Thank you for your help. I think I'm gonna go with 8hrs of prime 95. At least it's shorter! 
That's a good indicator right?
That's a good indicator right?
Whats your testing methodology?
- Thread starter opethdisciple
- Start date
Thats the thing. My OC passed about 4hrs of Real Bench. Both stress test and benchmark. It then did a a run of about 4hrs gaming in the Division and BF1 no issues.
Then after a week, yesterday it wouldn't even load The Division, it was blue screening whilst it was loading.
Then after a week, yesterday it wouldn't even load The Division, it was blue screening whilst it was loading.
Soldato
Set the multiplier to x44.
Set the Vcore to 1.35v
Set LLC to extreme (Gigabyte board, right?)
Set RAM to XMP
Run Realbench for three runs. If all is good and temperature below 80C, then just game.
If you get a BSOD do not up the Vcore, instead just drop the multiplier by x1 to 43 etc..
Set the Vcore to 1.35v
Set LLC to extreme (Gigabyte board, right?)
Set RAM to XMP
Run Realbench for three runs. If all is good and temperature below 80C, then just game.
If you get a BSOD do not up the Vcore, instead just drop the multiplier by x1 to 43 etc..
Dunno what R15 & 26.6 1344k mean?
But XTU looks good. Might go with that over Prime95.
How long should I run the stress test for, shortest time to give me a good level of confidence?
Soldato
Dunno what R15 & 26.6 1344k mean?
But XTU looks good. Might go with that over Prime95.
How long should I run the stress test for, shortest time to give me a good level of confidence?
R15 is Cinebench R15 which is a benchmark. 26.6 1344K is Prime95 26.6 running a custom burn test, in place FFTs with a minimum and maximum size of 1344K.
imo XTU and benchmarks are too easy to pass. I did an overnight run of XTU and crashed very fast in BF1. On the other hand P95 is incredibly demanding and in a not very realistic way. People can pass hours and hours of Prime and then crash opening chrome.
Isn't stress testing just the most fun thing ever
Man of Honour
Nothing beats real world usage, but generally I find that if it can take 12 hours of Prime95 (on mixed) then it's unlikely anything else will upset it. That said, I only overclock CPUs these days: overclocking the RAM has no real-world effect and I was never any good at overclocking GFX. But Prime is pretty good on the RAM side as well.
In light of the Ryzen 8hr small fft stability test, I'm going to just go with that.
I started doing the XTU stress test until you mentioned it was too easy...
I started doing the XTU stress test until you mentioned it was too easy...
R15 is Cinebench R15 which is a benchmark. 26.6 1344K is Prime95 26.6 running a custom burn test, in place FFTs with a minimum and maximum size of 1344K.
imo XTU and benchmarks are too easy to pass. I did an overnight run of XTU and crashed very fast in BF1. On the other hand P95 is incredibly demanding and in a not very realistic way. People can pass hours and hours of Prime and then crash opening chrome.
Isn't stress testing just the most fun thing ever
I do very similar to Doug2507 and it hasn't let me down yet - I use XTU to get a ballpark - once its passing and hour or two of that I up voltage very slightly (alternatively you could drop multi very slightly) and test cinebench - if its passing that so far its not fallen over on anything else. After all these years though I have almost a sixth sense for where actual stability is at and once I've got a ballpark from XTU can finesse it the rest of the way.
I don't use Prime these days.
If the CPU Freq & Vcore is the only thing I have tweaked, small fft is the right setting for me to test stability in Prime?
Running Prime small fft now,. Seems to add 10c more than any other stress test.
Running Prime small fft now,. Seems to add 10c more than any other stress test.
Last edited:
My testing methodology is overclock then play games, I don't see the point in doing endless stress test if you aren't using similar programmes. I don't encode videos etc so doing handbrake test and stuff like cinebench is pointless if I can play bf1 at 5.0ghz but cinibench would crash at 4.9 etc.You can be stable for hours on end in prime but five minutes of bf1 it will crash , as meridian said real world usage is best overall.
That is great if all you do is game and it holds upto gaming - but that doesn't mean the system is 100% stable and corruption could be creeping in somewhere. Even stress testing isn't bullet proof but a much better representation of all around application stability.
Soldato
it's also not so fun when you crash in game and it turns out that your last couple of hours of progress have been lost as the save games are corrupt.