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

[Oasis]

CO core 2,3,4 -30.

Which are you best cores and pbo power limits?

 

[outpt]

let's see: CO to -30 all core @4.7ghz. no real score change, but max voltage dropped from 1.431v to 1.313v

 

[Oasis]

let's see: CO to -30 all core @4.7ghz. no real score change, but max voltage dropped from 1.431v to 1.313v

Any AutoOC?

 

[outpt]

Any AutoOC?

Any AutoOC?

No RM auto over clock. Clock speeds to 47 multiplier, docp enabled. Don’t set temp limits just CO -30
Since voltages are down wonder if I can push up all core

 

[Oasis]

No RM auto over clock. Clock speeds to 47 multiplier, docp enabled. Don’t set temp limits just CO -30

so you can do a static overclock and PBO2 CO at the same time?

 

[outpt]

Curve optimizer is under pbo. I was able to. ymmv
Since 5ghz was a no no got to make it up some way

 

[Mussels]

Thanks

so like PPT 156, EDC 140 and then how much for TDC? 90? 100?

It's not like the numbers arent in the first post

(although i do intend to tidy up the text at some point)

 

[OneMoar]

-20 all core
nothing to report

 

[Mussels]

let's see: CO to -30 all core @4.7ghz. no real score change, but max voltage dropped from 1.431v to 1.313v

curve optimiser doesnt work with all core clocks... neither does PBO
You're seeing VID, the requested voltage - it may not be applying that actual voltage.
And 1.4v is going to degrade that CPU hella fast

What wattage and amperage are you drawing during those tests?

 

[mtcn77]

Thanks

so like PPT 156, EDC 140 and then how much for TDC? 90? 100?

Since your cpu is currently on manual multiplier, it wouldn't apply to you.
Post your temperatures which is so essential that I'm surprised you haven't already done so.

 

[Mussels]

VID and actual voltage are different, you cut off most relevant readings
seeing amps/watts is a better way to know what's being used

 

[outpt]

curve optimiser doesnt work with all core clocks... neither does PBO
You're seeing VID, the requested voltage - it may not be applying that actual voltage.
And 1.4v is going to degrade that CPU hella fast

What wattage and amperage are you drawing during those tests?

Eh forgot no CO on all core oc

 

[OneMoar]

1.4 - 1.5 is normal for lighter loads
All Core Manual OC do not exceed voltage is 1.300

 

[Ibizadr]

Curve optimizer isn't only put a random number in cpu cores and puffff magic happens and it runs at high clock with low voltage.
I'm in a run at least with one month off CO and corecycler runs. I'm able to run my games at 4.8+ GHz with no problems at max 60ºc. For example I have some cores at -26 and have cores at -10. Every cpu is different, what works for me can't work to others. If anyone want more info about how I'm testing and setting my pbo+Co I gave all info I have

 

[mama]

Those tuned parts are all stable and because I'm replacing my NH-L12s to a D15 I need to adjust my PBO power settings

Wait for the cooler then get into it.

 

[Mussels]

1.4 - 1.5 is normal for lighter loads
all core do not exceed voltage is 1.300

^ this
Auto can do higher volts because it obeys power limits, all core OC doesnt.
High amps kills the CPU's, and while no one really knows for sure we do have advice from some of the top tier OC'ers that degradation on zen 3 starts... somewhere

Edit: this info may be incorrect. Play safe with voltages.
Actually found a chart released by AMD about zen3, this might actually be a goldmine: "1.35v for sustained multithreading" -so i may change my recommendation to 1.35V all core if you're brave on zen 3 (but recommend 1.3v or lower)

This also makes sense, the more cores - the lower the safe voltage
(The voltages they list under "Stress" would be the goal here, i assume they accounted for Vdroop in the original number)

Apparently pre-launch, the 5800x was a 65W 5700x and they changed the PBO and rebranded it... it'd make sense why the 5800 performs so bad thermally if thats correct

 

[DemonicRyzen666]

^ this
Auto can do higher volts because it obeys power limits, all core OC doesnt.
High amps kills the CPU's, and while no one really knows for sure we do have advice from some of the top tier OC'ers that degradation on zen 3 starts... somewhere


Actually found a chart released by AMD about zen3, this might actually be a goldmine: "1.35v for sustained multithreading" -so i may change my recommendation to 1.35V all core if you're brave on zen 3 (but recommend 1.3v or lower)

View attachment 223043


This also makes sense, the more cores - the lower the safe voltage
(The voltages they list under "Stress" would be the goal here, i assume they accounted for Vdroop in the original number)

View attachment 223044

Apparently pre-launch, the 5800x was a 65W 5700x and they changed the PBO and rebranded it... it'd make sense why the 5800 performs so bad thermally if thats correct

Great info, however....
In that image in the link with the Infinity fabric that is chiplet to chiplet doesn't exist at all.
it's only the one from the chiplet to the IO die and back that's why going from CCD to CCD has a large latency hit still.

I've been watching my HWinfo on single cinebech my chip rarly uses it's 4,850mhz single thread it's usually only at 4,040mhz-4,339mhz most the run which makes no sense unless I'm temperature limited. I'm at 76C all core load is a bit high imo. I think I'm around 58C in single thread. This AIO is only a 25mm thick aluminum 360mm Radiator.

 

[tabascosauz]

Actually found a chart released by AMD about zen3, this might actually be a goldmine: "1.35v for sustained multithreading" -so i may change my recommendation to 1.35V all core if you're brave on zen 3 (but recommend 1.3v or lower)

That official slide is sus as hell for alleging that ~1.35V is stock all-core Vcore for Ryzen 5000. There is not a single instance of any 5600X/5900X/5950X that pulls more than even 1.25V in any respectable all-core workload, under stock conditions. One of the defining characteristics of Ryzen 5000 across the board is its revised boost algorithm that keeps Vcore low during all-core work, which is where those relatively amazing temps came from, compared to their Ryzen 3000 counterparts.

Unfortunately AGESA now imposes crippling limits on whatever it deems a power virus, so P95 is no longer a useful tool to test the chip's fitness limit and go off that. If you try that Vcore will be uselessly down at 1.0V. But I can tell you with certainty, that it isn't 1.35V.

Yes, stories of Ryzen 5000 degrading are far rarer than stories of Ryzen 3000 degrading, whatever that reason people believe. But the principles haven't changed. It's never been a matter of voltage, it's a matter of current combined with temps. Assuming you set the same OC, 4.6 @ 1.35V on a 5900X, for example:

• If you're only ever gaming on your computer, you will probably never see any ill effects during your ownership, because you will probably only be loading ~2 cores (stock Vcore for that would already be higher at ~1.35-1.45V due to only 2 core load). Ballpark 80-100W at most.
• If you're crunching/folding/rendering/chasing Cinebench scores for hours upon end every day, don't expect the chip to continue like that. Doubly so if you can't keep your temps under 75-80C.

Want to make sure your CPU will last? Don't run an all-core exceeding stock all-core Vcore, while loading it with all-core workloads frequently every day, at temps exceeding 75-80C.

There's a reason why the boost algorithm works the way it does.

• More cores? Vcore goes down.
• "Heavier" load (ie. heavier instructions, AVX)? Vcore goes down.
• Temps are higher? Vcore goes down.

When deciding on an all-core OC, your thinking should work the same way.

 

[Mussels]

That info was from 5000 series pre-release from MSI, there is a chance its wrong - if so i'll edit that post for people checking this thread in the future

 

[OneMoar]

That official slide is sus as hell for alleging that ~1.35V is stock all-core Vcore for Ryzen 5000. There is not a single instance of any 5600X/5900X/5950X that pulls more than even 1.25V in any respectable all-core workload, under stock conditions. One of the defining characteristics of Ryzen 5000 across the board is its revised boost algorithm that keeps Vcore low during all-core work, which is where those relatively amazing temps came from, compared to their Ryzen 3000 counterparts.

Unfortunately AGESA now imposes crippling limits on whatever it deems a power virus, so P95 is no longer a useful tool to test the chip's fitness limit and go off that. If you try that Vcore will be uselessly down at 1.0V. But I can tell you with certainty, that it isn't 1.35V.

Yes, stories of Ryzen 5000 degrading are far rarer than stories of Ryzen 3000 degrading, whatever that reason people believe. But the principles haven't changed. It's never been a matter of voltage, it's a matter of current combined with temps. Assuming you set the same OC, 4.6 @ 1.35V on a 5900X, for example:

• If you're only ever gaming on your computer, you will probably never see any ill effects during your ownership, because you will probably only be loading ~2 cores (stock Vcore for that would already be higher at ~1.35-1.45V due to only 2 core load). Ballpark 80-100W at most.
• If you're crunching/folding/rendering/chasing Cinebench scores for hours upon end every day, don't expect the chip to continue like that. Doubly so if you can't keep your temps under 75-80C.

Want to make sure your CPU will last? Don't run an all-core exceeding stock all-core Vcore, while loading it with all-core workloads frequently every day, at temps exceeding 75-80C.

There's a reason why the boost algorithm works the way it does.

• More cores? Vcore goes down.
• "Heavier" load (ie. heavier instructions, AVX)? Vcore goes down.
• Temps are higher? Vcore goes down.

When deciding on an all-core OC, your thinking should work the same way.

All Core manual overclock is NOT the same as Precision boost you loose a lot of safety margin with you go all core as you are can run the cpu WIDE OPEN with no throttling other then the TDP Limit which means parts of the core that may get REALLY HOT but not necessary trip a thermal event are at risk
I would suspect that degradation starts at some ware around 95-110A@85c for an all core manual overclock
at 120A @ 90c I would expect the cpu to incur instantaneous damage and we know a heavy AVX workload can get you there

and most games these days will use 4 or more threads so that point is moot a example of this is COD:Warzone It will bring 3 or 4 cores to 90%

 

[Mussels]

All core disables the safeties, basically.

They poll every 1ms, a thousand updates a second - that's where clock stretching and all that comes in, but it also lets cores disable or downclock to keep things safe.

Basically, static overclocks are scary on ryzen.

Page one is updated.

If you want your settings and info reported in the first post, post a screenshot of:
1. CPU model and cooling
2. PBO settings used
3. Open Hwinfo before you start R23 and wiggle your mouse around or whatever, so it can record your boost clock too.
Then start R23 and take a screenshot near the end of the 10 min run, so we can see the all core clocks under load and heat on those settings

Example:
5800x custom water
Curve: -10 all core
+200Mhz
PPT: 110W
TDC: 90A (notice how this wasnt at 100% like the others, its too high)
EDC100A
R23: 4.45GHz all core, 63C (4.925GHz boost)




And then @GerKNG efficiency settings:
Max efficiency:
PPT: 95W
TDC:60A
EDC: 90A
56C, 4.2GHz all core load, 4.9GHz boost
Note how the PBO values all match at 100% all together - Gerky really tuned these in!




A few dozen reboots later, my "performance" values are:
Curve: -20 all core
+200Mhz
PPT: 120W
TDC: 75A
EDC: 110A

Even on custom water it sits at 70C, and the clocks ease off a little at those temps making it fluctuate
Add in some GPU heat and i can see there being zero gains at all, vs the more efficient settings

Big oof: Ran OCCT's tester changing core every 3 minutes, but just one core at a time

All core: 70C

Single core: 80C ... cause the heats more densely packed

 

[Oasis]

Since your cpu is currently on manual multiplier, it wouldn't apply to you.
Post your temperatures which is so essential that I'm surprised you haven't already done so.

I will post PBO settings and Temps with the D15 after some more ram tuning

 

[mtcn77]

I will post PBO settings and Temps with the D15 after some more ram tuning

There has been a change of plans: post now, tune later. I wouldn't recommend too much perfectionism, it is a destination travelled not reached...

 

[outpt]

i can't help my self:4.95ghz. maybe there is hope and it's stable.

 

[ViperXTR]

Not sure if its just margin of error but i get somewhat a little bit more score on R23 on my older bios (Beta January 2021) vs the newer one (september, MSI Mag B550M Mortar), or maybe perhaps i just missed to teak something, gonna take a look back on my other settings
PPT 120, TDC 85, EDC 110, CO Best core negative 5, 2nd best negative 8, the rest is 13, max temp 82C on a Corsair H60 push pull (ST: ~4850Mhz, MT ~4550Mhz - ~4650Mhz)