Prime95 Stability - Is it really that important?....

Associate
Joined
19 Feb 2004
Posts
1,210
Location
in a house
Ok, so I am no stranger to overclocking, and in most tests, the only way to truly see whether your overclock is definitely stable seems to be to run Prime95 for a good few hours to make sure you have no errors.

In the past, I have run it and it has immediately failed, although windows and games (for the most part) are ok. For general day-to-day use, and gaming, if my PC doesnt crash - I dont care if it can't run a stupid test for X amount of hours.....

However, I tweaked some settings recently after rebuilding my PC and found that it was in fact unstable - another thing that confirmed this was a windows version of memtest that instantly brought up messages advising me that my RAM had a problem holding any information at all.

So, after a few adjustments and dropping clock speeds back etc and tightening timings - 333 x 9 on the C2Q6600 @1.3v giving me 3ghz and 800mhz on the RAM running at 4-4-4-12 @ 2.0v - the memory program reported no errors.

So, just before going to bed, I thought - why not, ill stick Prime95 on all 4 cores and go to bed, giving it a good 9 hours running time. The morning came, and to my delight, the PC hadnt rebooted, albeit the room was quite toasty, so all seemed well.

However, on closer inspection, the thread it was running on Core2 had failed after 18 minutes of operation. The other 3 were still running fine 9 hours later.

My thoughts are - what does this actually mean? Is the overclock unstable? Could it just have been a hiccup and the overclock is stable?

Once again, I cant see any problems in games, windows is always fine, and any other test I do works ok. I havent re-run Prime95 yet, which I may do later.......

Cheers,

Craig
 
Its different for everyone really, in some cases, your machine can be unstable in Prime95 but perfectly stable in gaming and applications.

Sometimes unstability in prime will cause reboots and lockups during games.. basically if its stable for what you use it for, its stable imo.

If it bothers you that much you can try upping the volts a little. 1.3 is low for 3.0 not all Q6600 can manage that.
 
Usually when one core fails, as said, you need to up the volts a little.

If you don't get it prime stable, when you have a crash in the future, you might not know whether it's down to instability or a software problem, IMO.
 
Last edited:
It might only take a notch or two of voltage to make it stable. Is 1.3v in bios? You might be only 1.2V under load in windows (it is on mine board with vdrop and vdroop)

So long as your temps are under control I would rather have a prime95 stable system for 8 hours than risk instability.
 
It might only take a notch or two of voltage to make it stable. Is 1.3v in bios? You might be only 1.2V under load in windows (it is on mine board with vdrop and vdroop)

So long as your temps are under control I would rather have a prime95 stable system for 8 hours than risk instability.

yeah, its 1.3 in the bios - but how do I know how much its actually trying to draw when under full load - is it something CPUZ can monitor in real time?

ive read a bit about vdroop and cant work out whether it should be enabled or disabled in the bios or not? anyone comment?
 
also, another question which has been niggling at my brain - their is an option on the P5K boards called a transaction booster - on a website (forget where) they have run some fairly concise tests to show that enabling this and setting it to 0, 1 and 2 tends to give about 200-500mb/s on each level you set it too, but most sites say they are unable to get a stable overclock at all with this enabled in any way at all.

does anyone have any experience with this?
 
yeah... depends on what your RAM is rated at and how fast you want it to run... I can run my OCZ 9600 at the highest level no issues when I am running @ 3.2 (400FSB) so RAM is running @ 800mhz (3-4-3-10) well under the 1200mhz it will do.. but If I am running it at higher frequencies then I have stability issues... if you use a programe like memset this is also called the "performance level" of the RAM
 
For a 24/7 use machine, i.e. a machine you use for personal use then yes, Prime is important.

I don't get users who don't like to run it on such machines, on benching machines I can understand but computers you actually use - I simply don't.

You wouldn't run stock speed memory that is faulty and generating errors in prime now, would you?
 
I dont think you need hours stable.. But for the most part people keep stuff on their PC thats important to them... therefore I think its only sensible that you have made sure that the system is reasonably stable... Nothing is going to tell you the system is a 100% stable and I wouldnt just run a small ffts prime but a blend as well.. but seems sensible to me
 
yeah, its 1.3 in the bios - but how do I know how much its actually trying to draw when under full load - is it something CPUZ can monitor in real time?

ive read a bit about vdroop and cant work out whether it should be enabled or disabled in the bios or not? anyone comment?

Yeah from memory cpu-z does show it in real time. And from memory Asus does have a bios setting which practicularly elimantes any voltage drop. Try it either on or off and see what difference it makes under load in windows and leave it on/off depending which gives the least voltage drop.

Like I said on my Abit board 1.3v in bios equates to 1.2v in windows under load in prime95 which might just be too little voltage for 3Ghz (it isn't for my VID chip of 1.25v) but every chip is different.
 
I don't get users who don't like to run it on such machines
It's always the same Mekrel, everyone wants to go to the party but no-one wants to stay and clean up the mess! :D

Having had a quick think about prime stability I believe it is a standard that's evolved for like minded computer builders to show their work is good. I can't test out all the hardware I would like to so I kinda depend on my fellow system builders and overclockers to supply information on the kit they are using and personally for me a few hours of prime stability is a lot more realiable source of information than Johhny-Gunho stating he clocked a chip to 4GHz or showing a 1M super-PI shot :o

It doesn't take too much effort to get a machine prime stable and also to take some nice screenies and post them up for all to see, very helpful and I do appreciate the efforts some people make to share! :cool:
 
I want a machine that's not going to slowly corrupt it's data, so I'd run Prime95 if I overclocked.
 
Big.Wayne said:
important to who? you or me?

I won't run a machine (overclocked or stock) that can't run prime stable for a few hours, I mean, it's not that hard to do now is it? :D

important in general. like I mentioned, I have no problems in day to day use of the PC, games etc are all fine, its just one little test, which stresses all 4 cores at 100% for a very long amount of time - in real world applications, there is nothing that will ever do this, which is where my question stemmed from.

Mekrel said:
For a 24/7 use machine, i.e. a machine you use for personal use then yes, Prime is important.

it is a 24/7 machine, the only time it gets turned off is when vista rapes itself




saying all that, what I am most perplexed about is that 1 core failed after just 18 minutes, yet the 3 othere stayed working fine for 9 hours?

also, my previous question stands - has anyone had any experience with the transaction booster?
 
Simple and worthy stress test for multi core cpu's.

- complete one Folding@home winSMP unit - about 20-36h - if it errors your pc/mac is not completely stable. If it completes one you've helped further protein reasearch and KNOW your system is stable.

Other programes can (and should) be run at same time, so no need to but your life on hold, but your system may feel 10-15% slower than normal especially with less the 2gb of ram.

If all you do is game and surf and watch movies then your never stress your cpu so don't bother with prime unless you want to see if your new cooler is better than your old one.

Can also be usefull if selling your cpu on MM to 'prove' an OC

As to your opening post - looks like you need a voltage boost to keep it prime stable, or not to keep it gaming stable.
 
Last edited:
saying all that, what I am most perplexed about is that 1 core failed after just 18 minutes, yet the 3 othere stayed working fine for 9 hours?

also, my previous question stands - has anyone had any experience with the transaction booster?

Can't help you with the transaction booster as it's not on my board.

However, the core issue is easy. There is always a core which fails first. On mine it's always core 1 which falls over on any stress testing when the others will go on for hours.
 
If all you do is game and surf and watch movies then your never stress your cpu so don't bother with prime unless you want to see if your new cooler is better than your old one.

.

Sorry, i don't agree. I've seen prime-unstable machines crash at meager tasks such as copying files and playing movies or even opening internet explorer, while being stable for days on end, something pulled that magic string to knock it over and it certainly was nothing to do with heat or load, once the machine was prime stable it was fine.

Once again i'll say it, and i'll say it until die..:p an overclocked machine should be able to do ANY and EVERYTHING with the same accuracy and stability it can while in a stock state to really be called successful. Theoretically a stable stock machine should be able to run prime or any stress tool infintely until a piece of hardware gives up etc.. i'd expect the same from an overclocked one, though i would'nt go that far, 24hrs is enough for me ;)
I generally run 3D intensive tests, memtests, prime/orthos, sometimes TAT stress when unsure about the cooling system and now a little intel burn before a system leaves my door. Higher standards? OCD? who knows? :)
I've been using prime since almost its inception and find it a good indicator of a stable system, not the end to be all and 100% accurate, just a good indicator, and coupled with a few other tests generally guarantees a rock solid rig.

As for the cores issue, all cores are not equal and just like single cored cpus, not all clock the same.. so tolerences and overclockability of a cpus cores can vary just as it can with seperate cpus. One might be stable at 1.3V for example while one may need 1.325V for the same clockspeed to be stable. The unfortunate thing being with a multi cored cpu you'll have to up the voltage for all to compensate for that one, thus increasing power draw and heat produced.
 
Last edited:
Back
Top Bottom