I upgraded from a Ryzen 7 5700X to a Ryzen 7 5800X3D, but my performance got noticeably worse. In Cinebench, the 5700X scored about 15,172 points while the 5800X3D only scored around 14,002. The 5800X3D reaches roughly 90°C in Cinebench and around 85°C while gaming with my DeepCool LE500 cooler and Snowdog Husky thermal paste.
Gaming performance also dropped significantly. In the same area of Resident Evil Requiem and with identical settings, I went from roughly 60 FPS on the 5700X to 35–40 FPS on the 5800X3D. I tried a -20 Curve Optimizer undervolt, but -30 caused system shutdowns. My 5700X previously reached about 80°C with PBO enabled and a -15 curve setting.
The system uses an ASUS TUF Gaming B550M-PLUS motherboard, an RTX 5070 Ti, and 32GB of Kingston Fury Beast DDR4-3200 CL16 running at 3600MHz with an 1800MHz FCLK. BIOS was updated recently. What should I check before returning the processor?
4 Answers
The cooler installation is still worth checking, even if it was reused from the 5700X. Reseat it with fresh paste and confirm the pump and radiator fans are working properly. Around 90°C in a sustained all-core workload can be within the 5800X3D’s limits, but if it stays near that temperature in games, it may reduce clocks. Since the cooler has already been reinstalled successfully and gaming temperatures were later reported around 50–60°C, cooling alone may not explain the FPS loss.
The 5800X3D normally runs hotter than a 5700X, but a large gaming performance drop is not expected. Check that the CPU is reaching its normal boost clocks and that it is not thermal-throttling. A Cinebench score by itself does not explain gaming performance, so monitoring clock speeds, CPU usage, GPU usage, and temperatures while the game is running would be more useful.
First verify the software and firmware setup: use a BIOS version with proper 5800X3D support, install the current AMD chipset driver, and confirm that the processor is being recognized correctly. Also check that XMP/DOCP, the 3600MHz memory setting, and the 1800MHz FCLK are stable. Incorrect CPU or memory settings can cause unexpectedly poor performance.
Try testing with PBO and Curve Optimizer disabled before applying an undervolt. A -20 setting may be stable enough to boot but still cause errors or reduced performance, and every chip behaves differently. Re-enable tuning only after establishing a stock baseline. If the board supports it, make sure Smart Access Memory or Resizable BAR is enabled as well.
I tested with PBO and undervolting disabled and the result was about the same. I’m using an RTX 5070 Ti, so I’ll also verify the Resizable BAR settings.
The FPS difference is happening in the same game and area, so I’m going to monitor clocks and GPU usage during a test rather than relying only on Cinebench.