The R7 5800X3D is designed for optimal cooling performance.
The R7 5800X3D is designed for optimal cooling performance.
Check the PBO area, then adjust the Curve Optimizer to -30 fully and benefit from improved performance at reduced temperatures. If -30 causes issues, try -25 instead.
I kept it for a few weeks. I paid close attention to matching the screw tension and ensuring the cooler was centered. It seems I might have placed the case on its side while setting it up. The temperatures on synthetic benchmarks are either similar or lower than what others report, except for the OCCT small data set. On Medium data it doesn’t reach 72°C, Cinebench R23 works well at 75-76°C, and even the stress test in Minecraft hit the low 80s—while the whole CPU was loading. It looks like the OCCT small data set is the issue, but I’m not sure why. Am I facing a problem with just one benchmark or are all these readings unusually high? OCCT is using AVX2, right?
Your temperatures aren't unusually high, just slightly above what you'd expect. This might point to a mounting pressure problem. You could try adjusting the voltage using the BIOS and lowering it slightly. There are great tutorials on YouTube that walk through this process—just watch and take notes. It's possible you're experiencing uneven power draw from certain components. More power generates more heat, but it varies depending on the apps running and many games don't stress the CPU as much as software tests. I don’t see a cause for concern, but there are several methods to help lower temperatures and ease your worries.
Check typical benchmarks for your setup. Cinebench R23 generally reflects CPU performance, while GPU scores depend on the card and drivers. Expect ranges based on your hardware configuration.
The app may experience noticeable temperature changes in identical CPUs because of silicon properties. De Bauer conducted some solid tests, indicating acceptable performance as long as results stay within 6%–8% of similar models. A maximum deviation of 7–8% is considered normal. Expect swings between 3% and 6% to be typical. Every CPU behaves differently, so setup matters. I tested an older 8700k with a dual fan Noctua Owl; the rear case fan had a bigger effect than expected. The test was challenging due to excessive noise, leading me to replace it for better results. Removing it didn’t save money, which wasn’t ideal.
I've been adjusting the wiring over the last hour to prevent the rear case fan from being tied down by the 140mm unit in the NH-D15-S. I replaced and repositioned fans, even adding a spare 120mm Noctua Redux model on the front of the heatsink (only the 140mm was used before). In short, nothing altered much. Swapping the front fan didn't noticeably change temperatures, and the extra front fan likely added just a small margin—probably due to measurement inaccuracies. Removing it didn’t show a clear temperature rise, so it seems the situation remains stable. The overheating in OCCT SFFT and Prime95 SFFT might stem from their heavy caching, especially on the 5800X3D. I'm not a hardware specialist.
I disabled all AVX settings in OCCT and it managed smaller data sets effectively this time... it stayed near 85-86°C but ran faster overall, using similar power to before but with better cooling.
It's well known that 5800x3D gets quite warm. You might need improved cooling or adjust PBO2 to lower temps as mentioned before, or just accept it—temperatures stay within normal ranges. In my setup, I hit around 79°C during the CPU with PBO2 at -30 (and with a cooler running less than max). Hope this helps!
That aligns well. I've noticed similar trends with hot conditions and AVX performance evaluations using limited datasets. What's your take?