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

[Pivo187]

Water temperature is less spiky than the actual hardware temperature. Therefore it is better to use the water temp sensor for pump/fan control. The stock fan profile in the H100i is to aggressive for most, so it will cause annoying fan speed changes. Making a more boomerang shaped curve is usually a good thing. Then find a temperature above the normal spikes to set the steepest increase in fan speed.

If you want to keep your fans connected to the H100i you need to use the older Corsair software to set the fan curves. This profile is saved to the H100i directly, so there is no need to run the software constantly.

If you want to use the motherboard to control the fans, you have to unplug them from the H100i first, then connect them to the motherboard directly. Use CPU and CPU_OPT headers if you have both. Alternatively CPU header with a splitter. Then your fans will follow the fan curve set in the bios.

The pump rpm output from the H100i should be connected to the CPU header on the motherboard if you keep your fans connected to the H100i, so that the motherboard can warn you if the pump is not running. If you move the fans to the motherboard instead, you should move this back to the AIO pump header.

I'm not using the stock profile. I've created my own. What is the correct way to have all of this hooked up lol

 

[WhenMusicAttacks]

I think all you need is to use AVX2 Extreme, that did the job for me.

I think corecycler is still much harder for each core thus better at detection right? Has anyone experienced a system where the former was stable but OCCT was reporting errors?

 

[izy]

I think corecycler is still much harder for each core thus better at detection right? Has anyone experienced a system where the former was stable but OCCT was reporting errors?

I am using corecycler first then im going to OCCT when OC-ing, i feel that corecycler is better for Curve Optimizer testing and OCCT for FLCK OC testing.

 

[WhenMusicAttacks]

I am using corecycler first then im going to OCCT when OC-ing, i feel that corecycler is better for Curve Optimizer testing and OCCT for FLCK OC testing.

Ok thanks, makes sense as OCCT puts load on all cores thus heating the chip a lot more

 

[Pivo187]

So using amd own ryzen master is not enough to test the cores for the curve op?

 

[izy]

So using amd own ryzen master is not enough to test the cores for the curve op?

From my own experience i would say no , at least not when you have AutoOc / PBO enabled, RM can give u somehow stable values but they are not error free, i mean my CPU can do -30 on all cores ( it think RM gives me a -29 -30 on the cores) and works OK without noticing anything wrong, even in Cinebench and stuff but testing with OCCT / CC i get errors so i have to relax the curve on some cores.

 

[Pivo187]

From my own experience i would say no , at least not when you have AutoOc / PBO enabled, RM can give u somehow stable values but they are not error free, i mean my CPU can do -30 on all cores ( it think RM gives me a -29 -30 on the cores) and works OK without noticing anything wrong, even in Cinebench and stuff but testing with OCCT / CC i get errors so i have to relax the curve on some cores.

Doesnt RM also change your PBO values or just Curve OP? I think after running the CO tests it changed my PBO settings from what I had manually input. RM majority cores are at -24 with the 2 others being 16 and 17. It also set my PPT to 120, TDC 75 and EDC 110. Should I change any of this or its good to go?

 

[Pivo187]

Bump anyone alive in here!?? So you can't tell corsair software to use the h100i temps bc those never really change much... They normally stay in the 30*c range.. So my question is what in the corsair software should be probing the temps of the cpu if you look above in the picture I posted it only gives me the option of MB temp#1-7, gpu temp, and again the useless h100i temp that never changes. I've also noticed that corsair software over exaggerate temps as well when compared to Ryzen master and also hwinfo64. When corsair software sees cores at 40, hw is seeing them at 34...So this could also be causing the random speed up of the fans bc it's reading temps at higher levels than the other two softwares. Also I seem to have figured out that the MB 1-7 values are actually the 1-7 cores lol so if you choose a core that normally runs hotter than the other cores the fans will ramp up much more... This corsair software is stupid. You should be able to just choose an average of the 8 cores temps and not have to choose the temps of one of the cores for the fans to go by.. This is def the problem that is going on with fans ramping up.. Is there any fix for this?!?

 

[HD64G]

So using amd own ryzen master is not enough to test the cores for the curve op?

You just need to lower the offset a tiny bit from the one the RM shows. I put -20 instead of -22 and -23 for -25. In my experience that is proven stable even with +150MHz boost increase. In your case with the figures you posted in the post previously to the last one, I would go for -22 instead of -24 and -15 insted of -16 and -17. Also your choice to lower the PPT to 120 will helps temps and stability even more. I would leave the TDC and EDC to the preset settings though to allow max boost clocks when the power limit isn't reached.

 

[Pivo187]

Corsair software can't use the h100i temps bc those never really change much they stay in the 30*c range.. So my question is what in the corsair software should be probing the temps of the cpu if you look above in the picture I posted it only gives me the option of MB temp#1-7, gpu temp, and again the useless h100i temp that never changes. I've also noticed that corsair software over exaggerate temps as well when compared to Ryzen master and also hwinfo64. When corsair software sees cores at 40, hw is seeing them at 34...So this could also be causing the random speed up of the fans bc it's reading temps at higher levels than the other two softwares. It seem that the MB 1-7 values are actually the 1-7 cores lol so if you choose a core that normally runs hotter than the other cores the fans will ramp up much more... It's dumb corsair chooses to use a temp of one of the 8 cores instead of an average of all 8.... I decided to choose the core that had an average temp of all 8 together. Is there any fix for this?!?

 

[Pivo187]

Bump arooski any one alive in these forums!??!?

 

[AMF]

try 199 ppt 150 tdc 150 edc soc 1.118v cpu voltage to offset auto option with +

 

[Pivo187]

try 199 ppt 150 tdc 150 edc soc 1.118v cpu voltage to offset auto option with +

That sounds way too high...lol

 

[gasolin]

try 199 ppt 150 tdc 150 edc soc 1.118v cpu voltage to offset auto option with +

Way to high, wouldn't make a possitive effect

It's not like i don't get the performance i should

 

[Pivo187]

Way to high, wouldn't make a possitive effect

It's not like i don't get the promance i should

View attachment 290396

View attachment 290397

View attachment 290398

What does lowering ppt do and highering edc? My ppt 120 is edc 110??

 

[gasolin]

What does lowering ppt do and highering edc? My ppt 120 is edc 110??

Explaining AMD Ryzen Precision Boost Overdrive (PBO), AutoOC, & Benchmarks | GamersNexus

stub With the launch of the Ryzen 3000 series processors, we’ve noticed a distinct confusion among readers and viewers when it comes to the phrases “Precision Boost 2,” “XFR,” “Precision Boost Overdrive,” which is different from Precision Boost, and “AutoOC.” There is also a lot of confusion...

www.gamersnexus.net

 

[Pivo187]

Explaining AMD Ryzen Precision Boost Overdrive (PBO), AutoOC, & Benchmarks | GamersNexus

stub With the launch of the Ryzen 3000 series processors, we’ve noticed a distinct confusion among readers and viewers when it comes to the phrases “Precision Boost 2,” “XFR,” “Precision Boost Overdrive,” which is different from Precision Boost, and “AutoOC.” There is also a lot of confusion...

www.gamersnexus.net

View attachment 290425

Is there an app that can test your cpu to create the best PPT and EDC for it kind of like using the master ryzen to get the best Curve op settings?

 

[gasolin]

I don't think so, like oc you have to find it yourself

 

[izy]

Is there an app that can test your cpu to create the best PPT and EDC for it kind of like using the master ryzen to get the best Curve op settings?

Just start from 142/105/105 and go lower till you are comfortable with temps/performance.

 

[Pivo187]

Im at 120 PPT and 110 EDC so Im mostly over your starting points lol

 

[izy]

Im at 120 PPT and 110 EDC so Im mostly over your starting points lol

Well you want to find the line where your cooling does not keep up and makes the CPU throttle , if you are going too low you will have to compromise ST with MT performance and i wouldnt compromise the ST, you will need to find the lowest values that doesnt affect performance (keeps the CPU lowest temps possible for that performance with the cooling you got) or if you prefer lower temps on the CPU find the values that you feel comfortable with for the performance / temps.

 

[Pivo187]

Well you want to find the line where your cooling does not keep up and makes the CPU throttle , if you are going too low you will have to compromise ST with MT performance and i wouldnt compromise the ST, you will need to find the lowest values that doesnt affect performance (keeps the CPU lowest temps possible for that performance with the cooling you got) or if you prefer lower temps on the CPU find the values that you feel comfortable with for the performance / temps.

I thought ppt and edc don't affect temps but tdc?

 

[freeagent]

PPT and EDC for sure, not sure about TDC though.. I don't think so..

 

[Emilychi]

Hmm how about 120 75 105.
Or should i just leave it on auto at this point with -15 corecycler

 

[gasolin]

corecycler ?