Optimal BIOS configurations for 7800X3D
Optimal BIOS configurations for 7800X3D
This is my first AMD processor since the Athlon XP days, so I am a little out of the loop with current BIOS settings. I know it's not OCable as Intel is, but i know something can be done with PBO and CO. I have paired the 7800X3D with an Gigabyte X670 AORUS Elite AX and 2x16GB G.Skill 6000Mhz CL30, cooled by a Lian-Li 360 AIO (GA II TRINITY PERFORMANCE) So far I have enabled EXPO, PBO and CO -30. I also saw in the BIOS PBO enhancement for which I can choose 70/80/90 Level 1->5, but I have no idea what this means. I assume the former are temps, but what do those levels mean? With the current settings I saw [email protected]. The temps stay around 60-65ºC in games, but sometimes i see jumps to 70-90ºC for 5-10s, which seems very weird, especially since there is no heavy on screen activity to justify it.
If you're lucky that won't give you any issues. If you start having idle lockups and/or games crashing for no obvious reason, back it down to -25 or -20. Frankly, CO has made absolutely zero real world difference for me on my 7800x3D, though it was fantastic on my 7900x. Otherwise there is no need to touch anything else. (there isn't really a need to touch anything at all) You're not going to gain any appreciable performance regardless of leaving it stock or touching CO with this chip, it just kind of kicks ass out of the box. The AIO was massive overkill but some buy them for looks, I get it.
The 70/80/90 level 1-5 configurations are already set up. They adjust the thermal throttle and voltage offsets. The 90 level 5 setting exceeds -40mV. It uses a 3D v-cache chip, causing it to overheat significantly. That’s why I removed the head from my 7950x3D. For simple testing, lower the SoC voltage. I operate at 1.1V. You might also adjust CPU/VRM voltages with Vcore and SoC load line reductions. Running on Turbo is feasible only with direct-die or heavy cooling. The 50.5x multiplier limit applies to the 7800x3D. The 7900x3D and 7950x3D CCD0 can reach 52.5x, but this only happens occasionally unless you use direct-die. I performed many R23 runs with data, including post-delay+direct die with a kyrosheet and LMOS. The voltage behavior of the 7950x3D without CCDs isn’t reliable, so consider full 16C/32T benchmarkers for accurate HWinfo64 data.
I apply those same configurations myself, adjusting CO multiple times to different negative settings without affecting my Cinebench results. As mentioned before, it’s built to perform exceptionally from the start. If stability is key at -30, go for it.
The game behaves oddly with fluctuating temperatures. When I launch it, the temperature stays around 80°C to 90°C for a few seconds before gradually dropping to about 60°C. In-game, it doesn’t matter where you are—whether in the middle of a desert or a busy city during a match. During peak times, spikes reach 80-90°C, but these changes happen slowly, like 1°C every few seconds going up and down. The actual activity inside the game doesn’t affect this. On my previous CPU (10700KF @5Ghz), I achieved under 77°C with 180W power draw. Despite similar die size, it uses less than 77°C while consuming around 70-80W.
Consider the metrics you should include in HWInfo. Compared to the 10700KF, there are more sensors which can be confusing. For temperature, CPU(Tctl/Tdie) resembles a GPU hotspot while the average CPU die temp reflects normal operation. The difference between CPU CCD1 and CPU Die might be due to the hardware configuration—having only one CCD could explain the variance. Regarding voltage readings, MSI Afterburner in-game shows inconsistent values; HWInfo gives different ones. In MSI AB you see 1.4-1.5V, but HWInfo lists 1.05, 1.25, 1.11... It seems MSI AB might not be fully reliable. For the CPU clock, in DD2 it was locked at 5050 MHz, whereas HWInfo shows an average of 4.6 GHz—this discrepancy should be noted.