Prime 95 Fatal error!

[Bayonett Priest]

my computer keeps freezing! and i did a prime 95 check and got:  fatal error rounding was 0.5 expected less than 0.4!

what does that mean and how can i fix it? i ahve xmp 2 and was a 4,69 Ghz an dit got quite hot 96 degrees celsius and i have it watercooled with my gpu with 2 360 rads!

should i do a bios update? havent done one i a while!

[WereCatf]

2 minutes ago, Bayonett Priest said:

what does that mean and how can i fix it?

It rather clearly says in the log "Hardware failure detected" -- you need to check if your RAM is defect with e.g. memtest86+ and if it doesn't show any errors, you'll need to lower your CPU's clocks until the error goes away.

[glenalz81]

What does the stress.txt file say?

[Bayonett Priest]

1 minute ago, glenalz81 said:

What does the stress.txt file say?

where do i see that? i see the stress.text fuile but thats only explenation what prime is! do you mean results.txt?

[Bayonett Priest]

11 minutes ago, WereCatf said:

It rather clearly says in the log "Hardware failure detected" -- you need to check if your RAM is defect with e.g. memtest86+ and if it doesn't show any errors, you'll need to lower your CPU's clocks until the error goes away.

i once a month ago needed to reseat the Ram cause the rgb on them went dark and had some problem inserting them again! might have scraped it a little bit maybie at the bottom(had soem problems seating it again cause a coolingtube was over the ram and it was dificould to put them in! heres how it looks! still has enough room to extract the ram though

[TheGlenlivet]

Remove XMP and overclock, then retest.  Does it still fail?

[Bayonett Priest]

Just now, TheGlenlivet said:

Remove XMP and overclock, then retest.  Does it still fail?

ok will try

[Bayonett Priest]

On 12/20/2019 at 7:53 PM, TheGlenlivet said:

Remove XMP and overclock, then retest.  Does it still fail?

yea i still have freezings every hour or so and the message! should i reseat my ram?

[Bayonett Priest]

On 12/20/2019 at 7:35 PM, WereCatf said:

It rather clearly says in the log "Hardware failure detected" -- you need to check if your RAM is defect with e.g. memtest86+ and if it doesn't show any errors, you'll need to lower your CPU's clocks until the error goes away.

sry had stuf to do! i did set the xmp to auto now! and its till happening! how do you run memtest? do i have to buy the pro version? never used memtest! is there maybie a tutorial?

or should i maybie reseat my ram?

[Falkentyne]

1 hour ago, Bayonett Priest said:

sry had stuf to do! i did set the xmp to auto now! and its till happening! how do you run memtest? do i have to buy the pro version? never used memtest! is there maybie a tutorial?

or should i maybie reseat my ram?

You should reorganize the prime95 windows.

Each window has its own thread.

 

What you should do is resize each window side by side, in two rows of 8 (that's how I like it) so you can see which thread actually failed.

Often you will find that one or two threads will ALWAYS Be the ones that fail, which are your weakest cores.  If it's always one thread that fails first, this is almost always a vcore problem.  If it's random threads that fail, then that could be a cache/ or IMC/L3 problem.

 

The way you have it, it's impossible to tell what thread crashed.

Try a layout like this.

 

 

Don't hate at my VID.  VID isn't vcore.  I was messing with SVID Offset and DC Loadlines.

This was the actual vcore.

 

[Bayonett Priest]

8 hours ago, Falkentyne said:

You should reorganize the prime95 windows.

Each window has its own thread.

 

What you should do is resize each window side by side, in two rows of 8 (that's how I like it) so you can see which thread actually failed.

Often you will find that one or two threads will ALWAYS Be the ones that fail, which are your weakest cores.  If it's always one thread that fails first, this is almost always a vcore problem.  If it's random threads that fail, then that could be a cache/ or IMC/L3 problem.

 

The way you have it, it's impossible to tell what thread crashed.

Try a layout like this.

 

 

Don't hate at my VID.  VID isn't vcore.  I was messing with SVID Offset and DC Loadlines.

This was the actual vcore.

 

did a 9 hour torture test while i slept! and this is the result!

[Falkentyne]

20 minutes ago, Bayonett Priest said:

did a 9 hour torture test while i slept! and this is the result!

Yeah you have two unstable cores at that voltage.  You lost three threads 2-3 hours in.

Whenever you lose a thread you should just instantly abort the test and adjust voltage or reduce clockspeed as needed.  No point in continuing a test after a thread drops.

[Bayonett Priest]

2 minutes ago, Falkentyne said:

Yeah you have two unstable cores at that voltage.  You lost three threads 2-3 hours in.

Whenever you lose a thread you should just instantly abort the test and adjust voltage or reduce clockspeed as needed.  No point in continuing a test after a thread drops.

so should i reduce the voltage? i have my clockspeed at 4,7! shoudl i drop to 4,3? i figure i have to set the xmp to manual and set it myself!

[Falkentyne]

You have two options.

Increase the voltage by about +25mv (0.025v) or reduce the clockspeed by 100 mhz and try again.

[Bayonett Priest]

2 hours ago, Falkentyne said:

You have two options.

Increase the voltage by about +25mv (0.025v) or reduce the clockspeed by 100 mhz and try again.

ok n ow i have a strange ocurence! did the test for 3 hours now no errors for now BUT i did set it to xmp 1 and set the core ratio to 45 and voltage to 1,31! but on the stresstest the clockspeed is at 4,05ghz strangely?