CPU Instability Question

Foxie3a · Apr 15, 2004

One of the main differences between the P4 and the P4 Xeon is the length that it is tested. A Xeon is tested to, and built to be able to be on for much greater lengths than a typical desktop chip is. Thus the server/workstation role.

Now here comes my question. When a processor becomes unstable due to being on too long, and you must reboot(after a few days, maybe even a week or more)...Well, do you really need to reboot?

What is causing the instability? Its not just heat I'm sure. And how does a simple little restart change anything? I am guessing if you have L1/L2 ECC on this would solve that problem without any data corruption.

Am I correct on my assumptions? Please help me with my questions, and thanks in advance.

c627627 · Apr 15, 2004

Don't forget the buggy, ineffective, resource hogging operating system...

Foxie3a · Apr 15, 2004

I'm not talking about software level crashing. I'm just talking about the pure hardware instability issues.

Gnufsh · Apr 15, 2004

When a processor becomes unstable due to being on too long, and you must reboot(after a few days, maybe even a week or more)...Well, do you really need to reboot?

I haven't had any problem with this and ran my computer straight folding for months. If you leave your computer on inefinately, sure eventually you will get an error or two, but the leaving on isn't the problem, it's just things like, well cosmic radiation hitting just the wrong spot and flipping a bit in your RAM, nothing caused by leaving it on. The only way rebooting would help is if it causes the system to lock up. I know windows eventually requires a reboot, especially under heavy useage, even without hardware errors. Or mayybe this is simplyy a problem I don't know about?

Captain Newbie · Apr 15, 2004

I've left several of my boxes folding for about four weeks continuously and haven't had any troubles. As gnufish said it's usually a phase of the moon type thing or a cosmic ray.

Scientifically there isn't any reason for instability and the Operating System (Especially BillOS) is usually to blame.

Zuzzz · Apr 15, 2004

the main reason for rebooting windows is to refresh the memory. Windows is known for memory 'leaks'. Over time these leaks can make a system slow sluggish or even crash. As gnufish hinted its the software nott he hardware causing those problems. I have had a good load of NT/XP and have been folding on some rigs for weeks at a time. The only reason I reboot is morethan likely pure habit. Hardware usually either works or it doesnt. If you have ruled out heat related issues then the hardware is ok. Its the software messing up

Z

Foxie3a · Apr 15, 2004

I'm not having issues, I just have a question.

But Xeons are tested to stay on much longer than normal desktop chips, I know that for sure. Then what is there to gain stability wise?

larva · Apr 15, 2004

Nothing; stable is stable. Intel is just trying to justify their pricing of the Xeon stuff.

Lan_Of_Malkier · Apr 15, 2004

Ditto Ditto

If you have to reboot, it's not hardware unless the hardware is crap. Your O/S is the culprit....

Xeon design is not for 'long haul stability' persay, as they are all designed to run @ 100% for the life of the proc. The Xeon design difference is in cache and information handling mainly. Which would, as you suggested, make it a better server proc.

For that reason they are much more expensive. Who is gonna buy a 'server proc.'? A Corporation or some other business! Which means they are willing to spend generaly whatever amount of money to get the server going. That's why the high prices.

That being said. WinXP can be ran for months w/o needing a re-boot because it is very efficient with memory and the likes.

Hope I helped clear the fog a bit at least.

Foxie3a · Apr 15, 2004

Maybe I am confused. Maybe it was that they had a greater longevity than a typical desktop chip. Along with running at 100% the entire time.

L337 M33P · Apr 15, 2004

Captain Newbie said:
Scientifically there isn't any reason for instability and the Operating System (Especially BillOS) is usually to blame.

There is a reason that hardware can and will fail after a certain length of time. Aside from thermal problems, electron migration or ESD/lightning strikes radioactivity plays a part. Radioactivity is responsible from random hardware errors that crop up every so often. ECC prevents most of these by keeping a parity bit on calculations, so the error rate of servers and such is even lower.

Radioactive emissions such as beta and gamma rays are in a constant flux around us - many sources account for it - radon gas, radioactive calcium in glass and bricks etc. These rays are capable of penetrating the CPU/memory/any semiconductor and if it is energetic enough a ray will knock some electrons out of the way in a CMOS transistor. The effect may be that a 1 gets represented as a 0, which could cause an error in a non-ECC system or just make an ECC system make a note in it's log.

Computers operating at high altitudes are known to fail or generate errors much sooner than ones at sea level - this is due to the increased background radiation from cosmic rays.

Quailane · Apr 16, 2004

Now here comes my question. When a processor becomes unstable due to being on too long, and you must reboot(after a few days, maybe even a week or more)...Well, do you really need to reboot?

Never happens. As long as the temperature doesn't get too high and your voltages don't go way too off, a computer should be able to stay on indefinately. (well, we're talking over a decade here. I don't think anyone really knows how long cpus being made today will last.)

MRD · Apr 16, 2004

Processors don't become unstable. If you have to reboot it's your OS. You almost never have to reboot if you are running unix. Windows 2000/NT/etc are not too bad, but still require some rebooting. Wind 95/98/ME require reboots at least once a day if not more to have acceptable performance.

Foxie3a · Apr 16, 2004

I'm not sure if I would say that. Doing normal stuff, like Office work, or even patched gaming...I would keep my computer on all weekend at a time with Win98SE.

Heat will eventually tear away at the CPU, no matter how cool you make it(well, I can think of some cooling that might do that job).

I think that processors are supposed to last 7 or more years. Roughly of course.

Wizdon · Apr 16, 2004

Thrity to forty degrees is not hot for a circuit to run for an indefinate amount of time. I don't believe that that amount of heat would damage it, even over a period of a decade or more of solid use. I'd be more inclined to say that the wear of CPUs running at low temperatures <40C is not due to the temperature but to other factors.

Yes, I've heard 7 years as well. That's taking into account the 'average user' (oooh how we love this term) who has no idea how hot his closed, fanless, HP case is inside. I think that with care, a CPU could last a very long time, though I doubt its usefulness after 7 years with the pace of this industry.

My brother has been running his Duron 800MHz since November 2000 non stop; he reboots rather than shuts down; and for most of that time at temps between 50-60C. Solid as a rock, shows no sign of wear... not that they would be easy to see or find.

ghostrider · Apr 19, 2004

Foxie3a said:
I'm not sure if I would say that. Doing normal stuff, like Office work, or even patched gaming...I would keep my computer on all weekend at a time with Win98SE.

Heat will eventually tear away at the CPU, no matter how cool you make it(well, I can think of some cooling that might do that job).

I think that processors are supposed to last 7 or more years. Roughly of course.

Where are you getting all these "facts" you keep stating? From all I have learned, a properly made chip should last an indefinate amount of time as long as the temperature it is running in is kept to spec. Having electrons flow through the chip should not cause the materials in the chip to break down.

Captain Newbie · Apr 19, 2004

ghostrider said:
Where are you getting all these "facts" you keep stating? From all I have learned, a properly made chip should last an indefinate amount of time as long as the temperature it is running in is kept to spec. Having electrons flow through the chip should not cause the materials in the chip to break down.

There does come a time where the logic gates themselves will waste away into nothingness after being run for a certain amount of time. It's caused by a multitude of things beyond the scope of this forum.

It is assumed, by most of us gearheads, that the average design life of a processor is ten years. This makes all our percent life loss calculations easier.

Even in spec, and even when things (like CPU's) don't have any visible moving parts, there are still things--electrons--moving, and they're raucous particles.

<edit add> I think that lifetimes gradually will decrease if the inflation of clock speeds continues and a lot more CPUs will fail a lot sooner

</edit add>

Gnufsh · Apr 20, 2004

ghostrider said:
Where are you getting all these "facts" you keep stating? From all I have learned, a properly made chip should last an indefinate amount of time as long as the temperature it is running in is kept to spec. Having electrons flow through the chip should not cause the materials in the chip to break down.

That would be incorrect. All CPUs will eventually die from things like electromigration:
http://www.csl.mete.metu.edu.tr/Electromigration/emig.htm

CPU Instability Question

Foxie3a

Normal Member

c627627

c(n*199780) Senior Member

Foxie3a

Normal Member

Gnufsh

Senior Member

Captain Newbie

Senior Django-loving Member

Zuzzz

Premium Member

Foxie3a

Normal Member

larva

Inactive Moderator

Lan_Of_Malkier

Member

Foxie3a

Normal Member

L337 M33P

Member

Quailane

Member

MRD

Senior Member

Foxie3a

Normal Member

Wizdon

Registered

ghostrider

Registered

Captain Newbie

Senior Django-loving Member

Gnufsh

Senior Member

Similar threads