PBO variations with temperature effects
PBO variations with temperature effects
Hi everyone, I've been working with overclocking for a while now. My 5900X was performing well on a crosshair v8 system until I dove into BIOS and adjusted the PBO scalar. At first, it was consistently around 4.9GHz with about 80 watts, though not all cores were active—most were sleeping. The system ran smoothly and temperatures stayed at or below 80°C during heavy use. I enabled everything in BIOS, set the temperature cap to 90°C, and adjusted the auto scalar settings to positive with a value of 1. After rebooting into Windows, I noticed the CPU would spike between 110W and 209W depending on load. It fluctuated a lot, matching the CPU temperatures. When I checked Task Manager, all 12 cores stayed at the same speed, which was great. But when temperatures reached 90°C or higher, I worried the PC might shut down immediately. After a restart, I tried a negative scalar (minus 1) and again set it to 1. The system started working normally, but I still saw the wattage swing so much. I thought a negative scalar would lower power usage, but it didn’t help much. I ran a stress test using nicehash and saw temperatures hit 110°C, which is definitely not ideal. I let it run for about six hours as a test, hoping the cooling setup would handle it. With my large radiator area and custom components, I figured it could survive. Why did it not blow up despite the 85°C max setting? Normally, if voltage spiked to 60–94°C and then dropped, the PC would shut down. Why was so much voltage being pushed even without load? And why did it seem stable overall? Any tips on fixing this weird power draw? Is tweaking the PBO still practical at 4.9GHz? Thanks a lot!
wow, ive spent months asking for data like this. i have a 5900x as well, very helpful! How could you tell which cores are the fastest? Suppose repeated stress testing 12x over or more could answer that one, just wondering if theirs an easier way THANK YOU!!!!
These are performance testing tools used to measure CPU speed. Cinebench evaluates overall performance, Aida64 focuses on memory bandwidth, and Prime95 tests sustained CPU load. They’re often confused but serve different purposes.
I started exploring further and found some interesting stuff. Between Cinebench R23, AIDA Prime and Nicehash, I’m running smoothly at 76°C, 53W (average 1.186v per core). My hash rate nearly doubled after adjusting the ccx0/1 ratios. Full load, 4.3GHz, no crashes at all! That explains why hardware info is so helpful. Oh, and a quick note—pen and paper are fun too, turns out ccx1 is preferred for most cores now. -why waste time with PBO? -what’s the core multiplier like when comparing ccx0 and 1? It looks like setting ccx0 to 44 and 1 to 46 might cause a crash, but if I get the right power and voltage, I could hit 4.9-5. The ccx0/1 pairs need to match perfectly—it’s all coming together! Whoop!!!