Are OCCT Errors BS?
Are OCCT Errors BS?
Testing is at 48 minutes with no BSODs. It’s stopped on the first error.
You might encounter minor computation issues that don<|pad|> a bit more subtle than a crash or reboot. I'm not very experienced with OCCT, but I've used Prime95 for similar checks since it gives me better insight into the problem areas.
Prime 95 displays the error details clearly? I've run several consecutive tests using OCCT to see if issues stem from code flaws. The first test produced errors repeatedly within short time frames under an hour. At 4.1GHz, any voltage below +0.102v (about ±0.1v) caused problems, except at +0.102v where it worked. The second test remained stable for about an hour. At 4.3GHz with +0.174v, it failed (48:21 error). The current reading is 1.29v with a minimum of 1.164v and a maximum of 1.224v, averaging around 1.195v. Today's results are similar. For 4.3GHz at +0.204v, the chip shows no errors after 20 minutes. Testing suggests OCCT is quite reliable. My focus is on large data sets, extreme conditions, auto threads, and AVX2 support. I've heard small datasets stress the cache more, while large ones better mimic real-world performance. Could you confirm the safe voltage range for this chip? I own a Ryzen 5 3600 with precision boost disabled. With it on, I observed around 4.2GHz at roughly 1.4v max, sometimes higher or lower. That’s significantly higher than expected. Clearly, one of the benefits is testing if I can safely undervolt and boost beyond stock speeds, which so far looks promising.
I also felt this kind of worry, thinking about how calculation mistakes might lead to poor performance and frame drops in games. I really enjoy overclocking, but gaming is a big part of me too. It's a tricky balance, isn't it? And by the way—are you a brony? Have you heard of SMOR3S, the musician from that fandom?
I see how to interpret the results to narrow down possible causes, especially when paired with targeted tests. It seems related to Prime95's FFT size testing. At low FFT sizes, data fits into CPU caches, which can overload the cores more. Larger tasks move data to RAM, causing slower performance. Because RAM speed is limited, this mainly affects the IMC and memory rather than the CPU cores. Based on my experience with Zen 2 CPUs, they're designed near their maximum limits at release. For stability across all scenarios, I wouldn't risk under-volting unless you need consistent performance. The safest approach is to lower the power limit only with PBO, as power consumption fluctuates too much for traditional fixed clock/voltage overclocking.
TL;DR - Prime 95 is interesting, but the performance gains aren't worth it. OCCT 4 LIFE! Using blend mode puts a strain on CPU and RAM. I get it, I’ve used it sometimes, but the worker windows are terrible and the program needs a major overhaul. I tend to stick with OCCT and RealBench/Cinebench for reliable results. Overclocking errors seem genuine. At 4.4GHz I’ve been stable for about 40 minutes. Hopefully it’ll hit around 1.38v-1.40v; an AIO would be nice since I’m already near max temps but still have some headroom. The stealth chip seems worth it. I usually run large datasets with AVX2 for maximum speed. Small data sets don’t seem practical under extreme heat, same with Linpack Extreme. I prefer stress tests like RealBench for an hour after OCCT stability, plus Cinebench to check numbers. I enjoy benchmarks afterward a few times to ensure I’m not losing performance. Is the stock boost real? Hard to tell. If you’re after a ready-made fix, it works—but thermal throttling remains a concern. My 3600 ran at 4.2GHz boost around 1.45v+, now I’m seeing a 200MHz jump with slightly less voltage. It’s not huge like my 7700K (4.9GHz at 1.4v-1.5v), but it’s impressive. I appreciate overclocking, and I don’t want to just install a CPU without proper control. The VRMs on the AORUS M seem decent, though the BIOS tweaks are a bit confusing. I’ll keep pushing it, but it really needs better cooling to reach 4.6GHz. Maybe around 4.5GHz if I manage the same voltage or a bit more airflow.