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

[jayjr1105]

NGL, I'm kinda burned out for the time being on tweaking. Seems very stable and cool unless I run OCCT with Large FFT's I get some errors. Small FFT set and it's free of any errors, likely because the extra heat forces the all core turbo to 4.2 or so instead of Large FFT's hitting 4.6. I will likely lower the CO offset from -10 to -5 and call it a day. System is running great otherwise with all the PBO stuff and my 3060ti undervolted.

It's weird, it just seems OCCT with Large data set, normal mode, avx2 it just hates my CPU. I can run every other combination and no errors. Just hates large data and always throws errors. This combination does allow the chip to all core boost to over 4.6. I am probably too heavily tuned for all core boost when I should be more towards single core boost for gaming like you guys said.

 

[jayjr1105]

Okay I did spend a little time messing around again. I'm 100% stable, no more errors. I'm still unsure on how to sway the PBO to focus on single threaded performance vs. an all core turbo. If I run a CPUZ benchmark, I am slightly above average on single core and way above average on multi.

Also just realized I'm like 3 BIOS updates out of date. Any harm in going fully up to date with a beta BIOS on MSI B550 Gaming Plus?

 

[mama]

Okay I did spend a little time messing around again. I'm 100% stable, no more errors. I'm still unsure on how to sway the PBO to focus on single threaded performance vs. an all core turbo. If I run a CPUZ benchmark, I am slightly above average on single core and way above average on multi.

Also just realized I'm like 3 BIOS updates out of date. Any harm in going fully up to date with a beta BIOS on MSI B550 Gaming Plus?
View attachment 214296

No harm in my view. I'm on the most recent BIOS for my MSI board and it's an improvement with my testing of it.

 

[outpt]

using the latest beta bios for my b550 tomahawk and it has cured a lot of problems. instabilites with the cpu and some usb disconects.

 

[tabascosauz]

NGL, I'm kinda burned out for the time being on tweaking. Seems very stable and cool unless I run OCCT with Large FFT's I get some errors. Small FFT set and it's free of any errors, likely because the extra heat forces the all core turbo to 4.2 or so instead of Large FFT's hitting 4.6. I will likely lower the CO offset from -10 to -5 and call it a day. System is running great otherwise with all the PBO stuff and my 3060ti undervolted.

It's weird, it just seems OCCT with Large data set, normal mode, avx2 it just hates my CPU. I can run every other combination and no errors. Just hates large data and always throws errors. This combination does allow the chip to all core boost to over 4.6. I am probably too heavily tuned for all core boost when I should be more towards single core boost for gaming like you guys said.

All-core tests are not how you test Curve Optimizer. Granted, for a stock 5600X with no +200 it's probably acceptable since all-core non-AVX and ST non-AVX shouldn't be far apart on clocks, but the point is to use OCCT or Corecycler to test one core at a time so it can test stability at the highest achievable ST boost clock. Once you set it up properly (been a while since I tested OCCT, a little rusty on how to set it up the core cycling mode), run it through all the cores one at a time for a fair amount of time each - Large FFT works well in both OCCT and Prime95.

The more cores you load, the more AVX you use, the more the freq drops (minimal for up to 2-3 cores, but drops appreciably after that). The more the freq drops, the more pointless the test is because the point is to see if your current curve (eg. -5) is stable on any given core at its highest boost clocks. You can still test AVX loads after making sure the SSE test is stable, just to be extra sure, but that comes after.

If all the cores are easily hitting the freq limit (4650MHz global stock limit), might be some value in trying out +200 boost override as it sounds like they have some extra ST headroom. eg. my 5600G and 5700G easily did 4650MHz (+200) and 4775MHz (+125) over their stock limits, with minimal undervolt.

 

[jayjr1105]

Here are OCCT's core cycling settings. Should I adjust anything here? BTW, I am now PBO + 200MHz and it's been very stable and very good temps. Games Boost to near 4.8 at times and all around I think I am close to my goal.

 

[tabascosauz]

Here are OCCT's core cycling settings. Should I adjust anything here? BTW, I am now PBO + 200MHz and it's been very stable and very good temps. Games Boost to near 4.8 at times and all around I think I am close to my goal.
View attachment 214402

I think I set mine a little longer than the default 3 seconds. About a minute per core. I tested stability using an extremely long Corecycler config, not OCCT. Just ran OCCT later for shits and giggles. In any case OCCT offers a lot of monitoring during the graphs so you can easily tell if it's actually working as intended.

Curve Optimizer stability is a pretty broad sliding scale in terms of what is acceptable for daily use and what isn't. It's up to you what you can put up with.

• Completely meaningless: able to boot Windows
• Potentially stable enough not to crash: able to run Cinebench
• Probably good enough for gaming: able to pass corecycler default config
• Probably good enough for most use cases: able to pass OCCT on decent
• Probably good enough for 99.9% of use cases: able to pass 5+ iterations of All FFT corecycler

 

[Mussels]

OCCT is stepping up! nice

 

[Jadz1ra]

Hello Guys,

I am a new user here and I would like share something pretty weird with 5800x.

Below you'll see what happens to the CPU at 100% as maximum state of the CPU and at 99%.
It was recorded in idle, no background apps running, only cpu-z and windows powerplan.

how CPU Ryzen 5800x react to maximum state of CPU - Powerplan

I would like to get to know if something similiar happen to you and if it is normal for cpu to change clock speed every single second.

 

[Mussels]

Hello Guys,

I am a new user here and I would like share something pretty weird with 5800x.

Below you'll see what happens to the CPU at 100% as maximum state of the CPU and at 99%.
It was recorded in idle, no background apps running, only cpu-z and windows powerplan.

how CPU Ryzen 5800x react to maximum state of CPU - Powerplan

I would like to get to know if something similiar happen to you and if it is normal for cpu to change clock speed every single second.

Yes. You're using a program that shows the speed of *one* core on an 8 core CPU, so its bouncing around showing various cores.
You have not locked it to 100% by changing that setting, that is NOT what it does.

 

[Taraquin]

So I messed with the curve optimizer some more. Set to -15 and started getting some errors in OCCT. Set down to -10 and still a few errors but less. Is this the only way to clear these errors up? Does adjusting EDC/TDC/PPT do nothing for these? Can I force PBO to just not boost as high? I'd be happy with a 4.5 all core tubo instead of 4.6.

I guess if I bring PPT down it will naturally downclock on all core loads correct?

Use core cycler and set CO individially Some core might handle - 30, some might only do - 5.

 

[DemonicRyzen666]

Yes. You're using a program that shows the speed of *one* core on an 8 core CPU, so its bouncing around showing various cores.
You have not locked it to 100% by changing that setting, that is NOT what it does.

My core 3 is my best on Ryzen master and preferred core in windows. cpu-z allows you to right click and select which core you're looking at.

In CTR2.1, this 5600X is rated golden sample, My cinebench23 scores do not reflect this at all. Although I'm only running Sata SDD and not a NVMe drive. My bios is set to CSM instead of UEFI too.
I did how ever bump my SOC voltage by one step, which seemed to have given me a boost, Ryzen master shows that it's I.O die wattage double, almost tripled in some instances which is odd.
I think I'm going to bump the bus speed up very slow. My ram is rated for 3200mhz but will do 3400mhz fine.

 

[freeagent]

5600X is 65w right? Overclocked it can do over 130w ppt..

 

[outpt]

Use core cycler and set CO individially Some core might handle - 30, some might only do - 5.

Core Cycler does NOT equal stability in every situation. I have used the program overnight and longer. Switch to gaming and it blue screens in less than 2 minutes.
Have gone -30 all core and various other settings. I am not though with the program at this point. According to event viewer every time it blue screens it’s core 4. I have put #4 to 0 and running to see what happens. Core 3 is at -30 as it “seems “ to be the most stable.

 

[jayjr1105]

Core Cycler does NOT equal stability in every situation. I have used the program overnight and longer. Switch to gaming and it blue screens in less than 2 minutes.
Have gone -30 all core and various other settings. I am not though with the program at this point. According to event viewer every time it blue screens it’s core 4. I have put #4 to 0 and running to see what happens. Core 3 is at -30 as it “seems “ to be the most stable.

Lately I run a small window of heaven benchmark or kombuster while stressing the CPU to get some heat from the GPU and to just give it some variety to potentially crash it if unstable.

 

[outpt]

Use a lightly threaded workload for stability.
As you stated unigine heaven is good. You don’t need to run the cpu full tilt. Just something that will loop.
I assume that you know this just covering the bases.

 

[freeagent]

Superpi 32m is good for single core stuff..

 

[Taraquin]

Core Cycler does NOT equal stability in every situation. I have used the program overnight and longer. Switch to gaming and it blue screens in less than 2 minutes.
Have gone -30 all core and various other settings. I am not though with the program at this point. According to event viewer every time it blue screens it’s core 4. I have put #4 to 0 and running to see what happens. Core 3 is at -30 as it “seems “ to be the most stable.

I did not say that, but corecycler can identify instability in many cases, but should be supplemented with other tests like occt, tm5 usmus etc

 

[DemonicRyzen666]

5600X is 65w right? Overclocked it can do over 130w ppt..

I'm right up to about 124 w PPT but I have PPT set to 310w ? only gets to about 40% on everything. :/

I'm so noob at this, I'm too use to ye olde phenom II/FX and A.O.D

I Think I might have to Nuc this windows; setup with UEFI bios setting instead of CSM. Also Switch from Sata SSD for NVMe.

I can't get over 1600 in cinebench23 in single thread 4,850mhz. My mutli-thread seems fine though 11,900-12,200 4550-4,700mhz

 

[Mussels]

I'm right up to about 124 w PPT but I have PPT set to 310w ? only gets to about 40% on everything. :/

I'm so noob at this, I'm too use to ye olde phenom II/FX and A.O.D

I Think I might have to Nuc this windows; setup with UEFI bios setting instead of CSM. Also Switch from Sata SSD for NVMe.

I can't get over 1600 in cinebench23 in single thread 4,850mhz. My mutli-thread seems fine though 11,900-12,200 4550-4,700mhz

you can set BIOS limits to a billion, but the chip has hardcoded limits the BIOS cannot go past - which is around 130W for that chip

 

[jayjr1105]

Well since I can't ever leave well enough alone, I splurged on a 5800X and will likely sell my 5600X to a friend. Looks like my gamble paid off so far. Seems the silicone for the more recent 5800X's are pretty good at least for thermals. Messed around for a good bit with PBO and such and this thing is barely warmer than my 5600X on all core loads and gaming and I haven't even touched curve optimizer yet. I did buy a new cooler last week (Fuma 2) and paste (MX-5) and perhaps those are helping as well. I liked the idea of having the most possible cores on a single CCD plus antonline ebay store had them for $350. Right now PPT 125 TDC 85 EDC 120

Alright guys, what's wrong here. After some reddit discussion, it appears my CPU won't get hot even with all PBO limits turned off during CB23. I even reset my BIOS and pulled the CMOS battery for a good 20 minutes. Re-enabled PBO and it just doesn't want to go higher than 120ppt on its own. What's the deal?

 

[Taraquin]

Well since I can't ever leave well enough alone, I splurged on a 5800X and will likely sell my 5600X to a friend. Looks like my gamble paid off so far. Seems the silicone for the more recent 5800X's are pretty good at least for thermals. Messed around for a good bit with PBO and such and this thing is barely warmer than my 5600X on all core loads and gaming and I haven't even touched curve optimizer yet. I did buy a new cooler last week (Fuma 2) and paste (MX-5) and perhaps those are helping as well. I liked the idea of having the most possible cores on a single CCD plus antonline ebay store had them for $350. Right now PPT 125 TDC 85 EDC 120

Alright guys, what's wrong here. After some reddit discussion, it appears my CPU won't get hot even with all PBO limits turned off during CB23. I even reset my BIOS and pulled the CMOS battery for a good 20 minutes. Re-enabled PBO and it just doesn't want to go higher than 120ppt on its own. What's the deal?

What MB do you have? Could it be MB lims for that CPU? Try curve optimuzer and I bet you would be pleased with the performance at 120 ppt

 

[jayjr1105]

What MB do you have? Could it be MB lims for that CPU? Try curve optimuzer and I bet you would be pleased with the performance at 120 ppt

MSI B550 Gaming Plus. Should be able to dump enough power into a 5800X. I eventually set limits to disabled in bios instead of motherboard. This was still the hottest I could get the chip

 

[VulkanBros]

Should I be worried? Just bought a 5800X and an EK-AIO 360 D-RGB.......haven't installed it yet......but after reading this thread, I am beginning to get a bit worried

 

[Mussels]

Should I be worried? Just bought a 5800X and an EK-AIO 360 D-RGB.......haven't installed it yet......but after reading this thread, I am beginning to get a bit worried

nah, some of them just run hotter than they should with single threaded loads, until you tweak PBO a little