My guide to boosting performance shows great gains with the Ryzen 5900X — finally, I feel like I cracked the code!
My guide to boosting performance shows great gains with the Ryzen 5900X — finally, I feel like I cracked the code!
All processors set to 4.6GHz without SMT. PBO, Dynamic OC, and turbo boost turned off in BIOS for a clean static setting where all 12 cores run at the same frequency. VCore 1.25 with L4 LLC on an Asus Dark Hero board. Highest L5 LLC and lowest at 1, giving top performance. After OCCT tests, no issues—passed all checks, including Blend and Small FFT portions despite heat. Runs lasted about 20 minutes; kept windows open to lower house temperature during stress tests. Temper stayed under 80°F even with high heat, thanks to cooling. Stability was excellent, no errors in Prime95 or WHEA tests. Ran LinPack XTREME and Prime95 for over an hour without problems. Even with AVX disabled, temps stayed in the low 70s to low 80s. Successfully cleared Y Cruncher and other benchmarks. Temps reached mid-80s without exceeding specs. Passed multiple LinPack runs and memory tests. I’m curious if recent AMD Ryzen 5000 CPUs have improved binning or if I got lucky. The results are promising, but I wonder about the stability of newer models compared to older ones.
EEEH, my PBO boosted 5900x gives better results. I see 4.65 across all cores during Cinebench and mid-80s temps. On a single thread, I usually hit 5GHz or more according to hwinfo64. You're hurting single-thread performance by forcing manual CPU overclocks when it can run at or above 4.6 all cores under PBO load.
I’m fine with that. I prefer solid cooling without tricks like PBO and appreciate a low voltage setting along with 4.6GHz being more than sufficient. Why is it said AMD CPUs struggle to overclock with these specs? Do you have any information about the VRM temperature or are you using air or liquid cooling? I’m using a Noctua NH-D15S cooler.
I own a NH-D15 on my 5900X. It's advised to avoid overclocking and stick with PBO due to limited single-core performance. After some adjustments, it performs well—adding roughly 2-300 MHz across all cores when under load and supporting 5GHz+ single-core speeds. Refer to the video for a detailed guide; this setup works perfectly.
Is that truly the main reason? Are the discussions around Ryzen 5900X and Zen 3 components really limited by standard base speeds, with only small manual boosts possible—around 15% to 20%? Most of these chips are considered weak and rely heavily on PBO, making it hard for anyone to know what works.
Absolutely, it works that way. Generally, achieving full manual control isn't possible without PBO's optimization, which can be fine-tuned through several hours of testing.
Some concerns about Zen 3's performance stem from its limited overclocking capabilities compared to Zen 2, while still being significantly improved over earlier generations—though it falls short of Intel's benchmarks without specialized tools like PBO.
It runs at 4500/4600MHz across all cores with PBO and some CO + EDC/TDC/EDC adjustments, while the 4900-4950Mhz version uses one core when possible. This pushes performance well beyond the base 3700MHz range. Locking every core to 4.6GHz actually limits better low-core-count efficiency.
You wonder where the FUD originates from with AMD Ryzen CPUs that struggle to overclock easily by hand. It might be linked to silicon quality, causing inconsistency especially when doing manual tweaks. PBO seems more reliable for pushing higher even with subpar chips, while manual adjustments often fall short compared to what you achieved. Many claim they successfully boosted clocks to 4.6–4.7GHz manually, yet doubt their performance would meet LinPack XTREME 1.1.5 standards—something your previous 5900X couldn’t achieve at higher voltage or lower clock speed.
Dark Hero supports DOC, and it doesn’t impact single-core performance.