Ryzen 5950X PBO plus CO combo
Ryzen 5950X PBO plus CO combo
Over the past few days I've been working to maximize my 5950x performance. After stocking it for about 1.5 years, it was time to test its limits. Stock CB23: 24,500 I began by searching for the PBO sweet spot. PPT: 250 (it never hits this mark) TDC: 150 (my motherboard max is 160 but it only added more heat without extra gain) EDC: 150 (to align TDC, which gave the best CB23 results) Max Boost override was set to 0 since I favor a natural curve over boost (Bosmang Precision Boost Overdrive 2 Guide). This led to a consistent core stress temperature of 75°C, gradually rising to 80°C as the coolant temperature in my H150i increased. This boosted my CB23 score to 28,800. The next step was using the Curve Optimizer to squeeze out more performance. I adjusted the CoreCycler script to automatically raise CO values per core during instability. After three days of testing (FFT720, HeavyShort, Heavy, ...) this improved my score by several points. The surprising part is Core0. While all other cores are within normal ranges, without a boost override it still needs a +2 for Prime95 stability. I suspect the issue might be another unstable core forcing the script to target the wrong one. Changing -5 0 0...0 caused crashes until I set it to +2. Am I just unlucky, or is Core0 problematic? Or am I overlooking something?
I discovered that running with HWinfo 64 and keeping the EDC near 100% during Cinebench tests gave the best results. TDC is the thermal Limit PPT, which is essentially power limit EDC on my 5600x. Getting it close to 99% was the goal. Example numbers: PPT 110, EDC 124, TDC 90 – I’m not sure about 5950x but I wanted to note it. Spoiler alert!
For me EDC always matches what I set unless TDC is below 118A or PPT is under 160W. Even with TDC at 115A, EDC will nearly reach 190A. If I raise EDC to 190, I lose about 600 points in CB. I know what you mean, but it doesn’t seem to impact my CPU the same way.
This matches the format of HWInfo observed during a CB run, around 250-150-150.
Here are a few references I came across. A few results are slightly better, others are a bit closer but still helpful. Hope this gives you more direction!
I’ll consider adding the boost override though—I’m a bit unsure since my curve is already +2 on core0. I’ve tried 250/160/190 but the scores came in lower, though they’re still within acceptable range. Let’s try pushing my curve further and see how the boost override affects things. Thanks for all the feedback!
I believed Curve represented a voltage-frequency relationship. Therefore, during boost overwrites the curve would shift since higher frequencies are involved. I thought your curve wouldn't be affected by any PBO restrictions. I ran tests using default configurations, PBO parameters, and cor0 also showed instability with negative offsets. Are you suggesting the curve applies to other PBO settings? A lot of trial and error was needed to explore this further.