5800x (and other Zen 3 chips) PBO settings/Temperature fix

[Cowboystrekk]

my 2nd pic lists my co offsets but I will put them in my initial post and here.

-8 -6 -20 -22 -20 -13 -10 -15

Also, the ram I have took a long time to get to the timings and frequency. its rated for 3200mhz. It was very difficult to get it to the timings and speed it is at now. But if you really think that it will increase my gaming performance noticeably, I will work on it. Aida64 reports my memory latency as 61ns. With XMP, that value was 67ns.



Thank you, I should have specified that I already have my CO offsets in the 2nd pic I posted. It shows ROPbench and has my offsets along with my PBO values. ROPbench Imo does the same thing as pbotuner2 plus more.


I would like to know something. I ran OCCT CPU + Ram test. It maxes out my limits further than just a CPU test. I wanted to know the effect of setting PBO limits in real time, so I played with them and watched my clock speed. If I was at 100% Utilization of my PBO limits, my clock speed would drop on TDC and PPT. With EDC even at 100% utilization, the "Sweet spot" seemed to be 85A. Any lower or higher and my effective clock speed would drop. I did the same thing with TDC and PPT. With PPT it was 160W. 10W below max usage during OCCT. Any lower or higher and my effective clock speed would go down. TDC was a little harder to gauge, but I noticed clocks being affected when TDC was below or higher than 100A. Im curious if this test I did is only specific to stress testing with OCCT, or If I set the PBO values I found, will they be correct for gaming as well?

It is strange that you get lower scores using these values. Is edc/tdc/ppt the same with and w/o co? Can you post a screenshot with hwmonitor running when running cinebench? I want to see corevoltage, frequency, temps and power.

Save your current ramprofile and try my suggested settings if you like. This ram looks like Micron rev E DR or rev B SR from the timings. They should both be quite easy to run at 3800 You may have to play a bit with resistances.

 

[forkinator]

It is strange that you get lower scores using these values. Is edc/tdc/ppt the same with and w/o co? Can you post a screenshot with hwmonitor running when running cinebench? I want to see corevoltage, frequency, temps and power.

Save your current ramprofile and try my suggested settings if you like. This ram looks like Micron rev E DR or rev B SR from the timings. They should both be quite easy to run at 3800 You may have to play a bit with resistances.

Regarding your first question, I'm not entirely certain if both runs were conducted with the same PPT, TDC, and EDC settings. . To clarify, I executed CB with my existing offsets at 160PPT/100TDC/85EDC. For the comparison benchmark where I disabled all settings, I simply enabled PBO. On my motherboard, enabling PBO hides all power and frequency options, leading me to assume that 'enabled' would default to auto settings for PPT, TDC, and EDC. I was hoping to confirm whether this is indeed the case.

Not sure if they are visible yet (moderation needs to be completed), but I have two comparison screenshots of my Cinebench PBO + Offsets VS PBO only scores in my first post. HWinfo was started right when Cinebench multi started, and a screenshot was taken right after Cinebench completed.

If you are specifically asking for HWmonitor, I will download it. I'll implement the settings you recommended and share the outcomes. A little info about my ram situation; This kit is new as my previous kit was identical, yet it failed to post at XMP settings. After owning the kit since 2019, I contacted Corsair a month ago, and they replaced it. That kit was version 4.32, which I suspect was Samsung D-die. This current kit is ver 3.31.

 

[Zach_01]

On my motherboard, enabling PBO hides all power and frequency options, leading me to assume that 'enabled' would default to auto settings for PPT, TDC, and EDC. I was hoping to confirm whether this is indeed the case.

PBO disabled or Auto = Disabled. CPU runs at default AMD factory settings

PBO enabled = The board sets the limits for PPT/TDC/EDC and with the values applied it’s practically unlimited. The CPU internal silicon management takes over to keep CPU healthy by keeping temp/EDC in check. Not the best way to use it as it’s non-tuned, just unleashed and inefficient in most cases if not all.

PBO advanced (selecting manual settings) = this is the go-to settings for those how like to tune the CPU performance, limits, V/F curve and/or additional boost.

 

[Octopuss]

I don't think I ever tried running Prime95 on all cores on my 5800X, but I just have since I am tweaking FAN speeds, and to my surprise, the damn thing runs at rougly 85-90°C on all cores. Isn't that a bit too high? Or is it normal?
I have some kind of beefy Noctua cooler, use liquid metal TIM on the CPU, and run various negative PBO offsets (it varies wildly core to core). No overclock, power limits set to motherboard in the BIOS. All cores boost is 4500MHz, which I assume is the default.

The bizarre thing is the cooler is pretty much cold, but I am sure it's seated properly. I mean the CPU is idling at 40°C or so (not sure exactlycca 32°C).

 

[Assimilator]

I don't think I ever tried running Prime95 on all cores on my 5800X, but I just have since I am tweaking FAN speeds, and to my surprise, the damn thing runs at rougly 85-90°C on all cores. Isn't that a bit too high? Or is it normal?
I have some kind of beefy Noctua cooler, use liquid metal TIM on the CPU, and run various negative PBO offsets (it varies wildly core to core). No overclock, power limits set to motherboard in the BIOS. All cores boost is 4500MHz, which I assume is the default.

The bizarre thing is the cooler is pretty much cold, but I am sure it's seated properly. I mean the CPU is idling at 40°C or so (not sure exactlycca 32°C).

Why do you care about what happens when you run a synthetic benchmark?

 

[Waldorf]

@Octopuss
i would still compare numbers used by the board, to what you should be using/not go over (for your chip).
there are some numbers that will have negative impact on perf, while they might not even be max board can do/handle.

 

[Octopuss]

Why do you care about what happens when you run a synthetic benchmark?

I... guess you have a point? But I use Prime95 to check stability, and perhaps if the CPU is throttling, potential instability might not show up.