Is the Ryzen 3 3100 4.3GHz 1.25V suitable?
Is the Ryzen 3 3100 4.3GHz 1.25V suitable?
On Prime95 Small FTT it showed no errors for ten minutes, then I stopped the test. Cinebench R23 scored 6406 points. Also, I played 30 minutes of Apex Legends and it was enjoyable. Cooler: Snowman T6 with MX-4.
I have a 5600G, which means some of my settings might not be relevant for 3000 series devices. Here’s what I did for tuning my 5600G:
Started HWinfo64, checked the "Core Effective Clocks" group. The standard core frequency reading is around 4.3GHz, but the actual effective core frequency determines performance.
On the group with normal core clocks, you’ll often see numbers in parentheses (like perf #2/3), indicating the window and AMD performance ratings for that core. Higher numbers mean a poorer core. This will guide your next steps.
1.25V at 4.3GHz performs marginally better than typical standards.
I used a 5600G and followed these steps for tuning:
Started with HWinfo64 running. Checked the "Core Effective Clocks" group. The standard core frequency reading might be 4.3GHz, but the actual effective core frequency is what matters for performance.
On the group above that showing normal core clocks, you'll often see a few numbers in parentheses (like perf #2/3). These indicate the window and AMD performance ratings for that core. Higher numbers mean worse performance. This guides your next actions.
Run Prime95 small FFT.
After Prime95 starts, reset the HWinfo64 readings (the clock dial symbol at the bottom). You should look for any threads dropping below the average effective clock—this signals instability. These unstable cores usually cause a thread crash in Prime95.
With this info, you can decide whether to increase voltage for stability or lower it.
On the 5000 series PBO Curve Optimizer, you can apply negative offsets per core or all cores. If you get values like -20, -30, -30, etc., and any of those -20 cores activate, then all cores will follow that stricter voltage setting. The only case where you get a full -30 is if none of the -20 cores are on.
If Prime95 runs smoothly for at least 30 minutes without crashing threads or the whole system, your setup is quite stable.
1.250 to 1.225 represents a reasonable voltage drop. Since 1.250 is reliable, start lowering to 1.242v, then 1.234v, and so on.
I avoid adjusting Ryzen vcore, as it negatively impacts single-thread performance by requiring higher voltages for full multi-core boost. This can reduce overall frame rates, especially in games that rely on a master thread with other threads supporting it.
This approach is preferable as it allows for better control over voltage settings rather than relying on a fixed voltage cap. This is particularly useful when you don’t want the system to automatically adjust frequencies based on core count. For instance, my 5600G operates at 4GHz across all cores at 1.2V, yet achieves 4.6GHz with a single core at 1.44V. When using Prime95 with full core loading, I’d expect 4GHz and adjust accordingly. However, if a constant 1.2V limit were applied, the higher frequency wouldn’t be possible because the CPU can’t sustain that shift in voltage. Still, during manual tuning, this detail remains important to consider.
I tested ClockTuner2 and Dram Calculator. The CTR2 includes two profiles plus a default, giving me standard idle configurations up to 18% loads, 4.4GHz on 1-4 cores at 55% loads, and 4.29GHz across all cores on a 3700x. A single-thread performance increased by around 30 points, while multi-thread improved nearly 400 points in Cinebench R20. Temperatures fell by over 20°C for the profiles because the CPU VID is limited to 1.325v. That’s essentially what the offset does—reduces the VID, which reduces CPU demand without changing CPU usage.
I could ask too. Using voltage offset might give similar results to Intel Turbo Boost. For example, if I'm only running at 1.35V with all cores fully overclocked, it would work like a boost setting. But when idle or under lighter loads, the voltage would drop below that level. I'm new to Ryzen overclocking and trying to understand if AMD can achieve something similar to Intel Turbo Boost and speedstep. My goal is to see if lower clock speeds and voltages are possible without heavy processing. Right now, it's been frustrating to get 1.35V on all cores whether it's idle or moderate use. I'm more interested in dynamic voltage overclocking. Thanks. 😀