Stress Test Run Time vs Overclock Method

[4Ryan6]

This seems to be a constant subject of disagreement so lets examine it and discuss it rationally, I’m going to start this off and then you tell me why you may agree or disagree.

The time needed for stress testing an overclock totally depends on what is actually involved in accomplishing the overclock.

A raised FSB overclock knocks everything out of specifications so various voltages, speeds, and timings, have to be tweaked to bring it back within specifications, Prime95 for instance has to be run extensively, in testing your settings to discover what various settings lead to a stable platform.

Once you’re confident you’re stable some time needs to be run to assure you are actually stable but even under those circumstances 24hr runs of P95 are just not necessary, if you have no errors in 4 hrs, you should be fine.

If that doesn’t sound right to you, then here’s a good rule of thumb:

Run prime95 for the maximum amount of time you will actually use your computer at a sitting and add an hour.

No one is going to use their computer for 24hrs straight, unless you’re having some kind of change off marathon, changing out users while the others sleep.

A raised CPU multiplier overclock does not knock everything out of spec, so quite a bit less stress testing is needed in reaching a stable overclock, so long term stress testing is just not necessary, 1 to 2 hrs with no errors is enough for that type of overclock.

I saw one claim of stability saying he runs Prime95, 24hrs for each of the 3 tests, Small FFTs, Large FFTs, and the Blend test, that’s totally ridiculous!

Long run Prime95 claims of stability and bragging in one post and crying in another because their favorite game crashed?

Why?

Prime95 and all the stress testing programs, do not test the graphics or the sound card, so you’re better off using a combination of benchmarking and stress testing, and don’t rely on just one, to claim your stability.

When Prime95 first came on the scene, FSB overclocking was just about the only way to overclock, there were few choices in those days, there were some unlocked multiplier CPUs on the market, but they were the most expensive CPUs available.

Overclockers prided themselves in getting every last mhz they could get, but today ghz is possible and not much trouble at all to get there, so try to keep that in mind when you simply raise your multiplier and CPU voltage to reach your overclock and everything else stays within specifications.

Extremely long term stress testing is just not necessary.

This is a subject that's been put off way too long, now lets hear what you guys have to say about it, keep it civil, with no insults. Ryan

 

[beenthere]

I disagree with a lot of your conclusions but you're certainly entitled to hold whatever opinion makes you happy.

Prime 95 for 24 hours is the baseline for a stable system IMO. If it can't do 24 hours of P95 without a hiccup then it's not really stable. It may be good enough for gaming or whatever but it's not technically stable.

I use OCCT for a severe quick check when OC'ing. If it won't pass OCCT it won't pass P95 for 24 hours IME. The nice thing with OCCT is you get instant results in an hour or less. When you find the point where OCCT will not longer run the extreme test without issues, you back up one notch and run P95 for 24 hours.

I'm not interested in the last Hz of CPU or GPU speed as that's just for bragging rights. I'm interested in getting the most from the hardware with 100% stability. There are enough issues with the O/S and software that I don't want to be dealing with three issues when I can eliminate hardware problems.

 

[Cygnus x-1]

Nice post ryan. I have to agree for the most part. In my experience, I have concluded that a 2 hour run in most cases is enough. I have tested this by going with 2 hours first and then 8 hours or more after. In each case, If I was stable after 2 hours, so was I after 8 or more. This has held true for the most part through multiple platforms, including i7.

All that being said, maybe if your running important data or a server, you might want to run stability checks longer, just to make sure. I personally just have a home computer. So the potential of corrupt data doesn't bother me, as I am completely backed up and a reinstall takes little time.

 

[4Ryan6]

As the Title says, "Stress Test Run Time vs Overclock Method", different overclock methods require different run times of stress testing.

Getting my Intel Q9550 to 4.0ghz was the most difficult CPU I have ever overclocked, it was back to overclock 101 for me, balancing all the voltages, CPU, PLL, VTT, GTL, NB, to get a stable overclock.

It's the most I'd ever run Prime95 juggling all those voltages to a stable outcome, but overclocking my Intel 2500K to 5.0ghz was a piece of cake, with an unlocked multiplier, it just did not have all those voltages to juggle to get to a stable overclock.

So there is less possibility of errors from Prime95 simply because the majority of the platform is still within specifications because the overclock didn't require taking it out of specifications.

Prime95 was designed in 1999, the last recommendations for 24hr run time that I found was around 2008, adopted by the overclocking community, we are flooded today with unlocked multiplier CPUs, do you not see a difference in what is involved today, as was involved yesterday, regarding the overclock process, and the 24hr run recommendation.

The 24hr run time is still locked in peoples minds like it's "Word", but CPUs today with unlocked multipliers just don't need that much run time, it's kinda like an adopted tradition that no longer fully applies.

Honestly, if I was running something mission critical, why stop at 24hrs if lives were on the line wouldn't 24 days be better (Exaggeration!), but if I was running something mission critical, I wouldn't be overclocking it in the first place.

Stress test programs like Prime95, OCCT, Intelburn, etc. actually stress more than just the CPU, they also stress the M/B, Chipset, Memory, Voltage Regulators, Power Supply, etc., so you need to be conscious about just how long you allow those programs to run.

I've spent way too much money on my hardware to leave a stress test program running on my computer all night long unattended.

 

[Cygnus x-1]

I think some work station owners run overclocked rigs. My point was in that type of situation maybe a little more testing is warranted.

 

[jsc]

Well, ryan, I am a proponent of 24 hour P95 runs. But then, I am still limping along with Core2 systems with locked multipliers.

And although I favor 24 hour test runs, I have tried to be clear that there is a lot of discussion about how time is really necessary.

 

[4Ryan6]

Well, ryan, I am a proponent of 24 hour P95 runs. But then, I am still limping along with Core2 systems with locked multipliers.

And although I favor 24 hour test runs, I have tried to be clear that there is a lot of discussion about how time is really necessary.

I've known of your standing on P95 and what you have accomplished with your setups overclocking the core2 systems.

Hopefully this thread can logically lead to sharing information between us all as to what kind of times are actually necessary when overclocking between the different methods, locked and unlocked multiplier CPUs.

My goal is not to push unstable machines, my goal is to arrive at a consensus of what is a necessary run time with today's unlocked multiplier CPUs.

I appreciate your input Thank You!

 

[arthurh]

Well, I agree with Ryan on this subject and that is not backed by any other fact other than my personal experience.

A lot of people have opinions as opposed to actual facts other than their personal experience.

That is not to say personal experience does not count!

If we do not include our personal experience then this site would not have nearly the following it enjoys now.

Wikipedia, LINK, scroll down to "Computer processors", weighs in on this subject but starts it with a disclaimer asking for LINKs to back-up any in-put.

They do provide links that describe in detail what OC'n is, how it is done and the many parameters of software stress testing needed to help, not ensure stability. As this article indicates, LINK, read 2nd paragraph under "Stability and functional correctness", there is NO absolute way to assure "technical" stability due to the nature of the beast.

If you actually start clicking on all the links provided regarding this process they take you on a long educational journey about what we are trying to acheive with the tools at hand. It goes along the lines of..., "The more I learn the more I realize how little I know".

I hope what I have said here is read carefully and not taken as a "you are wrong and I am right" post as I do mean well to all here. As we all gain experience with the many different CPUs, GPUs etc; out there and share our knowledge gained first hand with each other we all become better OC'ers.

Good for you and good for me.

 

[4Ryan6]

@arthurh, That 2nd link paragraph deserves quoting, it confirms some of what I've been trying to get across all along, secondly the correlation to time posted data vs our present day CPUs isn't covered anywhere that I could find so far.

Meaning all the past dated articles, do not really cover the unlocked multiplier CPUs we have available today.

Stability and functional correctness
See also: Stress testing#hardware

As an overclocked component operates outside of the manufacturer's recommended operating conditions, it may function incorrectly, leading to system instability. Another risk is silent data corruption by undetected errors. Such failures might never be correctly diagnosed and may instead be incorrectly attributed to software bugs in applications, device drivers, or the operating system. Overclocked use may permanently damage components enough to cause them to misbehave (even under normal operating conditions) without becoming totally unusable.

In general, overclockers claim that testing can ensure that an overclocked system is stable and functioning correctly. Although software tools are available for testing hardware stability, it is generally impossible for any private individual to thoroughly test the functionality of a processor.[8] Achieving good fault coverage requires immense engineering effort; even with all of the resources dedicated to validation by manufacturers, faulty components and even design faults are not always detected.

A particular "stress test" can verify only the functionality of the specific instruction sequence used in combination with the data and may not detect faults in those operations. For example, an arithmetic operation may produce the correct result but incorrect flags; if the flags are not checked, the error will go undetected.

To further complicate matters, in process technologies such as silicon on insulator (SOI), devices display hysteresis—a circuit's performance is affected by the events of the past, so without carefully targeted tests it is possible for a particular sequence of state changes to work at overclocked rates in one situation but not another even if the voltage and temperature are the same. Often, an overclocked system which passes stress tests experiences instabilities in other programs.[9]

In overclocking circles, "stress tests" or "torture tests" are used to check for correct operation of a component. These workloads are selected as they put a very high load on the component of interest (e.g. a graphically-intensive application for testing video cards, or different math-intensive applications for testing general CPUs). Popular stress tests include Prime95, Everest, Superpi, OCCT, IntelBurnTest/Linpack/LinX, SiSoftware Sandra, BOINC, Intel Thermal Analysis Tool and Memtest86. The hope is that any functional-correctness issues with the overclocked component will show up during these tests, and if no errors are detected during the test, the component is then deemed "stable". Since fault coverage is important in stability testing, the tests are often run for long periods of time, hours or even days. An overclocked computer is sometimes described using the number of hours and the stability program used, such as "prime 12 hours stable".

 

[casualbuilder]

Not to go totally out here on a whim, but what about the "Long-term stress testing will shorten the life expectancy of your cpu"? does this hold water as well? I feel that it does, as any use technically decreases the life, especially since most people only use around 60% (even in gaming).

 

[4Ryan6]

Overclocking by itself shortens life expectancy of the CPU, and long term stress testing isn't going to help that.

Read the 2nd Link provided by ArthurH.