What exactly makes prime95 torture test error?

[hadbeen]

So I've been trying my hand at some OC'ing and running prime95 to verify the clock setting's stablity. I notice that once I get my A8-3870k to 3.7ghz (on air) I get an error in prime95 saying "FATAL ERROR: Rounding was 0.5, expecting less than 0.4". I am wondering where the error comes from and how I can get a more stable system at 3.7ghz. I already have bumped the CPU core voltage to 1.504v but haven't messed with any of the other voltages. I have a bit of room to still raise the CPU core voltage and keep at the temp below 70C. Is this a memory error or is this the CPU erroring up? I am open to any suggestions and thank you!

 

[Josh Weston]

No, it just means that your CPU is not stable. Not enough voltage at a certain clock shall make the chip unable to perform, and it will start to mess up in its calculations, which is all Prime 95 is.

 

[jamesriley94]

Yeh, anything like FATAL ROUNDING, or worker stopped, or BSODs just mean the CPU (or RAM) isnt stable. Take the speed down or increase the voltage

 

[Ru5ty_t00l]

So I've been trying my hand at some OC'ing and running prime95 to verify the clock setting's stablity. I notice that once I get my A8-3870k to 3.7ghz (on air) I get an error in prime95 saying "FATAL ERROR: Rounding was 0.5, expecting less than 0.4". I am wondering where the error comes from and how I can get a more stable system at 3.7ghz. I already have bumped the CPU core voltage to 1.504v but haven't messed with any of the other voltages. I have a bit of room to still raise the CPU core voltage and keep at the temp below 70C. Is this a memory error or is this the CPU erroring up? I am open to any suggestions and thank you!

did u have any BSODs ,freezes or crashes so far?
if not I wouldn't worry about fatal errors on prime.

 

[jamesriley94]

did u have any BSODs ,freezes or crashes so far?
if not I wouldn't worry about fatal errors on prime.

Fatal errors still mean the processor cant perform the way it is meant to. What is a fatal error on prime could be a BSOD on a game or a different stress test.

You need to be testing for at least 12 hours with no errors whatsoever to ensure stability.

 

[Ru5ty_t00l]

can't that fatal error be because of incompatibility issues ,malfunction or sth? I don't know , just guessing

I have stressed my cpu for more than 12 hours twice completely stable but still had a bsod on a non stressed situation.

 

[Josh Weston]

Yeah, these stress tests don't prove everything, but they give you a good idea of stability. The fatal error is generally the CPU saying something's not right. Not every error shall result in freezes and BSODs - that's the whole point in having stress tests that give error messages.

 

[airdeano]

....I already have bumped the CPU core voltage to 1.504v but haven't messed with any of the other voltages. I have a bit of room to still raise the CPU core voltage and keep at the temp below 70C.

you might benefit from reading a couple of proper overclocking guides on the AMD
platform. there are many other attributes that will need attention rather than like
the Intel of multiplier and voltage. AMD use northbridge voltage and multipliers
and other realms that need tweeking for proper stability. just a quick 1 second
google and i count 8 guides...

airdeano

 

[mrapoc]

Any error will point towards instability of either the cpu or the ram (or anything else which has been modified/clocked actually such as an undervolted north bridge or even a psu which is on its way out i.e supplying weird voltages)

So take the prime95 test as a stress test not for just cpu, as it indirectly will test most parts of your system. Once you get an error, take it as a pointer that something is wrong and either investigate or use trial and error

 

[hadbeen]

are we talking CPU core voltage? I had it up to 1.6v and I was still getting the error..runs 3.6ghz just fine with 1.504v touches 60C with hours of prime95..I don't want to try increasing the CPU voltage much past 1.6v..I feel as if it would get too hot

 

[jamesriley94]

Woah! Get it down. you shouldnt go above 1.5v or it can kill the chip...

 

[ChrisGlobe]

Woah! Get it down. you shouldnt go above 1.5v or it can kill the chip...

This is AMD, not Intel

AMD chips can take a heck of a lot more voltage than Intel ones, so 1.6V isn't too bad. However, I'd be looking at dropping it to 1.41 if you can, and seeing if it'll boot, as that's within the "official safe range" provided by AMD: http://products.amd.com/pages/DesktopAPUDetail.aspx?id=30&f1=&f2=&f3=&f4=&f5=&f6=&

At 1.41v, you then need to start looking at increasing the North Bridge voltage.

Take a look at this: http://www.hardwarecanucks.com/foru...amd-a8-3870k-unlocked-llano-apu-review-4.html

 

[jamesriley94]

I know AMD chips take more voltage than Intel...but 1.6v is way too much. And yes, it can cause degradation at that kind of voltage. I'd be putting it to 1.5v maximum - preferably under that if possible.

 

[hadbeen]

I know AMD chips take more voltage than Intel...but 1.6v is way too much. And yes, it can cause degradation at that kind of voltage. I'd be putting it to 1.5v maximum - preferably under that if possible.

so maybe it's another issue making me fail the torture tests? either that or the chip is just not designed to run like that at 3.7ghz..it was $100 for an chip with integrated graphics...I guess I don't expect it to be amazingly powerful

 

[ChrisGlobe]

Said a few times in this thread now, you can't just up the CPU voltage, you need to go the North Bridge and a few other things too.

Do that, then come back and tell us what's happened. There is a guide in my post above, and reference to others on the first page.

 

[hadbeen]

Said a few times in this thread now, you can't just up the CPU voltage, you need to go the North Bridge and a few other things too.

Do that, then come back and tell us what's happened. There is a guide in my post above, and reference to others on the first page.

I am waiting on a thermometer gun to verify that the 90-92C temp2 reading on speedfan is just a faulty/nonexistent sensor..I don't think that there could be a temp in there that high when I'm idling..but when I increase NB voltage that tends to go up too..so just want to make sure it's not something serious then I'm gonna let the voltages loose.