'prime stable'

Soldato
Joined
14 Dec 2005
Posts
5,036
was using superpi, always passed the tests with that
now trying prime, obviously a harder test as it failed after 3 mins so have had to mess with settings
but what is 'prime stable' ?
theres the benchmark test and the torture test, also the settings in prferences, which options do I choose for a reliable test ?
thanks
 
I use the Blend stress test which does cpu and ram use something like sp2004 which is just like prime 95 but it has GUI interface to Prime95's Torture Test here

I have just finished doing a stability test..
2.91ghz.JPG
 
Last edited:
Use torture test.

Small fft = Cpu test
Large fft = Cpu + little ram
Blend = CPU+memory

The amount of time it runs without errors is very debatable. I personally class my setup as stable if it will run 8hrs of prime95 (each test) without errors. Some go for 24hrs+! Many people have had errors after the 8hr mark so in retrospect I guess my system may not be wholly 'stable' - its just a matter of how stable YOU think you need to be. An unstable system can create data corruption but I have had my setup fail p95 in seconds yet ran everything (games+encoding etc) for days without a problem. If it passes the 8hr mark then imo it is stable for the everyday user. My rig is now 8hr prime stable and has been so for over 4 weeks and I have yet to encounter a single instance of data corruption.
 
To be honest i am not a major fan of prime, i have used it once or twice and allegedly my pcs at the time were "unstable" yet without changing anything they would run anything else i could throw at them 24/7 which included gaming for several hours, folding, seti etc so it's not like they were idleing during that time.

To me my pc is stable if it will do what i want, when i want, for as long as i want it too. It's not like its the pc in charge of launching the next shuttle mission or the pc running some nuclear reactor. Lets face it sometimes pcs just crash for no reason what so ever.
 
I agree Prime stable isn't vital if the PC is just for personal use but if you're using it for science (folding, seti) it's important that the PC is producing valid results so prime stable becomes much more important as it tests against a set of known results.
 
been playing need for pseed most wanted on max settings this morning, it's fine, except one slight problem
i put another post in general hardware about that though as I dont really know what it could be
also played cod2 for a fair while a few nights ago
 
builder22 said:
been playing need for pseed most wanted on max settings this morning, it's fine, except one slight problem
i put another post in general hardware about that though as I dont really know what it could be
also played cod2 for a fair while a few nights ago
To get stable at 2.6 i had to run 2.9V through my memory, at 2.8V it would run COD2 fine (1600x1200 Max settings with 2xAA/AF) but TCM/RTW would crash the machine. COD2 is GPU intensive but doesnt push the memory very much.
 
ReeMISM said:
I use the Blend stress test which does cpu and ram use something like sp2004 which is just like prime 95 but it has GUI interface to Prime95's Torture Test here

DONT USE IT!

SP2004 is based on an old version of Prime (version 23) AND ISN'T OPTIMIZED for the Athlon 64 architecture, therefore you are wasting your time running it it means nothing.

You need version 24 of Prime95:
http://www.mersenne.org/freesoft.htm
 
I find the best game for testing my cpu overclocking and memory etc, is Counter Strike Source, if my pc is unstable it will always crash when playing this,


Yet if i play Battlefield 2 it wont crash atall :confused:
 
rxmac said:
To be honest i am not a major fan of prime, i have used it once or twice and allegedly my pcs at the time were "unstable" yet without changing anything they would run anything else i could throw at them 24/7 which included gaming for several hours, folding, seti etc so it's not like they were idleing during that time.

If you're not Prime stable, you've been returning bad (useless) work units for SETI and folding.
 
Yeah I'm another one that doesn't bother with P95, after an overclock I'll run 32MB PI and if it runs CoD2 for 2hrs i'm happy.

I realise its important for the SETI people but I bought my PC for gaming:)
 
mmj_uk said:
DONT USE IT!

SP2004 is based on an old version of Prime (version 23) AND ISN'T OPTIMIZED for the Athlon 64 architecture, therefore you are wasting your time running it it means nothing.

You need version 24 of Prime95:
http://www.mersenne.org/freesoft.htm

It is just another stress benchmark i don,t think it matters whether it is optimised for a64 as long as it finds weaknesses which sp2004 has.. I don,t think it means nothing.Will d/load prime "optimised" to check again and do a stress test for the same amount of time but i am betting it will just be as stable..
Also why would the latest prime be optimised surely it is doing exactly the same prime calculations..
 
Last edited:
daz said:
If you're not Prime stable, you've been returning bad (useless) work units for SETI and folding.

Well ok if you say so but i'd prefer it if you had some sort of proof to back
that statement up.

Currently been running prime now for approx and hour on both my overclcocked systems and so far so good, will leave it a few more hours and if it hasn't failed then i will say it's ok.

My main 24/7 folding pc is completely stock so won't be bothering to run it on there.
 
rxmac said:
Well ok if you say so but i'd prefer it if you had some sort of proof to back that statement up.

Because your CPU is generating errors against known results in Prime, you can be sure it's generating errors when pushed to the limit calculating other results too. The results from such a PC can't be relied upon for scientific research.
 
Anyone ever think that if it's causing errors in prime it would be causing the same problems in games?. Games use calculations for everything...

I bet half these people with 3ghz opterons in there sig's are'nt even stable which means nothing.

Get it to pass prime for a minimum of 12 hours on blend then brag about a good overclock or enjoy your games flawlessly :)
 
You could debate all day about how long you should run prime for to get your system 100% stable.Anywere between 6-24hours(in the faq,s) for me personally i don,t have the patience for 12 hours myself, but i do agree it has to be prime stable in order for it to be a solid overclock ;) ....
 
Last edited:
Hi,

1/6/12/24hours are a good amount of hours to run prime. I agree that people have different opinions of what *stable* means.

Prime95 is very sensitive I find, it can pick up on errors that a lot of other applications do not see. Also Prime95 is just one of many *tools* that you can use to test if your system is *Rock-Solid*.

I normally spend about a week burning a new system in and testing the overclocks are stable, once I'm done with that I just buckle-up and let rip with some gaming :)
 
Back
Top Bottom