PRIME95 error message???

[6TNINE]

i keep getting an error that says that there is something wrong w/ the rounding that .49999 > .40. i have no idea what is wrong? any help?

 

[MrThompson]

You need to back off on your overclock.

 

[6TNINE]

i dont get that? i just got prime95 today, i've been running at 1473 on my 1.2 axia y for about a month now. havent ever had a problem? how is that not stable?? i just dont get it. when i drop down to 1.2 it works fine. at 1.47 i get that error on the first try. thats crazy??

 

[MrThompson]

Prime's self test compares calculations with your system to a known database. If there are errors, your CPU is not totally stable. Your system may run your daily applications but if you do a CPU intensive task like ripping MP3s, there will be errors. Prime is a valuable tool for finding the limits of your overclock.

 

[cbuchach]

Prime95 is the best test I know of for testing if an overclock is in fact stable and the CPU making correct calculations. I fear too many people are in your boat who think their CPU is stable and overclocked to a reasonable extent when in fact it IS making calculation errors.

Some may be able to live with this for a slightly higher benchmark score, but personally I can not and use Prime95 to validate any overclocking I do.

 

[AvidOC]

can someone give me a link to this Prime95?

 

[redpriest_]

Make sure you haven't set your memory timings too aggressively. Your CPU might be stable at that speed, but being fed incorrect data. My memory does this too, even at near default CPU clock speed, if the memory is timed too high.

 

[NightTrain]

Prime95

 

[Garet Jax]

6TNINE,

Before backing off your overclock, you may try increasing the voltage to the CPU. I had the same problem when running my CPU @ 1.6. When I upped to 1.7+, the problem disappeared.

Having said that, if your voltage is maxed out, then you will probably need to back off on the OC to correct the problem.

 

[Richardito]

Basically it can either be your CPU and/or your memory. If it is your CPU feeding more voltage may fix it. If it doesn't then you need to lower your overclock until it gets stable. If it is the memory you need to back off on the FSB speed and/or lower the timing settings from CAS 2 to 3. You've overclocked it too much.

 

[MrThompson]

You should also try increasing the IO voltage if the memory is suspect.

 

[compuwiz1]

I've held off on these threads, simply because I don't want to have to debate this again, but I do not use Prime95 for CPU testing. On 1.2GHz and up CPUS, I have had it error out at default speed. Also with a bunch of test rigs, connecting them all to the net is not feasible.

FWIW, CPU Stability test 6.0 can and has failed chips that passed Prime95, on Intel Chips. Don't know what it is, but anytime you are using something that has to use the internet to do the work, you can have hiccups.

Prime95=1995, way before Tbirds.

 

[Dill]

have a link to CPU Stab. test 6.0 'whiz?

 

[compuwiz1]

CPU Stability Test 6.0 on this site.

 

[Mapidus]

<< I've held off on these threads, simply because I don't want to have to debate this again, but I do not use Prime95 for CPU testing. On 1.2GHz and up CPUS, I have had it error out at default speed. Also with a bunch of test rigs, connecting them all to the net is not feasible.

FWIW, CPU Stability test 6.0 can and has failed chips that passed Prime95, on Intel Chips. Don't know what it is, but anytime you are using something that has to use the internet to do the work, you can have hiccups.

Prime95=1995, way before Tbirds. >>



Weird. Maybe it is not Prime95 that is in error. The factory test might not be good enough -- remember the Pentium III 1.13GHz fiasco where Intel shipped chips that supposedly passed their factory tests, but these chips faild under certain situations in real life. Same thing with the original 500MHz G4s that had timing problems that the factory test did not detect.

Prime95 does not need network access to do the torture test.

For stability testing though, I would not rely on one program, especially since each of the ones available only test certain units and instructions. Prime95 stresses the ALU well, but not the FPU. I believe CPU stability test is more comprehensive in testing both ALU and FPU in the latest versions. I also like to use 3dmark in addition to the above programs because it tests things such as MMX and SSE.

Like Compuwiz, I trust CPU stability test more because I have see situations where it showed errors, but prime95 ran fine -- most likely because Prime95 tests only the ALU. I have also see systems that passed both Prime95 and CPU stability test, but failed certain tests in 3dmark because of bad SSE units.

 

[MCS]

I think you'll find that Prime95 does in fact stress the FPU.

 

[compuwiz1]

<< Prime95 does not need network access to do the torture test. >>



Please explain further. I was of the belief that it had to access the server to run.


I can say that I don't generally get any complaints with regard to stability testing. I do also run 1 hour or 3DMark 2001, just to verify that the CPU can get that far OC'd, then onto CPU Stability Test 6.0 for 12.

 

[MrThompson]

Just select &quot;Torture Test&quot; from the options menu. It works fine without access to Prime's server. FWIW, I don't consider a system stable unless it has passed the torture test.

 

[MCS]

I just use the Self Test, takes around 16 hours. The Torture Test just loops the Self Test as far as I am aware?

 

[Cherub]

I have found that Prime95 will give errors even when 3DMark2001 runs perfectly, and as many people here can testify, 3DMark2001 just exists to crash. In fact, I recently ran every test program I know of, many claiming to include Prime tests, and never had any problem. Yet I still got errors in Prime95.

However, sometimes even a tiny adjustment can make a system stable for Prime95. I received immediate errors when running my 1 GHz AXIA Y at 1450, but I can run for hours at 1425.

(On a similar note, 155 fsb ran in every program I have, except the 3DMark programs. It's good to have programs that are particularly hard on one aspect of the system, for targeted tests.)

 

[Mapidus]

<< I think you'll find that Prime95 does in fact stress the FPU. >>



It does use the FPU, but I'm really not sure how much it stresses it. The Lucas test portion of it appears to use the FPU a lot. I use several programs to test stability(at least P95, CPU stability, and 3dmark) because there does not appear to be a single program that seem to be reliable in finding error by itself. I have seen cases where Prime95 did not fail but CPU stability failed. I do not recall the reverse, but there are probably cases where it would occur. The oddest case I have seen is one where Prime95 ran for 48+ hours without an error, but WinAmp would only run for about 10 seconds before freezing up the whole system.