How long to run prime95?

ferociouscorpion

Distinguished
Apr 6, 2009
11
0
18,510
I overclock my E5200 to 3.33 mhz
idles around 39-42 c and load using prime94 goes up to 71 c at first but has been staying at 60-65 c using stock hs and fan

I've had it run prime95 for almost 12hrs now and it has no problem.
Is it safe to say that it is stable now and I could stop running prime95?
 

optiprimox

Distinguished
Apr 14, 2009
24
0
18,510
I've done around 2-4 hrs of both and have had good stable systems. I don't think you really need to run it 24hrs or more, as some extremists imho
 
I am one of those "extremists" who does a final 24 test run.

I was working up what I thought were the final settings of a Q6600/Gigabyte motherboard system. I decided to just let it run overnight. Next morning, it had crashed at the 14 hour point. Repeated that evening. It reached 18 hours before it crashed. After that, I test for 24 hours.
 

foolycooly

Distinguished
Dec 26, 2008
702
0
19,010
Interesting note for the 'extremists.'

I ran prime95 blend for 13+ hours overnight w/o a hitch at my current settings. A couple of weeks later I got BSODed while playing an online flash game. I have had zero BSODs since, and do not plan on tweaking anything further. I have not BSODed during any heavy gaming sessions or during any heavy multitasking...just thought it was interesting to note that even if you run prime for 13+ hours, there is a possibility something can be missed.

I am also an advocate of enjoying your PC...run prime for a few hours and if it passes, enjoy what you paid for :p
 

MaDMagik

Distinguished
Oct 12, 2008
864
0
19,010



Same situation here, I used to run it for 12hrs, forgot about it once and it crashed around 18th hour. I run it for 24hrs since.
 

michaelmk86

Distinguished
Dec 9, 2008
647
1
19,015
There is no point to run prime95 for more than 20 min.
If you pass 20 min running prime95 with out crash that mean that cpu is 100% stable!!!
Running prime95 for 24 hours is just stupid I can say anything more.
 

foolycooly

Distinguished
Dec 26, 2008
702
0
19,010


This is so incredibly wrong.
 

MaDMagik

Distinguished
Oct 12, 2008
864
0
19,010


If you can(?) say anything more than how about some explanation regarding your theory?
 

michaelmk86

Distinguished
Dec 9, 2008
647
1
19,015


Prime does not reflect realistic use scenarios. It could easily fail at 25 hours or at 48 hours so what are you going to do? leave your pc forever running prime to make sure? of course not. A friend of mine was running a prime unstable pc for 5 months with no single problem while playing games etc. So prime is just a rough indication that if you push your cpu at its absolute limits it may fail which of course you will never do under normal use.

So leave prime for 20 min and then enjoy your overclock benefits in realistic scenarios. ;)
 

MaDMagik

Distinguished
Oct 12, 2008
864
0
19,010
Well with gaming, web surfing etc maybe it really isnt a must. But my main PC is on 24/7 and when you do video encoding and have cpu under 85-95% load for hours it becomes a bit more important. If there was no chance for cpu instability causing damage to my work maybe I wouldnt care about it that much either.
 

yorkaa

Distinguished
Jan 25, 2010
3
0
18,510
wel, both michaelmk86 and MaDMagik raised up good points,

people who run prime for days and only used to play games, surf net,
they really living in their perfect world
 
I usually run it for about an hour for temperature test, and maybe 2-3 hours for stability test.

As overshocked stated, all PCs will crash eventually even on so called "stable" PC.

As long as I get no more than 1 crash a month, I'm totally satisfied. So far I haven't had a single crash in 5-6 months and I haven't done more than 3 hours of Prime95 tests.

My main goal for running Prime95 is to monitor the max temperatures.
 
We all do different things with our computers. If all I did was gaming and it crashed every once in a while, big deal, so what? (Sounds like a song, doesn't it? :) ) I could just as easily chalk that up to a buggy video driver.
 

jjd99d

Distinguished
Apr 20, 2010
1
0
18,510
Prime95 24 hours is accepted as stable with the O.C. community as stated and accepted in Wikipedia. But after all it's your computer. If your happy with it telling you 2+2=1 then live with it. I just hope I don't ever get any data or torrents that have been on your system. Can you say "Data Corruption"
 

ooostephen

Distinguished
Jul 9, 2010
6
0
18,510
Has anyone done the 24hr test on a stock system? Would you expect it to fail at some point, or is stock the most stable and expected to run the test indefinitely?

I have concerns over shortening the components and materials life span by running the torture test for too long. I ran the 'blend' torture test for ~20min, and have never had any problems as an advanced user, even with Flash Pro, IIS, 3 monitors, games/dvds/etc..., and other dev apps.
 

CaptSkip8

Distinguished
Jul 8, 2010
5
0
18,510
I'm new to this OC stuff, and though my current build is my 4th, this is my first that I'm going to OC.

I'm slowly working through the fine post by 4ryan6, AMD Black Edition Overclock Guide, and have completed only the first changes of stepping-up the DDR3 to run at 1600.

AMD 965 BE 125W
ASUS M4A78T-E
G.Skill F3-12800CL9D-4GBRL
Zalman 9500A

I'll be running Prime95 for ~30 minutes after each tweek, and then for a few hours when I think I've got the CPU speed where I want it.

I'm new to OC, but not to OCD. :lol:
 

DrGonzo

Honorable
Aug 12, 2013
8
0
10,520
To everyone that says "You must run Prime 95 for 24 hours or longer to verify a stable system."

First, why not 25 hours. If your circular logic tells you that running it for 4 hours means nothing because it could fail at 5, how have you arrived at this "24 hour" number? It's completely arbitrary and I say you're all full of nonsense. Most "factory" PCs will fail these tests. Believe it.

My overclock is completely stable yet fails IBT on Very High one out of fifteen times no matter what I do. My overclock is COMPLETELY stable. I have ZERO problems. In fact, the only time I've encountered trouble is when I ran Prime 95 over night and awoke to find that windows had recovered from an unexpected shut down. This freaked me out so I started tweaking settings. After about ten BSOD's, I said to myself, "why the hell am I doing this?". I went back to my initial settings and have been happy ever since. Don't believe the hype!!! The only way to stress test your system is by running your system in YOUR normal capacity.
 
Just wanted to add something to this thread even though it's fairly old. There is a point that is rarely pointed out on any forum threads related to stress testing with Prime95 to verify stability of overclocks. CPU stability isn't the only or even the main cause of BSOD's. There are many other factors that can cause a system to crash including driver issues, memory issues (Unrelated to overclocked RAM, just to clarify), hardware faults not caused by voltage and problems within an applications code or the kernel itself in some cases.

Even a poor quality PSU can and does cause similar errors when voltages are not clean and delivered in the expected and necessary amounts. So just because you get a bsod or your system crashes at some point after setting up an otherwise unremarkable overclock, does not mean in and of itself that the overclock is unstable or that you failed to run Prime long enough.

That being said, running it for only a short time or a couple of hours does not allow it to run through all the FFT lengths which is the REAL reason it was originally recommended to run it for 24 hrs. 16.5 hrs is generally long enough to have tested all relevant FFT lengths according to every reliable source I asked.
 

chromic

Honorable
Jan 6, 2013
1,091
0
11,360
My "stable" is that i can run p95 smallfft for 20-30mins/ do cinebench and run furmark.

(havent had any bsod/crash in real life situations when doing this)

Playing/Streaming/watching movies/benchmarking, even tho the OC is "unstable" :D
 

groovesalad1

Reputable
Sep 8, 2014
4
0
4,510


1."Data Corruption?" If you dont want the possibility of corrupt data on your computer, why do you so easily and willingly accept that same possibility of corrupt data in your research? you cited wikipedia as a source. the website written by the users. The majority of it is well written, and most likely written by one or many individuals very knowledgeable on the subject. Though very useful, its not sufficient to be blindly accepted as fact. Especially not by any sort of specialists. Assuming those you called the O.C community are experienced and know this stuff better than myself and most of the world, it's probably safe to say they dont spend much time on wikipedia. They specialize in this area, and would use a more reputable organization/website/forum/media, etc... for their purposes. You dont learn authentic mexican food by eating at taco bell...

2.Then, you belittled somebody else's opinion by comparing it with extreme oversimplification and, I'm going out on a limb here, obviously false. I'm not calling myself an expert in any way. I just built my first computer recently. Had no idea what a burn-in was or how to stress test. The computer I built is way overkill for the gaming I do. I ran prime95 for 25 mins and called that good enough. So far, it's proven to do more than miscalculate 2+2 and empty my wallet.

 

__ChEM__

Reputable
Sep 24, 2014
1
0
4,510


Ha lots of people on this thread a bit too desperate for high overclocks. Been there done that when I was a few years younger, now I just accept that some chips are better than others and I don't have the cash to keep buying chip after chip for the sake of an extra few 100 MHz like people getting up towards 5GHz must do (either that or they've had a lot more luck than me, every single time I've OC'ed).

No one is telling you not to run an unstable system, just don't get upset when you find bits of your mp3 albums get skipped over and your large rar's, avi's etc suddenly don't work etc etc etc.

People posting that they don't get BSODs: just because it doesn't kill your whole session doesn't mean that the processor is functioning 100% correctly. It's getting through 4-5 x 10^9 cycles every second (ok well full load it is at least) so if you have a 0.001% malfunction rate the number of errors is still immense. You might not see it or care about it in normal usage but trust me you will see it in the data you're storing, especially after running unstable for 6 months+.

I've just been OCing for the first time in a while, got it decent now and done a check of my main array. 16 inconsistencies found already from a single afternoon of fiddling and it's only halfway there. If you don't have a mirror / parity to check against, that kind of corruption is now stuck on your hard drive forever and you can only hope it's not killed anything important (for the record as all decent OCing guides should say right at the top btw, keep your OCing confined to a dummy bare OS that you can sacrifice then plug your proper drive(s) back in once it's stable). The idea of running 24/7 like that (my configs were final tweaks with prime running stable for 6 hours before failing here, nothing heavy) with any kind of valuable data is not something I would ever consider.

Letting go emotionally of your hopes of getting near 5GHz after splashing the cash might be hard but having experienced both, I am firmly of the opinion that data loss is far far worse.

Groovesalad1, in the same post you say you've got little to no experience but then you're getting argumentative with jjd99d in a quite nonsensical fashion. Yes wikipedia isn't much of a source but the point is that 24h is a generally accepted standard from experienced overclockers. That is just a simple fact that you can't argue over, wherever it's written. It might not be a perfect test but statistically consistently passing 24h is a hell of a lot better than failing every 4-6h.

And the 2+2=1 point may be simplified in a way but in essence it's completely correct. Your CPU is pretty much an overly complicated calculator, entirely based on mathematical functions. The data your system is built on (whether you're talking about the stuff you're running and experiencing on your screen or the data on your HDD, network traffic etc at a more fundemental level) all goes through it, and if it's demonstrably unstable, it is spewing out a hell of a lot of wrong numbers ala 2+2=1 which will have a magnified effect in complex data, which really ain't great mate.

And the only way to verify stability is to process data (e.g complicated prime calculations) through it, see what comes out and then verify the answers against known correct data. This picks out slight miscalculations which will screw up a few bits and bytes hither and thither, but may not crash your computer. For the record there are probably better soak tests to do than prime95 as it's not that comprehensive, it's better as a max heat test more than anything, although whether you could ever work a multi-core cpu to get that hot in normal usage outside of some professional workstation stuff is also questionable imo, I tend to use normal stress test temps as the max as I know I won't ever really max out 6 cores for more than a tiny while.

HTH