I am constantly turning my desktop computer on and off. Sometime it seems that Rosetta setpoints rule my life.

I have just discovered the XP hibernation function. I have never seen it mentioned as a solution to long setpoint times. If you leave Rosetta in memory I would think that you should be able to turn off your computer at any time.

Is there any downside to using hibernation constantly? With WIN98SE restarting the system regularly seemed to help reset everything back to normal. Firefox and other programs supposedly "leak" memory.

Do desktop systems running XP home need a "full" restart every so often.

Jack

---

Whether you need to reboot your machine frequently or not with any version of Windows depends on who you ask. My machines are generally running 24/7 for as long as they can - which for some machines are hours because I need to reboot to load a new kernel or driver when I'm working on those things, and for others several weeks or months. I definitely don't have a habit of rebooting "just for the sake of it".

Generally, if an application is leaking memory, it should only leak within it's own space, so if you shut down such applications, the memory should be released. There were several "should" in that last sentence, so no need to inform me of those nasty applications that cause kernel memory leaks and such - but in my experience Firefox or Outlook do not belong in that category...

Drivers may have memory leaks too (I worked on a driver that at one point leaked about 4K every few minutes, until you eventually ran out of kernel memory - not good, but I fixed that particular one, and I stopped when it leaked less than 1K per day under fairly heavy load - seeing as this was a HUGE driver project, it wasn't entirely feasible to track down EVERY SINGLE possible leak...).

However, drivers are restarted when you hibernate and come back out again, so driver memory leaks shouldn't be a problem. Unfortunately, drivers are also not told that they are coming out of hibernation, so there's some "interesting" problems that can occur by Windows thinking that it can use the device just as if nothing had happened, and the driver has completely started over from scratch - can get you into trouble as a driver writer very quickly if you're not carefull and experienced in those things... But if you try a hibernate a few times, it should be OK.

--
Mats

---

Hi Jack,

a colleague of mine uses hibernation on his astronomy laptop since he installed an old camera on the machine more than a year ago. It has a driver that refuses to work correctly under Win XP and needs a reinstall once the machine is shut down completely, so he just sets it to hibernate. He says he has had no issues with that since.


Regards,

Christoph

There are no problems, but it shuts down and restarts quicker.

But it will not solve the savepoint problem*. Not sure if suspend will or will not (uses only slightly more electricity than off and hibernate) unless it for some unknown reaon keeps the fans spinning like some computers do.


* well when i used it it never did, but that was in earlier versions, don't know why it would change now.

---

Team mauisun.org

There are no problems, but it shuts down and restarts quicker.

But it will not solve the savepoint problem*. ......

* well when i used it it never did, but that was in earlier versions, don't know why it would change now.

I notice that no one is disagreeing with you. That would explain why no one told me to us hibernation to solve my thunderstorm mandated shutdown problems last summer.

I guess I will try it anyway, in a few days. I already have "leave in memory" selected in BOINC preferences.

I usually leave my computer on when I am in the apartment and awake. Since I am retired, that averages about 8 hours a day, 7 day a week. Also Windows update requires me to reboot monthly, and sometime my virus program automatically updates. There just seems to be many reasons to reboot.

Jack

JPL managed to fix the Mars Rovers S2K problem. JPL's Rover Operational Planning software was going to crash at 1000 sol's

---

Hibernation

I agree that logically speaking, hibernation should be invisible to the app; but like others I have found this not to work out quite so well in practice. Problems typically arise as the wall time jumps forwards, and other little sillies that got in the way. Try it and see, you may have better luck.

Reboot

Win 98 and Me need to be booted once a week or so, when they start to misbehave.

I run Win 2k 24/7 doing BOINC and general work (I am using it to type this). It is perfectly ok being rebooted only when update asks - as you say about once a month.

I would guess the same would be true of XP and any forthcoming successor operating systems in the same branch of windows. If you can afford to pay for the power, leave Rosetta on around the clock.

River~~

---

Stewjack,

My wife and I both run Rosetta in "leave in memory" mode, on Win2K laptops, as does a colleage on an XP laptop. We use both suspend (standby) and hibernate all the time.

In general suspend (which I believe powers down and lowers clock rates to the greatest extent possible while preserving RAM content and CPU state) gets used when closing up and casing a laptop for a, "during the day" relocation, while hibernate (which I believe saves RAM content and system state to a file, and then powers off) gets used for longer-term shutdowns (overnight, etc) where suspend would drain the battery.

We don't have any problems with BOINC/Rosetta operation in this mode, but it is fair to point out that the hardware and drivers are not really old and crufty.

My folks and inlaws have all been trained :-) to use hibernate on their XP desktops, simply because restoring previous state from the hibernation file is a good bit faster than a full start of XP. Once again no problems reported, on less than 5-year old hardware.

Either solution lets your crunching resume without any loss of work in my experience (compared to having to restart at the last checkpoint). Thus if you aren't going to burn watt-hours crunching when you aren't using your computer, it is a valid way to avoid marginal loss of work.

I don't believe either technique will help you substantially with application memory leaks, since the memory used by applications is either maintained (suspend) or restored (hibernate). As Mats points out, hibernate may help with any driver-level memory leaks.

Windows updates and the ^&$*(# Adobe reader updates that insist on reboots are the ones that get in my way!

Cheers,
Alan

---

...
Either solution lets your crunching resume without any loss of work in my experience (compared to having to restart at the last checkpoint). Thus if you aren't going to burn watt-hours crunching when you aren't using your computer, it is a valid way to avoid marginal loss of work.
...

Thanks Alan. With the latest tasks from Rosetta that run for hours without checkpointing (eg the series that began FRA...) it is more than a marginal loss of work, and certainly worth testing two or three times.

From your post I deduce that it did not work for me as all my hardware is >5yrs old. Since it works for some and not others, the real answer is that you will have to try it to see, and that the newer the box the better your chances.

R~~

...
Either solution lets your crunching resume without any loss of work in my experience (compared to having to restart at the last checkpoint). Thus if you aren't going to burn watt-hours crunching when you aren't using your computer, it is a valid way to avoid marginal loss of work.
...

Thanks Alan. With the latest tasks from Rosetta that run for hours without checkpointing (eg the series that began FRA...) it is more than a marginal loss of work, and certainly worth testing two or three times.

From your post I deduce that it did not work for me as all my hardware is >5yrs old. Since it works for some and not others, the real answer is that you will have to try it to see, and that the newer the box the better your chances.

How new your hardware is shouldn't effect it. If it couldn't suspend or hibernate it would
a) not give you the option (often happens when the power states are not supported or bad drivers)
b) you would know about it in anger :-D
while it happily crashies or trashes the system.


In Windows terminology

Hibernate does copy everything to a file (i beblive it call Sleep S4 state technically)

Stand By actually turns everything off* (well as off as a computer gets unless you switch it off at the wall) but keeps power to the memory. The CPU should be turned off, harddrives will be off etc.


*Well proper stand by does, called S3. There is also a S1 state where the CPU and fan stays on (I think often called Suspend on modern laptops). This could be older hardware/drivers cannot cope but still try to go into the S3 state and crash ;-) Often you need to setup S1, S3 or S1/S3 in the BIOS in a couple of year old computers or with jumpers on the motherboard on older computer. Often also having to set 5VSB jumpers as well, since at around that time some ATX power supplies couldn't cope with the current needed.

I may have to do some more investigating since I use a laptop as well and I'll check on Vista (desktop) how it's coping with the checkpointing.



OK did some searching on the Stanby / Suspend naming.
Win95 name = Suspend
WinXP name = Stand By
Vista name = Sleep

I still think laptops often use suspend nameing meaning S1 Fn+somthing key, could easily be wrong though.

---

Team mauisun.org

...
How new your hardware is shouldn't effect it. If it couldn't suspend or hibernate it would
a) not give you the option (often happens when the power states are not supported or bad drivers)
b) you would know about it in anger :-D
while it happily crashies or trashes the system.

...

(my highlighting)

Shouldn't but does.

The crucial point is that coming back from hibernate is not just a matter of putting all the memory back the way it was and everything will run from where it left off. That is OK for code that "thinks", but is not OK for code that controls hardware. Hardware drivers need to know what to do to reset their hardware after a power on that occurs away from boot time, or their hardware will never return properly.

In addition there can be timing related issues where more than one task is intended to work together.

I had a laptop that had bios hibernate, and that worked fine with BOINC on win98. To give you an idea of the vintage it was a Celeron 300 or maybe even 200.

When I put XP on, XP broke the BIOS hibernate. So I went into the BIOS, dsabled BIOS hibernate, and tried XP hibernate. That failed too.

When I say failed, the BIOS hibernate literally failed (XP would never come back from hibernate). You could tell parts of the OS were running because it would ping etc, but there was nothing on the screen and no video output. Clearly there is a "wake up the video" call that the win 98 mb drivers had and the XP drivers did not. I am not clear if this was about the OS waking up the video chips, or about the motherboard asking the OS to redraw the screen, but whichever it was there was nothing but black onevery pixel.

XP hibernate would come back OK if Word Excell etc were running, or if no tasks were left runing. If BOINC was running then whatever app it was would always come back with a lost heartbeat message (my guess is the the app was given the cpu before the client, and the app saw that the clock had jumped forward with no hearbeat from client -- but that is only my impression). I believe the issue was driver related. Also XP hibernate always broke the net connections - to get them going again needed a reboot so I ended up using a usb net connection even tho the laptop had on-mb ethernet.

An operating system that belongs in the same era as the mb will have a specific set of drivers for that mb (made either by ms or by the mb supplier). When an OS runs on a vintage mb it uses a generic set of drivers that cover a range of mbs and do the basics, but don't necessarily iron out all the wrinkles.

R~~

...
How new your hardware is shouldn't effect it. If it couldn't suspend or hibernate it would
a) not give you the option (often happens when the power states are not supported or bad drivers)
b) you would know about it in anger :-D
while it happily crashies or trashes the system.

...

(my highlighting)

Shouldn't but does.

The crucial point is that coming back from hibernate is not just a matter of putting all the memory back the way it was and everything will run from where it left off. That is OK for code that "thinks", but is not OK for code that controls hardware. Hardware drivers need to know what to do to reset their hardware after a power on that occurs away from boot time, or their hardware will never return properly.

In addition there can be timing related issues where more than one task is intended to work together.

I had a laptop that had bios hibernate, and that worked fine with BOINC on win98. To give you an idea of the vintage it was a Celeron 300 or maybe even 200.

When I put XP on, XP broke the BIOS hibernate. So I went into the BIOS, dsabled BIOS hibernate, and tried XP hibernate. That failed too.

When I say failed, the BIOS hibernate literally failed (XP would never come back from hibernate). You could tell parts of the OS were running because it would ping etc, but there was nothing on the screen and no video output. Clearly there is a "wake up the video" call that the win 98 mb drivers had and the XP drivers did not. I am not clear if this was about the OS waking up the video chips, or about the motherboard asking the OS to redraw the screen, but whichever it was there was nothing but black onevery pixel.

XP hibernate would come back OK if Word Excell etc were running, or if no tasks were left runing. If BOINC was running then whatever app it was would always come back with a lost heartbeat message (my guess is the the app was given the cpu before the client, and the app saw that the clock had jumped forward with no hearbeat from client -- but that is only my impression). I believe the issue was driver related. Also XP hibernate always broke the net connections - to get them going again needed a reboot so I ended up using a usb net connection even tho the laptop had on-mb ethernet.

An operating system that belongs in the same era as the mb will have a specific set of drivers for that mb (made either by ms or by the mb supplier). When an OS runs on a vintage mb it uses a generic set of drivers that cover a range of mbs and do the basics, but don't necessarily iron out all the wrinkles.

R~~

...
How new your hardware is shouldn't effect it. If it couldn't suspend or hibernate it would
a) not give you the option (often happens when the power states are not supported or bad drivers)
b) you would know about it in anger :-D
while it happily crashies or trashes the system.

...

(my highlighting)

Shouldn't but does.

The crucial point is that coming back from hibernate is not just a matter of putting all the memory back the way it was and everything will run from where it left off. That is OK for code that "thinks", but is not OK for code that controls hardware. Hardware drivers need to know what to do to reset their hardware after a power on that occurs away from boot time, or their hardware will never return properly.

In addition there can be timing related issues where more than one task is intended to work together.

I had a laptop that had bios hibernate, and that worked fine with BOINC on win98. To give you an idea of the vintage it was a Celeron 300 or maybe even 200.

When I put XP on, XP broke the BIOS hibernate. So I went into the BIOS, dsabled BIOS hibernate, and tried XP hibernate. That failed too.

When I say failed, the BIOS hibernate literally failed (XP would never come back from hibernate). You could tell parts of the OS were running because it would ping etc, but there was nothing on the screen and no video output. Clearly there is a "wake up the video" call that the win 98 mb drivers had and the XP drivers did not. I am not clear if this was about the OS waking up the video chips, or about the motherboard asking the OS to redraw the screen, but whichever it was there was nothing but black onevery pixel.

XP hibernate would come back OK if Word Excell etc were running, or if no tasks were left runing. If BOINC was running then whatever app it was would always come back with a lost heartbeat message (my guess is the the app was given the cpu before the client, and the app saw that the clock had jumped forward with no hearbeat from client -- but that is only my impression). I believe the issue was driver related. Also XP hibernate always broke the net connections - to get them going again needed a reboot so I ended up using a usb net connection even tho the laptop had on-mb ethernet.

An operating system that belongs in the same era as the mb will have a specific set of drivers for that mb (made either by ms or by the mb supplier). When an OS runs on a vintage mb it uses a generic set of drivers that cover a range of mbs and do the basics, but don't necessarily iron out all the wrinkles.

R~~

River, new hardware has the same problems (though it did mention where older hardware can cause problems more often. If the driver does not support hibernate properly and has a bug it of course does not work properly, let alone WinXP also having hibernate bugs in it itself unless you apply the fixes.
As you just mentiones your OS would hibernate, so it's the OS that is causing the problem dur to using old hardware. Of course if you used a K5 processor with AT power supply, it would not hibernate.
Again around the era of you very old hardware era ACPI was only just getting implemented properly ;-)

I have a laptop that has no problem what so ever hibernating in WinXP, but certain ATI drivers cause display problems, same as you the screen doesn't come come back on, but that mainly happens on suspend though. Another much newer Pentium-M processor laptop has similar problems with it's Nvidia GeForceGo-FX video.

Normally if an installed driver does not support Hibernate it disables the option, same if your BIOS doesn't support ACPI (or you disable it).

Anyways all this aside, my real meaning of the original age of hardware shouldn't play a part is, if it works and it comes out of hibernate properly, be it 5 years old or 1 month old computer, it should not cause one computer to restart from a save point and the other to continue where it left it.

---

Team mauisun.org

Anyways all this aside, my real meaning of the original age of hardware shouldn't play a part is, if it works and it comes out of hibernate properly, be it 5 years old or 1 month old computer, it should not cause one computer to restart from a save point and the other to continue where it left it.

I don't think we are disagreeing, just talking about different parts of the issue.

If a task backtracks to a save point then by definition hibernate has failed, as hibernate should come back to where it left off, as if the data never left memory at all.

And while there are always driver incompatibilities, they become more and more likely the bigger the gap between the release date of the hardware and the release date of the operating system, because OS vendors spend most effort catering for the h/w that is around at the launch of the OS; and h/w vendors spend most effort catering for the OS that is on sale at the time they launched their h/w.

R~~

R~~

Hi River,

FWIW, when I referred to "age of hardware" I was really taking a shortcut to bug-free behavior, and support for various standby/hibernate states for the drivers. So I'm in agreement with all the traffic from you and FluffyChicken.

In my distant (Win98SE laptop) past, I wasn't fortunate enough to experience having problematic drivers disable hibernation ... Instead it just failed. Some educated guesswork blamed the added-on Orinoco wireless networking card. If that NIC was left operating, it was BSOD time when it tried to return to operation.

For a while my only means for successful hibernation was to make sure I had shut off the radio and done a safe stop-and-eject on the PCMCIA card. Along came some driver and client updates, and the machine could hibernate without error.

My colleage who is hibernating an XP laptop Rosetta cruncher (recent purchase and setup, all latest and greatest vendor drivers in place), is reporting intermittent failures to hibernate (with some sort of "insufficent resources" message) if he tries to hibernate while attached to the laptop docking station. Once again, extra drivers for the station's hot dock/undock support, its network adapter, etc. ... Could be part of the problem.

Bottom line is you might be able improve your results with either a bit of operating procedure and/or driver updates.

---

As the individual who started this thread I would just like to say that you have scared me off hibernation completly.

My problem may have become less serious with the Rosetta 5.36 application. I have just started working on a series sf FRA work units with save points every 3 min!

I can live with occasional work units with hour-or-longer setpoint intervals.

---

As the individual who started this thread I would just like to say that you have scared me off hibernation completly.

That is a pity, as my real advice, and that of others, was to try it and see. At most you lose the WU that is currently crunching, and if it does work then you potentially gain something every time you hibernate.

All the rest, the scary stuff, is about the ways it can go wrong. But FC was right, if it works it works, and if it don't it don't, and it is reasonably obvious which is which.

R~~

As the individual who started this thread I would just like to say that you have scared me off hibernation completly.

That is a pity, as my real advice, and that of others, was to try it and see. At most you lose the WU that is currently crunching, and if it does work then you potentially gain something every time you hibernate.

All the rest, the scary stuff, is about the ways it can go wrong. But FC was right, if it works it works, and if it don't it don't, and it is reasonably obvious which is which.

R~~

Ditto, Keep using it since there is nothing wrong with it. Forget the scary stuff ;-). I use it all the time on laptops and desktops.

---

Team mauisun.org

I think the warnings about what can go wrong is more like the warnings you get with the medecines - if you take the time to read any medication, it usually warns for all sorts of side-effects. But most people use the medication without any side-effects, a few get mild side-effects, and very rarely does someone come up with a really strong reaction. But they still write about all the bad things that could happen from the medication...

I agree, try it and see if it works for you. In most cases it does...

--
Mats

---

Works well for me with the WCG UD proprietary agent and BOINC - 5 projects (UD/WCG/Rosetta/TANPAKU/SIMAP (latter on periodic projects). Use it regular as power is not stable here. The UPS allows orderly close. Before, put the agent in sleep mode, hold shift key before mousing the start button in WinXP (if the hibernation option does not show up in the close option), hit hibernate and it takes a minute to close. Standby still requires some juice to maintain memory state, whereas hibernation does not. Very rarely does it jump back to the previous checkpoint save (due the mentioned clock sync issues).

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

Thank you for all your help, but I have made my decision. I plan to stop monitoring this thread.

I think that I understand my own situation best. Just a few points.

I have a Compaq and can not use any of the XP disk recovery options. The computer was well worth the price because, it had been stored in a closet for three years and was given to me for free. I still had to download SP1 and SP2 over my dial-up connection! I don't want to think about reinstalling my OS.

Once again, Thanks for all your help.

Jack

...The computer was well worth the price because, it ... was given to me for free.

Now then, that is quite a strong argument :)

Go well
R~~