Overclock is totally stable apart from 1 game..

roozor

roozor

roozor

Associate
Joined
1 Jun 2004
Posts
417
Running only 4.6ghz on my 4670k + MSI z97-g43

Primr 95 12 hours stable, AIDA64, all other games eg; Star Citizen, Far Cry 4, Bioshock Infinite.

Dota 2 crashes every time I play it. Full hard lock with "whea uncorrectable error"

If I go back to stock CPU speeds I can play fine. Any advice?
 
Have you tried more volts? Ive found some games are just better at crashing overclocks than other. I was playng crysis and gta for hours the other night, then logged into wow to give someone a hand and crashed in 10 mins, 0.01v sorted it.
 
If it's crashing in a game it's definitely not unconditionally stable. Games aren't even that demanding

WHEA is normally vcore or cache related.
 
Last edited:
If it survives Battlefield 4 it is stable.

Also 12 hours of Prime and you haven't cooked you CPU - bravo that man, mine went from cold to surface of the sun in around 0.25 seconds!

As for advice is it runs fine on stock something isn't stable in your OC.
 
As above, whea event 19 cpu parity error is pretty much often vcore related. What's your vcore at and also what's your VRIN, (cpu input voltage). As a rule of thumb, this should be 0.5v higher than vcore. In my case, 1.300 vcore needs 1.800 on input. Cache voltage on gigabyte is known as VRING, though if running the cache at stock 4000mgz, I cant see it being the issue. Id also advise against p95, haswells run way to hot under this, something they're not designed for.
 
Thanks for the replies guys.

Might try more volts then. Currently on 1.261 core and 1.9 on the ring. Temps are absolutely fine with my new h100i absolute max has been 79 in prime and Intel cpu burn. Games it's only 47 to 50 degrees.
 
Stress tests don't always determine stability. Using the system normally does, use it for day to day things, if it doesn't crash your stable, if it does your not stable.

Asus realbench is good for stressing in real world scenarios. I usually go for 4 hours or 10 passes on the stress test or benchmark.
 
Have a look at your memory if cranking the CPU gets you nowhere. I was pulling my hair out for ages trying to solve some WHEA kernel errors, thinking it was the CPU. Turns out it was running my memory at XMP with my CPU OC meant I needed a little bit more voltage for the memory controller.
 
I've totally wiped settings and started again.

Put the mem down to 1333 at 1.5v.

Went straight for 4.4ghz on adaptive and set voltage to 1.250v. Crashed with the Whea uncontrollable error after 40 mins of Intel Stress Test.

This was actually a pre oc'd bundle from OCUK at 4.4ghz. Not sure what's going on now. Seems to have lost the min overclock it was once capable of!
 
Probably going to be a setting still. Make sure LLC and input are the same as they were

russell664 said:
Stress tests don't always determine stability. Using the system normally does, use it for day to day things, if it doesn't crash your stable, if it does your not stable.

Asus realbench is good for stressing in real world scenarios. I usually go for 4 hours or 10 passes on the stress test or benchmark.
Click to expand...

Real bench is good for people who do a lot of multi tasking and rendering. 2 to 4 hours is plenty enough for most users on the stress test
 
Personally I've ran 5 runs of Asus real bench, (not the stress test) followed by months of gaming and thankfully its been fine. My biggest issue at the start was ram errors, corrected by upping SA and IOA/D volts. Run the cache at stock mostly as its only useful in bences when clocked.
 
X264 stress test is a good testing program

Stress's the CPU in ways real world use will(unlike synthetic benchs Aida/IBT ect).So you dodnt see temperatures from it that are massively higher than you would in real world use(intensive gaming ect).My CPU hits 73c in 264x,while CPU intensive gaming for long periods my max temps are around 68-70c for example

Can set the number of times it will run the loop before running it.Usually i run around 40-50 loops(4-5hrs worth),if it manages that then its stable.

https://mega.co.nz/#!3tAGnAqZ!QbCz2r1fG0WjM8DgGYeExngGypaHftAzPUgTSn2kAdk

^clicky for any1 interested


OP sounds like you may need a little more Vcore,quite low voltage for 4.6Ghz imo,i need same vcore as that for 4.4.And i think that i got a pretty good OC'er
 
Those stress tests are rubbish, I've no idea why people bother.

I've heard of so many people say it's prime stable for 24hrs, yet crashes in a game. Just set a batch encode going in handbrake or ripbot.....if you manage 12hrs of ripbot, you're stable :D
 
Thanks MrMD. I'll check it out. I'll try for 50 loops.

Had to message OCUK to ask for the original bios settings as I can't even get it back to 4.4 stable which they sold it to me with. After I updated bios the saved profile doesn't load.
 
I personally overclock and test overclocks / temperature at my real world setting playing games and i tend to go to my more demanding games like crysis 3, if i find temps to be acceptable and its stable for several hours of play im happy with it and following that principle have never encountered any instability issues after i have said to myself that it's now stable! ( Though this may not be the ideal solution but it's not let me down before )
 
Been testing all weekend and eventually settled on 1.256 volts at 4.4ghz.

Again, every single game is stable. Played hours or star citizen, elite dangerous, far cry 4, bf4 etc....but it's only Dota 2 that will through my system into BSOD.

I'm getting clock_watchdog_timeout.
 
The trouble is when a lot of people are stress testing their CPU it is throttling and they don't realise, so it passes the tests with flying colours running at a canter and they think it's stable. Another factor is that they test their CPU with something like Prime95 all by itself, then when they run a game they've got a 300W GPU pulling load from the power supply and pumping scorching hot air into the case (and onto the CPU).

The best way to stress test a system is to run Prime95 whilst looping something like Unigine Heaven to load the GPU, you have to keep an eye out for any signs of throttling though because it won't just punch you in the face and tell you, a good sign you're throttling is that temperature will jump up and down a lot or will simply be unrealistically low. Run Prime95 with your CPU at stock, make a note of the load temperature and then run it overclocked.. if you don't see a substantial increase in temperature then you're getting some kind of throttling.

clock_watchdog_timeout in my experience basically means that you've overclocked too far and one of the cores stopped working.

Like I've said before, there is no 'definitive' stress tester that everyone should take as gospel - a stable overclock should be able to run everything it can at stock, no exceptions.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom