View Full Version : What is considered "Prime Stable"
scrible88
02-28-2005, 03:16 AM
How long exactly do I have to run Prime95 for my system to be considered "prime stable?" I'm running it @ my 24/7 settings which are in my sig. It currently has been going for about an hour error free. I am running the "blend" test.
Do I have to run the "blend" and "maximum heat and power consumption" tests for a long period of time to be considered "prime stable?" How long should I run each?
I've been overclocking for years but have never really tested a system to see if its 100% prime stable.
Highland3r
02-28-2005, 03:27 AM
It varies from person to person... Usually give mine ~ 6 hours if it passes that then I'm happy. If theres hangs/crashes etc in games though it gets a longer run 10+ hours just to make sure...
scrible88
02-28-2005, 03:41 AM
It varies from person to person... Usually give mine ~ 6 hours if it passes that then I'm happy. If theres hangs/crashes etc in games though it gets a longer run 10+ hours just to make sure...
I'm going to let mine run while I'm at school/hanging with friends this evening. It should be going for a good 12-ish hours.
It has never had any trouble with any games, etc.
Napoleonic
02-28-2005, 04:09 AM
IMO, small fft test would be more stressing the cpu while blend is for ram, cmiiw
dnottis
02-28-2005, 05:33 AM
Small FFT will die faster if the OC isn't stable - I will run that for just about an hour to check CPU stability, but then I blend test for about 19 hours just to be sure :)
cantankerous
02-28-2005, 11:41 AM
To me if prime passes at least 1 hour I am happy. The rounding errors you get in prime will never been seen on games etc as they don't round off as critically as prime does. I have had prime run forever just to have 3dmark etc crash on me. I usually run memtest to be sure that is stable. Then I run prime for about 1 hour as on a really flaky system it will fail almost instantly. The rest I rely on 3dmark runs and 32M super pi.
pudds
02-28-2005, 12:13 PM
Well ive had prime errors at the 20 - 24 hour mark so i run for at least 48 hours. Once on max heat max power and once on blend. Also memtest numbers 5 and 8 for 24 hours each.
[XC] moddolicous
02-28-2005, 12:32 PM
This is what I do:
Prime (small fft) for 6 hours
Memtest for 6 hours
3D01 looped for 6 hours
And maybe superpi 32mb. This always produces the most stable system for me. If you running a server, maybe multiply each of those by 4. If your comp can pass these, then it should be stable in games.
jinu117
02-28-2005, 02:01 PM
moddolicious method is what i use pretty much :) Only thing is, I add about 6 hours of straight CS:S on it. Seems to be most demanding game or time being as it crashes system faster than any above test will do if I did overclock too high up.
bushmeal
03-01-2005, 05:20 AM
Large FFT is a more stressful test IMHO. I only test large FFTs. I can easily pass prime using small FFT's. The tough one to pass is the Large FFTs test. The small FFT's test just doesn't stress (heat + mem) the system enough. Happy priming.
mdzcpa
03-01-2005, 05:20 AM
Blend overnight.
NiCKE^
03-01-2005, 05:25 AM
Is there any diffrent by running GO SP2004 and Prime? Since it looks the same and you can choose the same.
If it can fold for a couple days, it's good enough for me.... :hehe:
dnottis
03-01-2005, 06:23 AM
I was just testing a new CPU - damn thing primed for just over 12 hours, SuperPI 32M with no problems - but 15 minutes into Battlefield Vietnam the system hard locked - just goes to show you that TRUE stability may not be experienced until you actually USE it!! Ah to hell with the prime95 and SuperPI 32M theories.
BTW dropped it down 5MHz on the bus and I was able to play BF Vietnam for about an hour and a half afterwards with no problems.
I was just testing a new CPU - damn thing primed for just over 12 hours, SuperPI 32M with no problems - but 15 minutes into Battlefield Vietnam the system hard locked - just goes to show you that TRUE stability may not be experienced until you actually USE it!! Ah to hell with the prime95 and SuperPI 32M theories.
BTW dropped it down 5MHz on the bus and I was able to play BF Vietnam for about an hour and a half afterwards with no problems.
Fully agreed. Prime95 and Super PI stable is no big deal.
I too face the same problems you mentioned. I can run Prim95 TT (I run Small and Large FFT concurrently using SP2004). It can prime for 24hrs straight but when I play D3, the game hang after 30mins. Lowering CPU clockspeed solves the problem. So much for Prime95.
I also face problems where I am running 2 x Prime95 but my IE/Firefox crashes. Prime didn't report any error though. It continues to run fine.
Take note that Running Super PI is basically a division exercise. Its basically 22/7 and find the result up to 32 million decimal places. Prime95 is factoring massive numbers to find prime numbers. Its basically division too. when you play games, its less of that but mainly addition. Its takining your keyboard/mouse inputs and calculate coordinates and draw the wireframe model. There is also the trigonometric functions which are totally different from 4 basic mathematical functions. Many others.
Also, Prime95/Super PI involves working with a single large number. Its very difference from games which work with numerous small calculations (calculating coordinates of the wire frame model). There is also enemy AI etc.
Thus, bottom line is it doesn't mean your comp can run Prime95, its stable. Its just a common misconception that when you comp can run Prime95, its stable. I stilll use it but I don't rely solely on it.
Having said that, there is no such thing as 100% stable. No hardware and software is perfect. Bugs are everywhere even on CPUs (check out AMD and Intel's errata document).
Stumpjumper5200
03-01-2005, 07:06 AM
I could care less if my system can run Prime95 or some other torture test for a billion hours, that's not what I do with it. I play games, so if it doesn't crash when playing games, I'm happy.
bushmeal
03-02-2005, 07:00 AM
My experience is that if I test and pass memtest, prime and 3dmark, I'm usually stable everywhere else.
Some people in here don't know how to build and overclock a stable system it seems.
You have to test 2d in prime95. 24 hour torture test is what I run. That ensures cpu, memory, and chipset are fine. Then.....
3D is a whole different round of tests.
I've got to just say durrr on the belief that if it's prime95 stable it's 3d stable.
Just because it is prime95 stable doesn't mean your system has the power to push a high end overclocked 3d card too.
dnottis and ISS your belief that prime95 somehow makes your system fully stable in all task is your problem not the program itself. The program is telling the truth. And if your running prime95 and other programs crash, Guess what thats telling you...
Yep, your not stable.
alexio
03-02-2005, 08:13 AM
What about running 3dmark05 and Prime95 at the same time, looping it for 8 hours or so? That would ensure 100% system stability I think.
That would allso be nice for testing if your psu is a bottleneck in stability.
drftkng
03-02-2005, 08:34 AM
What about running 3dmark05 and Prime95 at the same time, looping it for 8 hours or so? That would ensure 100% system stability I think.
That would allso be nice for testing if your psu is a bottleneck in stability.
That would seem like a pretty good idea to me. I have run SuperPi in the 32meg test while running that Realtime Lighting Demo (the one with the pretty shinny balls) Figured that would be a good test also.
wfarid
03-02-2005, 10:24 AM
do any of you guys dual prime? like run 2 primes at the same time?
pudds
03-02-2005, 10:27 AM
do any of you guys dual prime? like run 2 primes at the same time?
I never see the point in that as there both doing the same calculations and just nicking CPU time from each other.
cadaveca
03-02-2005, 07:26 PM
I prime95 with small fft's, and run either the crytek mechanima or rubyx850 demo, with a continuous ping to another pc in my network. This way you really push the chipset. running just prime 95, is not enough, unless you are going to run in for a week. One of the other testa will complete use your ram, and that one's good too...but small FFT's will crash a system before anything else, so far that i have seen,
Some people in here don't know how to build and overclock a stable system it seems.
You have to test 2d in prime95. 24 hour torture test is what I run. That ensures cpu, memory, and chipset are fine. Then.....
3D is a whole different round of tests.
I've got to just say durrr on the belief that if it's prime95 stable it's 3d stable.
Just because it is prime95 stable doesn't mean your system has the power to push a high end overclocked 3d card too.
dnottis and ISS your belief that prime95 somehow makes your system fully stable in all task is your problem not the program itself. The program is telling the truth. And if your running prime95 and other programs crash, Guess what thats telling you...
Yep, your not stable.
I never said I believe that prime95 makes my comp fully stable. I am just quoting examples of the problems I face with a so-called Prime95 stable PC. If your comp is Prime95 stable, its, well just that. Prime95 stable. Its just that CPU could factor huge numbers w/o problems. Thats it. As mentioned, the FPU consist of more than just 4 basic mathematical functions. Prime95 does not make use of every function inside the FPU. In fact, we will hardly fund a program that could stress every single function in the FPU at once. and thats just the FPU. The CPU comprises of more than just FPU.
AMD and Intel has a reason for marking their CPUs are certain speeds. Their suite of testing programs is far more comprehensive than anything we have and while we are basically testing and ensuring the CPU is stable to run our apps, they have to ensure the CPU is stable at literally anything throw at it.
HiJon89
03-02-2005, 09:08 PM
For complete system stability checking, I run two instances of Prime95 Large FFT's and run the ATITool artifact checker at the same time. The ATITool artifact checker gets my video card MUCH hotter than any game and at least 5C hotter than RTHDRIBL.
perkam
03-02-2005, 09:21 PM
Some people in here don't know how to build and overclock a stable system it seems.
LOL Had MDZCPA not posted before PRIOR to that post I might have overlooked it but that was way out of line. If you're trying to help or know something we dont, pls dont get :banana::banana::banana::banana:y.
As for being Prime Stable, I believe OPP posted once (dont know if he was joking lol) that if the pc is prime stable (i think he meant blend) for 30 mins its fine. Which can only mean about an hour or two of prime95 should do it.
Though if you're running it for 24/7 stability your temps over time will be the main factor. It could run the CPU at x ghz on a cold day for 24 hours and *blip* change in weather, ambient gets too high and you've got a very snazzy looking paperweight.
For benching, prime stable usually means prime stable, but only a 3d bench will tell if its bench stable lol.
I'd pm Fugger/OPP on this and they'll dissect the matter much better.
Perkam
kappa2001
03-03-2005, 05:42 AM
I couldn't care less about prime.
I run it a few times using max heat and it would give errors within 3 or 4 minutes.
What is curious is that running @ the same speed i used for prime i can game for the whole day if i want (games like CS: Source or SOF II )
Perkam sorry if you take my statement as ":banana::banana::banana::banana:y". I'm just trying to stop new people from believing some peoples lies that if there cpu isn't calculating math correclty somehow it's stable.
Every cpu since my first 200MHZ MMX I overclocked has passed these test I run has been rock stable and has never failed me. If other people want to have data corruption problems and system stability problems while running there ram and cpu at speeds that it can't calculate correctly, go right ahead.
I can tell you by FACT if it falis prime95 your cpu and ram aren't working correctly and you WILL have problems whether you can distinguish them or not.
As far as Oppainter goes.. Yes he's the best world record 3dmark overclocker in the world but calling a system stable after it runs prime95 for 30 minutes is a joke. Maybe it's stable for his purposes but it's not a verified stable system.
Obviously an overclock will differ between the summer and winter if temps are a lot different. No reason to say F it though. It only takes 24 hours to reverify stability. If you leave yourself a little headroom, the 10f difference in temps won't matter.
Actually, there is no such thing as a stable overclocked system. When you overclock your CPU, you lose stability, simple as that. The important thing, however, is how much stability you are willing to sacrifice in order to gain that extra speed.
Everyone has their own view of system stability and there isn't a need to argue which is considered stable cause none are completely stable (none of our overclocked systems will be able to survive in a heavy sever load environment, neither can it run 3d rendering for weeks or even months lilke these workstations used to render CG scenes for movies).
The most important thing isn't to make sure your comp can run Prime95 or 3Dmark for hours. Its whether it can run all your apps and games w/o problems. There is no point in having a comp that runs 3Dmark or Prime95 ir any stress testing software for days but crashes when you are running your apps/games.
SPL15
03-03-2005, 08:58 AM
For stability tests I run prime but I do a custom test with 1024 FFT and use 512 megs of ram and I also run Folding or Seti and I run ATI tool and use the 3d hairy box thingy. Sometimes I will run my antivirus checker too because it uses a lot of hardrive reads while doing all of the above.
I run it for a few hours and I figure if it can handle that for several hours it can handle pretty much anything. Some people say run it for a day, but... its a whole day and I dont like waiting.
nebuchanezzar
03-03-2005, 09:11 AM
The eternal "what means stable" question. I like prime( actually I hate the evil :banana::banana::banana::banana::banana: but it works pretty well) and I usually us Go Prime or SPrime whatever you call it with the different gui cuz when it fails it has a sound I can hear when cleaning or goofing on my guitar or whatever. When just getting into the high OCs I usually run small FFTs as thats basically JUST the CPU. If that fails I back off. Once I find where thats cool then I run blend or large FFTs(or custom 1024 -4096) and usually leave it overnight. I'ld prefer 8+ hours but 2 or more is OK by me. I've also run prime and ATiTool and media player all together cuz on my present infinity my sound/usb drops at anything over 240FSB within an hour. For gaming stability i might loop 3dmark01 for like 25 runs without the screens between tests but usually I just start playing whatever game I feel like and save a lot :D . This is such a subjective question. If your games don't crash and you're cool with how it runs...it's stable. Just cuz I like 8+ hours of prime doesn't mean you need it. A really flaky system will usually crap out in under 15 minutes anyways.
vBulletin® v3.7.0, Copyright ©2000-2008, Jelsoft Enterprises Ltd.