Prime errors

Andy298

Andy298

Andy298

Associate
Joined
14 Jun 2006
Posts
847
Location
Macclesfield
Trying to get to 2.3Ghz and got this error after 10 minutes...

FATAL ERROR: Rounding was 0.484375, expected less than 0.4
Hardware failure detected, consult stress.txt file.

Upped the volts from 1.450 to 1.475 and got similar error

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

Upped volts to 1.50, ran for about 30 minutes and fell over with a different error

FATAL ERROR: Resulting sum was 5318893326385496, expected: 5318893326647399
Hardware failure detected, consult stress.txt file.

Not sure if I should try to up the volts a bit more whilst still on stock cooling. I've had the rounding errors before when trying to get to 2.2 on 1.425 volts but managed to get past them by upping volts. Any ideas what this error is?
Sorry if it's a stupid question but pretty new to overclocking :o
 
your using a crap program for testing.

you should use memtest and goldmemory for ram testing. and use snm for cpu testing. you can find your max stable overclock in around 2 hours max using the above tools rather than use crappy prime and wait 48 hours to find out what snm will tell you in under 2 hours.
 
I strongly dissagree, prime95 is one of the best stress test programs in existance.
If raising the volts doesn't cause it to pass, try lowering the cpu multi to rule it out and i bet it will still fail.

This suggests that either ram or htt has gone too far, check your htt multi is 4, or 3 if your over 250fsb.
Check using cpuz that your ram speed is below 200mhz, if not put it on a divider.
 
Cyber-Mav said:
your using a crap program for testing.
Click to expand...

Ignore this, Prime95 is considered standard for Stress-testing, and is widely praised as the best stress testing program avaliable for the CPU, with 24 hours on all tests being considered 100% stable (some people say more/less)

You'll want to run the Max heat test (uses little RAM) to see wether the problem resides in your RAM or the chip. I assume you're using the 'blend' test at the moment... that usually chucks out rounding errors quickly if the RAm isn't stable.

I found that i couldn't beat 2.2Ghz on my 3000+ until I dropped the RAM divider and the HT multiplier to 333 and 3x respectively.

Also watch the RAM timings, you'll want to slacken them off a bit (to 3-4-4-10)if you're Overclocking the ram (above 200Mhz if it's DDR400 = overclock) then tighten them up when you're happy with the CPU speed. If you want better speeds at tighter timings, the rule (as always) is more volts=more stable... some RAm doesn't like high voltages, some does, it's best to do some research and discover what yours can take...

I run mine at 2.6Ghz, with the RAM as DDR333 which gives a FSB of 290, Effective RAM speed of ~235Mhz, and i've got the HTx set at 4x (with many more volts, i'd recoment leaving this at 3x unless you've got good chipset cooling)

Use a program like CPU-Z or PCWizard from CPUID to watch the speeds (i'd recomend CPU-z.. it's the one you see posted as proof of everyone's overclocks, thanks to it's handy 'Verify' feature.

It's funny how Overclocking has evolved, It used to be jamming a signal generator onto the Parellel bus and clocking it up until the thing shut down, then running at just below that with a quick mod to jam an extra 0.2v through everything.. These days we play with voltages, HTT, Timings, FSB, Multipliers and god knows what else to get a decent OC... Still fun though

Goodluck :)

Oh, and watch your temperatues if you're incresing voltages... Use Speedfan or somthing to measure them..

If you need any more help, just give me a shout (I've been playing around with the 3000+ for a while now so should have no trouble helping you break yours :p)
 
Last edited:
Thanks for the replys guys.
Mav, I have to disagree with you regarding prime. I'd say most of the folk on this board use it, whilst I've only ever seen a handful of overclock posts from people who use SNM.
My HTT multi is on 3 at the moment. 4 is the default as I'm running a skt754 and max bus is 800MHz. My ram is on a divider also at 166 (DDR333) and temps are at 42ºC under load. At 2.2Ghz with 1.450v Prime was stable for well over 20 hours. Prime test I'm running is large FFT's, max heat and power.
Mikebert4, I might drop you a PM if I get stuck :p
 
i pity the fool who uses prime as a cpu tester.

if you dont believe me then go read up on the XS forums, those guys really know their stuff, and if you think prime is a better cpu tester than snm then your a fully qualified professional fool. nuff said. :rolleyes:
 
must agree with Cyber_Mav in this case even tho he lacks tact..

personaly.. i had an experience where it took 21 hours for prime to crash a pc i was testing whereas snm managed to do the same in 7 mins flat!!! Says it all really!

@Mav -- LOL BRING BACK THE A-TEAM
 
I think it all depends on whether you want 100% stability, for the folding team a calculation of 2+2=4.0000000000000000000000003 just isn't good enough because it'll knacker their results. On the other hand I've run a bucketloads of pcs with overclocks which aren't prime stable but perfectly fine for 24-7 windows and gaming.

Personally I try about an hour of prime max, if that fails immediatel I assume its really dodgy, if it manages a fair while I'll give it a couple of superpi runs and if it passes them its generally safe enough for me.

100% stable really isn't necessary for most people :)
 
Mr Men said:
That post is from 2004? got something more up to date?
Click to expand...

lol.. Mr Men, while we wait for Cyber_Mav's reponse with something more uptodate... why don't you give snm a try.. after all whats the worst it could possibly do to you... blow your cpu? blow your board power mosfets? blow your PSU? Thats the basic risks.. nothing to be afraid off! ;)
 
Last edited:
I guess its down to personal choice. I prefer Prime to SnM as it is more sensitive to errors than snm(imo). As for XS, you will find just as many people who use prime as those who dont, probably more in actual fact.

People who attack prime are usually those that cannot pass it and then say the program is crap while kidding themselves they have a stable system. Personally I'd rather have less bragging rights and be safe in the knowledge that my system is stable.
 
matt100 said:
I'm probably going to look irresponsible if I suggest 1MB superpi aren't I?
Click to expand...

hmmm.... errrrrm..... yes. :p
well... a 32M run can deem your PC pretty stable.

though the phrase 'Prime stable' has become THE standard for a -stable- system.

you can almost gaurnatee that a PC which is stable on Prime for 24 hours on all three tests will not fail. Nor will it chuck out the old 2x2=4.000000000001 which rounds to 5 error which screws up the 'ole FAH stuff

SNM is a good program IMO, but I just prefer to have that absolute certianty that my system is stable.
 
using prime to test for cpu stability is just as stupid as watching I Robot in media player to test gfx card overclock.

http://www.extremeprometeia.com/forum/showthread.php?t=2512

scroll to the bottom of this page:
http://testmem.nm.ru/snm.htm

look at the chart. snm is the only program that stresses a cpu out so it actually reaches its maximum design TDP.
only a fool will try to argue against the facts. snm is THE best cpu stability tester ever created.

Prime is probably the best program ever created to hunt for prime numbers.
 
so......why dont you just use both?? if an overclock passes prime and snm then id assume it is most definatly stable :)
 
only one problem. when i ran prime on my old system it passes over 35 hours with no errors with my cpu clocked at 2.5ghz.

i then found out about snm over at the xs forums and decided to give it a try and see how it went. ran the fpu test and withing 7 mins the pc crashed. could only get snm stable at 2435mhz.

after that experiance i gave up on prime altogether. and its not just me, other people i know used to be prime users but have converted to snm. snm all you have to do is run fpu test for no more than 2 hours and if it passes your pc is rock solid on the cpu front.
 
You must log in or register to reply here.
Back
Top Bottom