Now I have managed to sort out various hardware issues and have tightened RAM timings to get latency down a tad. my Ryzen 9900X is also running very nicely.I have 4 sticks of mixed Intel and AMD Corsair Vengeance and have managed to tweak the timings a little without raising the voltage as additional cooling is next to impossible. The CPU has been delided and I have dialled in the Curve Shaper, OCCT is stable for an hour with zero errors with either RAM or CPU, all software seems to be stable. Is OCCT considered a good enough stability test or do I need to thrash the CPU with Prime 95?
Is OCCT a good stability test?
- Thread starter 13oots2
- Start date
I do use it for fault finding. I recently had random crashes while browsing (maybe once every 2-3 hours) when using an old AMD 3600X CPU. Only when I used the CPU+RAM test in OCCT did I get errors that pointed to a weak CPU core/memory lane.
For me gaming is the ultimate test if that is what the system was built for
For me gaming is the ultimate test if that is what the system was built for
I do use it for fault finding. I recently had random crashes while browsing (maybe once every 2-3 hours) when using an old AMD 3600X CPU. Only when I used the CPU+RAM test in OCCT did I get errors that pointed to a weak CPU core/memory lane.
For me gaming is the ultimate test if that is what the system was built for
Am not getting any errors with A.I or a bit of gaming, the A.I work hits pretty much everything hard. Had problems for months after I installed my new graphics card, HDR 4K would freeze or crash pretty much any media player, would get random freezes or crashes every 10 minutes or so, narrowed down the problem to a poor connection on a SATA cable. I must've moved it slightly when piddling around with the graphics card, gave the SATA connectors a small squirt of WD40 Contact Cleaner and all has been well since. I thought it was system stability, never occurred to me that spinning rust with poor contact could cause that problem.
I use it for testing more for the CPU + RAM as the GPU tests apparently lack behind more bespoke tools for GPU testing.
These days I use it more for giving the CPU a 1hour 'stress' test to make sure cooling is sufficient.
These days I use it more for giving the CPU a 1hour 'stress' test to make sure cooling is sufficient.
With me, it goes back at least 25 years, I used to test overclocked results of SETI results for a large forum SETI team, I would compare a known good submission to an overclocked result. I would use file comparison software to see if they matched, if they did the result would be added to the CPU league table. At the time I had the fastest Duron which was a 900, that was until the Duron 1300 came out. It was surprising how many overclocked CPUs were throwing a couple of digit errors yet appeared stable. It was important that the data being sent back to SETi was correct as a false result could indicate intelligent life out there, considering the state of this planet, that could've been mind-blowing.
Aah Duron. Who needs an IHS when you have little rubber feet?
It didn't occur to me that a chip might "work" but spit out invalid results. That stress testing makes perfect sense.
I'm thinking of guys stress testing their overclocked gaming machines for an hour instead of playing Fortnite...
A certain amount of stability does matter for gaming as well though - those people with CPUs that crashed every few % whilst compiling shaders might have probably seem see similar crashing behaviour in games/activities that are more CPU bound.
Associate
- Joined
- 10 Sep 2006
- Posts
- 2,016
- Location
- UK
OCCT is fine. In my experience I have had it catch errors that other tests did not, but the reverse is also true.
If you want to ensure rock solid stability you would ideally use multiple tests.
If you want to ensure rock solid stability you would ideally use multiple tests.
I use stress testing when I overclock the CPU so to get the CPU voltage as low as possible without it failing.. Because more or pointless voltage means more unnecessary heat.
Last edited:
I get that, but why not just make the CPU do the thing you want it to do (play games/search for aliens/whatever) as a test? Does it save time?
The data fed back to SETI had to be accurate, what is the point of sending back erroneous data, you are not only wasting your CPU cycles but could cause extra work at the other end. The point of overclocking it to get a faster computer that is 100% stable and accurate, in some cases 99.99999% is not good enough, the odd random freeze is one thing, important data another. I want a faster computer but don't want a workload to crash 2hrs into a 2hr 15 minute render, that is a waste of time and electricity.
Last edited:
The CPU isn't working hard when gaming though, your entire pc is working equally, moreso the GPU... So games might run totally fine but crashes when your doing cpu intensive tasks like converting video