Overclocking Basics?
Overclocking Basics?
Clock speed enhancement is a single-thread boost. My 3700X's base boost frequency is 4.4 GHz (reaches it only under stock or PBO settings), while my overclock setting runs at 4.2 GHz across all cores, which exceeds the chip's sustained boost of 4 GHz during CB20. It seldom surpasses 4.2 GHz overall.
I understand... However, there are always exceptions. Many factors come into play. One point I covered was in the quote you mentioned only partially.
Ryzen delivers strong performance at a surprisingly low voltage, but when cooling options are tight it slows down significantly. 1usmus shared an excellent OC solution for Ryzen, https://www.guru3d.com/articles-pages/cl...mus,1.html. Boosting stock single-core boost speeds through all-core tuning depends on luck with silicon and effective cooling.
Absolutely, I understand. It's important to highlight that cooling is a major concern for these characters, and I should have included the full sentence as requested.
I discovered with Zen+ chips that the processors perform excellently even at low temperatures. The CPUs delivered the expected performance improvements as temperatures dropped. I found that using TEC cooling allowed me to run at around -30°C while maintaining full core boosts of 4300MHz. However, the cooling system limited the maximum load temperature to about 25°C based on water flow adjustments. For instance, a 4GHz all-core/threads setup with TEC cooling needed only 1.188V. The testing was confined to benchmarks, highlighting significant power savings and reduced leakage. Still, the chip's architecture restricts higher clock speeds—things like the piping affect this. Longer delays in instruction per cycle mean lower IPC, while shorter delays boost it. Compared to FX chips, ZEN offers better efficiency, partly due to size and density differences. Generally, high voltage isn't essential for achieving higher clocks; most users opt for sub-ambient cooling instead.
Well, ctr showed me a bit lower than what I thought was stable after manual setup. It seems the chip has thermal boundaries, so it's not pushing silicon limits. Ctr 1.1 isn't very helpful right now—notes say stress tests are affected, and even with low voltage and frequency temps are too high to run them. There are only small improvements under 55°C, which is interesting. I'm eager to see how voltage scaling behaves.