Overclocking Ryzen 3000 is pointless and speed is merely a figure
Overclocking Ryzen 3000 is pointless and speed is merely a figure
I own the Ryzen 7 3700x and previously ran it using CTR at 4.45Ghz for system and desktop tasks, 4.4Ghz in 4C, 4.35Ghz in 6C, and 4.2Ghz across all cores. After adjusting all cores to 4.1Ghz at 1.250V, I didn’t observe any performance decline during regular tasks or games. The cooler operation with the fixed voltage eliminated temperature spikes. I prefer this setup over increasing frequencies for better cooling and stability.
I disagree.
It does lead to significantly higher heat generation, but that's just a superficial improvement in performance. With a modified PBO, my 3700X behaves similarly to a 3800X (stock) during multi-threaded and single-threaded operations when evaluated using the correct benchmarks (CB20 and CB23). Without it, it performs like a fairly average 3700X in those same tests. I can't see any real trickery there.
However, I do understand it requires adequate cooling (I use a 240mm AIO) to sustain multi-thread efficiency for demanding tasks such as long video transcoding. On the stock Wraith Prism, I'm pretty sure it...
Even PBO serves as a distraction, offering increased heat and power without improving the real performance.
I don't agree with that assessment. While it does lead to a significant rise in heat production, it seems more of a superficial improvement rather than real performance gains. With a modified PBO, my 3700X behaves similarly to a standard 3800X across both multi-threaded and single-threaded tests (as per the benchmarks). Without this enhancement, it performs just as well as a typical 3700X in those tests. I find it hard to believe this is just a trick for show.
However, I do recognize the necessity for strong cooling—my setup uses a 240mm AIO. This would help maintain performance during intensive tasks like video transcoding. On the stock Wraith Prism, I’m confident it would slow down noticeably, possibly to a regular 3700X with a loud fan.
For lighter synthetic workloads such as Prime95 small FFTs, it will still cap around 4125, which is close to what you expect. Still, it can handle moderate to light loads better than expected, though not as effectively as under optimal conditions. For demanding tasks like gaming, the performance drops noticeably.
Yes, but it remains just a figure unless the benchmark differences exceed a few hundred points, then you'll see a real improvement in daily activities, gaming, and even under heavy workloads like rendering. I'm not sure if you've experimented with it yet, but CTR seems to be the top choice right now for keeping that 4400 stable at manageable power levels, far superior to PBO. As I mentioned earlier, it's just a number and you won't notice any gains from 4200 or 4100. 1usmus has previously said that with Ryzen's design, there won't be any performance increase beyond 4000.
The issue with contemporary overclocking is complex:
Many believe overclocking means pushing the turbo limit higher. But most modern processors already reach near their maximum, so adding just 0.1 to 0.2GHz is rarely noticeable without special techniques. This becomes even more challenging when enabling features like full-core turbo boosting, which limits further gains.
Despite this, the idea persists that overclocking always delivers significant performance gains—essentially free speed boosts. In reality, processors are often soldered close to their limits, and beyond around 4.0GHz, only minor improvements appear. While we hope for a direct performance-to-effort ratio, physics ensures a more balanced rise in power usage.
I notice some noticeable enhancements in the rendering process, though the slight improvements during a 30-minute video transcode don't stand out much. In this regard, AMD has definitely made significant strides with their boosting technology and chiplet sorting methods. The architecture behind chiplets allows these components to serve as versatile building blocks across various CPU designs—ranging from entry-level desktops with minimal cores all the way up to high-end servers and massive workstation processors. This flexibility means that even the low-end 3700X models are being utilized effectively, leaving limited room for overclocking.
My aim in pursuing overclocking has always been to upgrade to a higher tier in hopes of achieving better performance without pushing the system beyond its limits. The results I'm observing with my 3700X PBO suggest I've achieved my objectives, at least in terms of gains. However, the modest improvements observed don't fully reassure me about the value of the 3800X's current price. This has sparked further debate among enthusiasts.
Additionally, many users have reported substantial manual overclocking success using more established silicon processes. Yet, even these efforts seem to be trending back toward lower voltages over time. It’s possible that some users have pushed their CPUs beyond safe operating limits, only to notice instability as the components age.
The latest chips significantly improve performance across all core clocks. Overclocking remains inconsistent in the silicone selection, but occasional successes occur. My 3600 achieves a 4.4ghz boost with 1.28v and a droop as low as 1.269v. PBO over-volts the chip excessively, causing overheating. It operates more efficiently at 4.4 compared to using PBO, which reduced all core clocks to 3.9. Results will differ among users. The era of frequent 30%-50% overclocks on air or water is now behind us.
With these Ryzen CPUs, overclocking is definitely possible but comes with some caveats.
If my 3600x runs well on an Asus Rog Strix F Gaming, and my son's 3600 on an MSI B450 Pro Carbon also performs similarly, I’m unlikely to achieve a stable OC on the 3600x.
His 3600 maintains steady performance at around 4.225g with boost enabled, while mine reaches about 4.275g during Minecraft and Detroit Human. The rest of the time it stays at 4.225g when limits are increased, PBO applied, and a small voltage reduction is used.
I’m confident it’s worth it because my 3600x hits 4.325g OC at 1.30v, giving it a very close McCbride 20 score (within 15 points) for a 2700x.
I still had to consider whether the extra risk of higher voltages for minimal performance gains is worth it.