After overclocking, the CPU tends to run cooler due to improved efficiency and better thermal management.
After overclocking, the CPU tends to run cooler due to improved efficiency and better thermal management.
I recently changed from an i7-4790 to a Ryzen 5 3600, and since it's an unlocked CPU, I decided to attempt a light overclock. With no prior knowledge or experience, I began by increasing the base clock to align with the CPU’s boost clock and fixed it there—shifting from 3.6 GHz base to 4.2 GHz. All other settings remained unchanged except for the base clock adjustment on each core and turning off clock boosting. After the change, performance improved noticeably, with a Cinebench score about 400 points higher. During stress testing with AIDA64 Extreme, I observed a significant drop in CPU temperature (16°C lower) and a slight reduction in core voltage (0.3V lower). Interestingly, the voltage fluctuations before overclocking were much more erratic, ranging between 0.948 and 1.320 volts, while after overclocking it stabilized around 1.044 to 1.092 volts. What could explain this behavior? Edit: If you’d like my PC specifications, check my profile page. Edited August 21, 2020 by GrannyDane Specs matter, right?
I'm experiencing some problems with the evidence for this CPU. It seems to load AVX instructions at 4.2GHz, but there are no clear frequency details—just a stated 4.2GHz. The main concern is the screen shots; I can't see the processor's multiplier. Could you share OCCT screenshots with AVX2 enabled across all threads? Also, using HWInfo64 would help monitor everything better. Apart from that, the top golden sample looks good.
Sorry for the late update, I played MHW all day yesterday to check if the system remained stable with the OC. (I understand it's not a very CPU-heavy game, but I wanted to have some fun.) I didn’t notice any issues in the game. I followed the test you suggested, although since I’m not familiar with either of the programs, I’m unsure if the screenshots capture enough detail for you. Let me know if you need anything else. First, I ran it with the OC enabled and HWInfo monitoring: as you can see, it produced many errors when using the OC. I did some quick research and discovered a compatibility problem between HWInfo and OCCT on 1st generation Ryzen processors. Even though my system isn’t a 1st gen model, I thought it might still be worth trying. I then ran the second test using AIDA64 graph monitoring instead (no stress test): it stayed error-free until around the 4:30 mark, after which it started crashing shortly afterward. Perhaps because it was right after about 10 minutes of stress testing? I rolled back the settings with a reboot and reran the stress test, and this time with HWInfo it showed no errors at all. So there must be something I missed. The next step could be adjusting the voltage with OC, since it was running below 1.1V while non-OC was near 1.3V.
It's operating at 4GHz with a 1.250v setting, which causes throttling during loading instead of hitting 4.2GHz. Each test seems to last only a few seconds. The counters remain at zero, and I haven't received any updates yet. So far, no results at full speed except idle timers. At higher loads, the frequency adjusts to match the lower voltage core.
The initial tests indicating 10:00 were executing infinitely, which I stopped at the 10-minute point manually. The remaining ones showing 0:00 came from adjusting the countdown to a fixed duration, and I captured the screenshot once it finished. Both photos are from the same trial. You’ll notice the second image labels "Timed" as "00:10:00" in the top left. The first image lists a minimum clock speed of 4,192 MHz under "Core 0 Clock." Doesn’t that seem sufficient?
Yes, a slight rise in vcore can help reduce errors. Values around 1.125v might make a difference.
I adjusted the voltage for around two hours and found a stable +0.012V. I also experimented with setting it to 0.000V instead of Auto, but errors appeared quickly. It felt warmer than before yet still cooler than the default settings. The results remain consistent!
Additionally, the saving option is absent in OCCT for an unknown reason—possibly connected to the pop-ups prompting a $1 contribution during the initial stress test.