Yes, it can be viewed as stress testing the CPU under heavy workload.
Yes, it can be viewed as stress testing the CPU under heavy workload.
I'm talking to a friend about his CPU stress test. He mentioned he's running cpuminer-avx2, which made me puzzled. I asked if it wouldn't be better to use software designed for stress testing like P95, OCCT, or Cinebench. He explained that since his current method takes three hours per run and fully utilizes the CPU cores, it qualifies as a stress test. He also said he intends to mine using the CPU itself. I pointed out that simply running any app that hits 100% CPU usage isn't always a proper stress test—it might not push the processor to its limits or generate enough heat for meaningful results. Using a dedicated tool would be more efficient and help catch issues that software might miss. He also suggested adjusting software settings rather than physically tuning the CPU, comparing it to changing game settings for optimal performance. I asked for other opinions on this, as I tend to be firm and could overlook something. This isn't related to the profitability of mining with a 3900x in Australia without solar.
In my opinion, real-world testing is one of the most effective stress tests. Prime95 works well but mainly targets the CPU, not the whole system like triple A configurations. For verifying an overclock, tools such as F@H or mining would be sufficient if you focus on those specific instructions. AVX should also be evaluated (though it's unclear if mining relies heavily on it) since it often puts the CPU under maximum load. Ideally, you'd want to adjust the argument descriptions to highlight potential CPU strain. Setting up AS5 is particularly time-consuming, requiring around 200 hours of burn-in testing.
the method of mining influences performance, there are several instruction sets a CPU supports, i found realbench useful for a full 48-hour stress test combined with other evaluations to assess stability.
Stability really depends on the person. If I can handle WPrime 1024m using an older s939 chip and it runs about 25 minutes at a 50% overclock, then it must be stable enough. lol. But that's just about checking stability during benchmarking. I don't mind if it isn't running Crysis right now.
Yeah, that's a fair take. Burn-in isn't really about stress testing in the traditional way. If software were a good stress test, people would use it mainly to check overclocks—which is something very few actually do. At least according to what I know. First, you overclock and then use dedicated tools to stress test it, and finally confirm with real-world applications.
This might be the setup you prefer trying. Some individuals have unique preferences and make varied selections. Your choice is entirely yours to embrace their beliefs. I favor OCCT and IBT primarily, but of course everything else is fine. Of course, this applies only when seated at the bench. You must generate stress to produce heat for the burn. In my view, it's just wordplay. It's similar to TDP—really focusing on Thermal Design Point. The key term to keep in mind is Thermal, which is often misunderstood as power consumption. The other two elements are irrelevant.
Thank you all for your quick responses. It looks like I made a mistake and I should apologize...