you can see, there's significantly more writes than reads to the SSD.
-
Invariably for me, standby mode has always resulted in breaking things requiring a restart, so I disable it. Hibernation has been even worse.
What I do do now though is use the power button to put the monitors in standby.
Looks like I'm not having it too bad, then. Mostly... because I haven't had much issues other than this occasional cooling problem (that I'm sure I can fix powering the pump from another source, but since the issue is not systematic, I don't want to waste time trying things in hopes it will never happen again...)
If I had good reason, then I would try to resolve the issues coming out of standby but reliability is more important to me. I initially experimented with it but quickly concluded that it was a lost cause at least on this system. It may just have too much expansion hardware like Ethernet ports, external drives, hardware RAID, etc.
Of course one final fix would be to get rid of watercooling. The fans always spin up, so a purely air-cooled system would be 100% fine. I've installed watercooling on this machine because it's overclocked and I couldn't really find a proper air-cooled system at the time that would be efficient enough AND fit on the motherboard! (Big ventirads tend to not have enough room due to RAM sticks... anyway.)
I went the air cooled route this time but have not been completely happy with the performance compared to my last air cooled system. My last system had a 125 watt Phenom II 940 with Hyper 212 and my new system with a 65W Ryzen 3700X and perhaps even more impressive heat sink does not run as cool. I suspect he Scythe Fuma 2 cooler has a manufacturing defect but it works well enough that I am reluctant to take the system apart just to replace it. It just seems like it should work better than it does.
Although I could leave the computer on as some of you do, I still prefer standby if at all possible: this machine draws about 140W on idle. This isn't tiny.
That seems awfully high unless you have a lot of mechanical storage installed inside like I do.
-
What a swap size you would use if a whole system memory, lets say, 256GB? ::)
Please re-read what I have written. The answer is there. It’s in the second sentence of the message.
-
Although I could leave the computer on as some of you do, I still prefer standby if at all possible: this machine draws about 140W on idle. This isn't tiny.
That seems awfully high unless you have a lot of mechanical storage installed inside like I do.
This is a relatively powerful config, with a 2011-socket motherboard, overclocked Core i7 (the 2011 series was powerful but pretty power-hungry), 64GB RAM, NVidia GTX1060, a couple PCIe cards..., 2 SSDs. The only mechanical "storage" I have is a DVD reader/writer. It has watercooling too, which certainly draws more power than just an air-cooled solution.
Of course by "idle" I mean the computer fully on, but CPU load at "0"%.
A more recent CPU and motherboard would probably lead to a lower power consumption. Especially if not overclocked.
-
I seem to be the odd one out here in that I never hibernate or sleep but turn off the PC every night. Or morning if it's been a long night.
-
I tried that for a while, but inevitably I end up with 3 or 4 PCB layouts open that I'm working on, various programming projects, browser tabs I'm using, a few emails in my queue to reply to, it's just too much hassle to figure out what I was doing and get it all opened back up so I try to reboot as little as possible, typically 2-3 times a year.
-
We no doubt all have different ways of working. When I am in the workshop, for instance, periodically I will clean the work area and put away tools that are surplus, etc. At the end of some work period I generally put everything away and clean up. When getting going with something I like to just start, not spend time clearing up the last mess before I can even think about what I was going to do. Perhaps my PC workflow follows that kind of mindset.
But, particularly with PCs, what happens if you get a powercut? Or the cat walks over your keyboard during the night? That's a lot of work to try and figure out if you need to save whatever changes, of if it's changes to revert.
The other reason your kind of workflow doesn't work for me is because things one often sees tend to fly under the radar after a bit. My bench has loads of half-finishes projects, some of which I forget I even started, because I put them aside to do something else, and after a bit they are just bench furniture.
On a practical level, turning the thing off saves me 250-300W. I just checked and counted 12 mains plugs (or wallwarts) which are switched with the PC, so that's a fair number of potential fire sources to not have nightmares about too :)
-
As always plan for failure not success and you’ll never kill anyone or lose your work
-
But, particularly with PCs, what happens if you get a powercut? Or the cat walks over your keyboard during the night?
For me, nothing much really. I have an UPS, so powercuts are no problem. Cat walking over keyboard won't do much either. It will make the computer get out of standby. But since it requires entering your password when it resumes, if the keyboard is pressed randomly at this point, nothing happens except filling up the password field. If the cat ever managed by pure chance to actually hit the right keys to enter the proper password and the enter key, then, well. Guess I would start believing miracles are possible! ;D
-
Obviously you save your open files before you put your computer in standby (or even hibernate if you are cautious) :-// If the power fails, you're no worse off than if you had turned it off.
And, a desktop PC that draws 250W in standby is seriously broken. I've never seen one that draws that much even in a running/idle state.
-
And, a desktop PC that draws 250W
I think you missed the other 11 plugs (or wallwarts). But besides that, the figure is for when the system is idle, not hibernating or sleeping.
But since it requires entering your password when it resumes
Jesus! No! One of the things that put me off Linux-on-the-desktop was having to enter the bloody password all the time. (It's since been changed, but the attitude of the devs was that they knew better than me how I should use my system).
-
I have an UPS, so powercuts are no problem.
Sometimes a UPS will cause a problem where there wouldn't've been one without it. I have a client's UPS which he returned because it was causing his PC to reset periodically. A server here suffers a similar issue (different UPS make). My normally reliable (for APC) APC UPS doesn't remember when a previous battery runtime test lasted all of 5 seconds and tries it again the next month.