Associate
Hi guys. I have read a lot about PBO and undervolting using curve optimiser for my 5800X but I have hit a wall and wanted some advice.
Basically I mainly use my PC for gaming and wanted 2 things:
- to eke out a few more FPS in CPU limited games like Microsoft Flight Simulator 2020 (especially the minimum framerate and 1% lows).
- to reduce my temperatures since I have a relatively small form factor case.
I tested different power limits with PPT, TDC and EDC and found that the following settings gave me the best thermals with my case and cooling: PPT 110W, TDC 75A, EDC 100A. The efficiency of the 5800X drops off a cliff after 105W and by tuning these values all 3 are hitting 100% together in HWiNFO64 which apparently suggests that they are balanced. I dropped my temps in Cinebench and gaming benchmarks by about 5-15c depending upon the workload whilst keeping my scores within 5% of stock.
I then tried curve optimiser and went progressively from -10 to -30 all cores which was stable in the OCCT single thread testing using the following recommended settings for stability: Data Set - Large; Mode - Extreme; Load Type - Variable; Start at cycle - 1; Threads - Advanced; Physical Cores - core #0; Virtual cores - Physical only; Core Cycle - every 5 sec; swap active/inactive cores - disabled. I got no errors in the log down the bottom left (I assume this is where they should appear?).
I then bumped PBO boost override up to 50mHz then 100mHZ with the same -30 and still got no errors in OCCT V10.1.5 but I started getting frequent BSOD crashes in Destiny 2. I dropped curve optimiser down progressively to -10 but still had crashes so I disabled the PBO boost override and now my system is stable after several 4 hour gaming sessions.
So to where I need help: I was led to believe that this setup on OCCT should have identified which cores would crash and allow me to set the curve optimiser individually so my best cores have a lower negative offset (-5 for example) with a larger offset (-15 to -20 maybe) on the weaker cores which won't be boosting so high. But since it never found errors (or I am not looking in the right place for them) I just kept pushing further. So should I try Core Cycler instead or some good old Prime95 testing?
I realise it is perhaps unrealistic to both reduce voltage on the curve optimiser (technically increase frequency at a given voltage) and set power limits and also expect to increase my boost clocks with an offset. But it seemed like lots of people on different forums were managing this since the 5800X is more efficient at lower voltages and power and by limiting the voltage sent to weaker cores with individual curve optimisation you could save more energy for the best cores. This together would allow higher boosting since there is less heat. Is it even worth trying to set a boost override on the 5800X for gaming or just try to go further again with the core offsets to improve efficiency further and allow higher sustained boosting? I gain score in benchmarks with the extra 100mHz but not many fps.
Sorry for the length of my post. I appreciate any help you can offer. My RAM is on the default XMP settings since tuning it last year just led to system crashes and at best 1% performance boost over XMP. All voltages are set to auto on the latest BIOS for the X570 Tomahawk. My case fans are all Noctuas with a fairly agressive curve to keep CPU temps around 70-80c and my cooler is the Noctua NH-U9S with 2 fans. Thank you in advance
Basically I mainly use my PC for gaming and wanted 2 things:
- to eke out a few more FPS in CPU limited games like Microsoft Flight Simulator 2020 (especially the minimum framerate and 1% lows).
- to reduce my temperatures since I have a relatively small form factor case.
I tested different power limits with PPT, TDC and EDC and found that the following settings gave me the best thermals with my case and cooling: PPT 110W, TDC 75A, EDC 100A. The efficiency of the 5800X drops off a cliff after 105W and by tuning these values all 3 are hitting 100% together in HWiNFO64 which apparently suggests that they are balanced. I dropped my temps in Cinebench and gaming benchmarks by about 5-15c depending upon the workload whilst keeping my scores within 5% of stock.
I then tried curve optimiser and went progressively from -10 to -30 all cores which was stable in the OCCT single thread testing using the following recommended settings for stability: Data Set - Large; Mode - Extreme; Load Type - Variable; Start at cycle - 1; Threads - Advanced; Physical Cores - core #0; Virtual cores - Physical only; Core Cycle - every 5 sec; swap active/inactive cores - disabled. I got no errors in the log down the bottom left (I assume this is where they should appear?).
I then bumped PBO boost override up to 50mHz then 100mHZ with the same -30 and still got no errors in OCCT V10.1.5 but I started getting frequent BSOD crashes in Destiny 2. I dropped curve optimiser down progressively to -10 but still had crashes so I disabled the PBO boost override and now my system is stable after several 4 hour gaming sessions.
So to where I need help: I was led to believe that this setup on OCCT should have identified which cores would crash and allow me to set the curve optimiser individually so my best cores have a lower negative offset (-5 for example) with a larger offset (-15 to -20 maybe) on the weaker cores which won't be boosting so high. But since it never found errors (or I am not looking in the right place for them) I just kept pushing further. So should I try Core Cycler instead or some good old Prime95 testing?
I realise it is perhaps unrealistic to both reduce voltage on the curve optimiser (technically increase frequency at a given voltage) and set power limits and also expect to increase my boost clocks with an offset. But it seemed like lots of people on different forums were managing this since the 5800X is more efficient at lower voltages and power and by limiting the voltage sent to weaker cores with individual curve optimisation you could save more energy for the best cores. This together would allow higher boosting since there is less heat. Is it even worth trying to set a boost override on the 5800X for gaming or just try to go further again with the core offsets to improve efficiency further and allow higher sustained boosting? I gain score in benchmarks with the extra 100mHz but not many fps.
Sorry for the length of my post. I appreciate any help you can offer. My RAM is on the default XMP settings since tuning it last year just led to system crashes and at best 1% performance boost over XMP. All voltages are set to auto on the latest BIOS for the X570 Tomahawk. My case fans are all Noctuas with a fairly agressive curve to keep CPU temps around 70-80c and my cooler is the Noctua NH-U9S with 2 fans. Thank you in advance
Last edited:
