Asus Dark Hero + 5950X Bios/PBO/Temps questions for non gaming PC

professional loser

Limp Gawd
Joined
Mar 22, 2018
Messages
134
Was not sure if I should start this thread under MOBO or CPU section, anyway...

I just built my new workstation a few days ago, I play zero games and the PC will be used for content creation that utilize all 16 cores when rendering. Stability and longevity are my main concerns rather than few % gains from any dangerous/risky overclock!!!
That said, if I get extra performance without any real risk, then I am all for it!

MOBO - X570 Dark Hero, CPU - Ryzen 5950x, Cooler- Noctua NH-D15 (2x 140mm fans), Case- Meshify 2XL, Noctua 140mm fans everywhere, excellent airflow etc, Memory- G.Skill Ripjaws, 2x, 32GB, DDR4-3600, DIMM 288, F4-3600C16D-64GVK, Hynix- H5ANAG8N[M/A]JR-VKC.



When using Cinebench 23, Multithread (DOCP Standard Enabled on all readings/settings - CL16-22-22-42), I get the following readings-


Settings 1-
Vanilla Default F5 Bios setup, Sustained all core reading of 3824MHz and Max Temp 63 C, Score 25467

Settings 2- Extreme Tweaker > PBO Enabled, Sustained on all cores reading of 4300MHz, Max Temp 81C, Score 28831

Settings 3- Extreme Tweaker > PBO Enabled + Fmax enhancer Enabled, Sustained all cores reading of 4449 MHz, Max temp 71C, Score 26831

Settings 4A-
Extreme Tweaker > PBO Enabled + Fmax Enhancer Disabled, PBO scaler Enabled x3, Thermal Throttle Manual 85c, Sustained on all cores reading of 4424 MHz, Max temp 81C, Score 28870

Settings 4B-
Extreme Tweaker > PBO Enabled + Fmax Enhancer Enabled, PBO scaler Enabled x3, Thermal Throttle Manual 85c, Sustained on all cores reading of 4449 MHz, Max temp 70C, Score 26289

Settings 4C- Extreme Tweaker > PBO Enabled + Fmax Enhancer Disabled, PBO scaler Enabled x4, Thermal Throttle Manual 85c, Sustained on all cores reading of 4324 MHz, Max temp 81C, Score 26279

I really don't know what to make of it all! lol




EDIT- UPDATED results further down the thread!
 
Last edited:

Light1984

Gawd
Joined
Dec 31, 2004
Messages
588
What has worked best for me is PBO advanced enabled with all core negative 15 on curve optimizer, +200 on boost, fmax enhancer disabled, power targets set to auto. I have a 5800x though, so YMMV.

Edit: the pbo under extreme tweaker is the old one, use PBO2.0 under AMD overclocking. Leave extreme tweaker set to auto.

Watch this tutorial. This is the best explanation I've found.

 

IndyColtsFan

Limp Gawd
Joined
Jul 4, 2017
Messages
489
I have a similar config as you and get even higher temps on Cinebench, but I may be running R20 vs R23. At first, I thought I might need to reapply thermal paste or maybe switch to an AIO but after doing a lot of reading, I think even my temps are fairly normal for the 5950x. I think you probably have better airflow than I do, which would account for the 5-7 degree difference I’m seeing. I’m just going to stop worrying about it and if it becomes an issue, then I might try the “new thermal paste” route or an AIO.
 

soulesschild

Supreme [H]ardness
Joined
Feb 18, 2007
Messages
6,176
What has worked best for me is PBO advanced enabled with all core negative 15 on curve optimizer, +200 on boost, fmax enhancer disabled, power targets set to auto. I have a 5800x though, so YMMV.

Edit: the pbo under extreme tweaker is the old one, use PBO2.0 under AMD overclocking. Leave extreme tweaker set to auto.

Watch this tutorial. This is the best explanation I've found.



This is the guide I have followed and gotten great results.
 

professional loser

Limp Gawd
Joined
Mar 22, 2018
Messages
134
Hello again. Been testing for last 1 day. Lot to report but need to understand something clearly before I post my findings. When using HWinfo64, there is CPU temp and CPU package Temp. I assume CPU Package is the real temperature? CPU package temp corresponds to the temp that Ryzen Master shows.

I have a similar config as you and get even higher temps on Cinebench, but I may be running R20 vs R23. At first, I thought I might need to reapply thermal paste or maybe switch to an AIO but after doing a lot of reading, I think even my temps are fairly normal for the 5950x. I think you probably have better airflow than I do, which would account for the 5-7 degree difference I’m seeing. I’m just going to stop worrying about it and if it becomes an issue, then I might try the “new thermal paste” route or an AIO.

So what you are saying is that as long as the temp is under 85-90c, you are not bothered? Correct?
 
Last edited:

IndyColtsFan

Limp Gawd
Joined
Jul 4, 2017
Messages
489
Hello again. Been testing for last 1 day. Lot to report but need to understand something clearly before I post my findings. When using HWinfo64, there is CPU temp and CPU package Temp. I assume CPU Package is the real temperature? CPU package temp corresponds to the temp that Ryzen Master shows.



So what you are saying is that as long as the temp is under 85-90c, you are not bothered? Correct?
Correct, at least for now. Stability is the most important thing for me so I’d consider reducing overclock if necessary. Even if I reapplied thermal paste, I doubt I’d see a huge improvement and I’m a little leery of AIOs after I had a pump die on one. I have zero interest in a custom loop either.
 

professional loser

Limp Gawd
Joined
Mar 22, 2018
Messages
134
OK! Here comes, ROUND 2!!!!! I have about 20 different combinations but will post the most relevant few-
DOCP for all, G.Skill Ripjaws, 2x, 32GB, 3600 CL16

A1- CURVE- 15

Extreme Tweaker all default apart from, PBO Fmax Enhancer Disabled
AMD Overclocking-
PBO- Limits- Motherboard
All Cores Negative -15
During benchmark all cores 4450MHz

Temp CPU package 85 C, CPU 74 C

Multi core test: C20:11531, C23: 29466/1656

______________________

A2- CURVE- 20

Extreme Tweaker all default apart from, PBO Fmax PBO Fmax Enhancer Disabled

AMD Overclocking-
PBO- Limits- Motherboard
All Cores Negative -20
During benchmark all cores 4475MHz

Temp CPU package 86 C, CPU 76 C

Multi core test: C20:11521 (I did not do C23 as the score was lower than the above reading anyway plus more heat)

____________________


A3- CURVE- 25
Extreme Tweaker all default apart from, PBO Fmax PBO Fmax Enhancer Disabled

AMD Overclocking-
PBO- Limits- Motherboard
All Cores Negative -25

During benchmark all cores 4450 - 4475 MHz

Temp CPU package 88 C, CPU 78 C

Multi core test: C20:11614 C23: 29695

Single core Test- CRASHED the PC

____________________

A4- PER CORE CURVE
With default Bios settings and vanilla PBO, I observed the max boost clock of various cores during multithreaded tests and divided the cores into three categories.
Then applied PER CORE curves in PBO-2.
The Best cores got Minus 12
Average cores got Minus 17
The Worst cores got Minus 20

Temp CPU package 85 C, CPU 74 C

Multi core test: C20: 11511, C23: 29376

____________________


Notes-

1- With all the above settings I also tried adding Boost Clock ranging from 50MHz to 200MHz. Did not seem to make any difference.
2- PBO Limits when using curve optimizer -
-
If I choose AUTO then it would act as if PBO was off. 3800MHz on all cores when using C23 bench!
- If I manually input the standard 5950X settings of: PPT- 142W , TDC- 95A , EDC- 140A then I assume it is same or similar to Auto as the cores in multi thread test would again stay at around 3800MHz
- PBO- Limits set to MOTHERBAORD is when I got the highest scores.

3 - PPT, TDC, EDC


a- According to Ryzen Master during standard Asus Extreme Tweaker > PBO ON + F Max Disabled, PPT = 395W, TDC = 255A, EDC =200A


Aus TweakerPBO.PNG

b- According to Ryzen Master, when PBO limits are set to MOTHERBOARD, All Core Negative 12, the PPT ,TDC & EDC settings are same as above.

PBO-PowerCurve-Minus12.PNG
 
Last edited:

Light1984

Gawd
Joined
Dec 31, 2004
Messages
588
Yea, pbo limits set to motherboard is going to generate the most heat. I would leave them set to auto or manually set them for longevity sake. The extra heat is probably not worth the score bump.
 

IndyColtsFan

Limp Gawd
Joined
Jul 4, 2017
Messages
489
OK! Here comes, ROUND 2!!!!! I have about 20 different combinations but will post the most relevant few-
DOCP for all, G.Skill Ripjaws, 2x, 32GB, 3600 CL16

A1- CURVE- 15

Extreme Tweaker all default apart from, PBO Fmax Enhancer Disabled
AMD Overclocking-
PBO- Limits- Motherboard
All Cores Negative -15
During benchmark all cores 4450MHz

Temp CPU package 85 C, CPU 74 C

Multi core test: C20:11531, C23: 29466/1656

______________________

A2- CURVE- 20

Extreme Tweaker all default apart from, PBO Fmax PBO Fmax Enhancer Disabled

AMD Overclocking-
PBO- Limits- Motherboard
All Cores Negative -20
During benchmark all cores 4475MHz

Temp CPU package 86 C, CPU 76 C

Multi core test: C20:11521 (I did not do C23 as the score was lower than the above reading anyway plus more heat)

____________________


A3- CURVE- 25
Extreme Tweaker all default apart from, PBO Fmax PBO Fmax Enhancer Disabled

AMD Overclocking-
PBO- Limits- Motherboard
All Cores Negative -25

During benchmark all cores 4450 - 4475 MHz

Temp CPU package 88 C, CPU 78 C

Multi core test: C20:11614 C23: 29695

Single core Test- CRASHED the PC

____________________

A4- PER CORE CURVE
With default Bios settings and vanilla PBO, I observed the max boost clock of various cores during multithreaded tests and divided the cores into three categories.
Then applied PER CORE curves in PBO-2.
The Best cores got Minus 12
Average cores got Minus 17
The Worst cores got Minus 20

Temp CPU package 85 C, CPU 74 C

Multi core test: C20: 11511, C23: 29376

____________________


Notes-

1- With all the above settings I also tried adding Boost Clock ranging from 50MHz to 200MHz. Did not seem to make any difference.
2- PBO Limits when using curve optimizer -
-
If I choose AUTO then it would act as if PBO was off. 3800MHz on all cores when using C23 bench!
- If I manually input the standard 5950X settings of: PPT- 142W , TDC- 95A , EDC- 140A then I assume it is same or similar to Auto as the cores in multi thread test would again stay at around 3800MHz
- PBO- Limits set to MOTHERBAORD is when I got the highest scores.

3 - PPT, TDC, EDC


a- According to Ryzen Master during standard Asus Extreme Tweaker > PBO ON + F Max Disabled, PPT = 395W, TDC = 255A, EDC =200A


View attachment 356108

b- According to Ryzen Master, when PBO limits are set to MOTHERBOARD, All Core Negative 12, the PPT ,TDC & EDC settings are same as above.

View attachment 356109

Yeah, your R20 temps are around what I'm seeing too and around the same performance (11200-11600 in my case, IIRC). You might be 1 degree cooler and that's probably due to airflow differences (My NH-D15S has a 120 mm fan in front and the 140 mm in the middle).

I just installed my RTX 3080 last weekend and haven't benched since then to see if it is dumping enough heat into the case to alter my previous results.
 

professional loser

Limp Gawd
Joined
Mar 22, 2018
Messages
134
Yea, pbo limits set to motherboard is going to generate the most heat. I would leave them set to auto or manually set them for longevity sake. The extra heat is probably not worth the score bump.
Yeah, your R20 temps are around what I'm seeing too and around the same performance (11200-11600 in my case, IIRC). You might be 1 degree cooler and that's probably due to airflow differences (My NH-D15S has a 120 mm fan in front and the 140 mm in the middle).

I just installed my RTX 3080 last weekend and haven't benched since then to see if it is dumping enough heat into the case to alter my previous results.

So I set PBO limits manually!

A5- CURVE - 12
Extreme Tweaker all default apart from, PBO Fmax PBO Fmax Enhancer Disabled

AMD Overclocking-
PBO- Limits-
PPT 200W, TDC 200A, EDC 150A
All Cores Negative -12
During benchmark all cores boost to 4700MHz but settle on 4425MHz

Temp CPU package 77 C, CPU 66 C A BIG DROP IN TEMP!!!!

Multi core test: C20:11360 C23:29588

That is a respectable, safe looking boost, but I don't know what the long term consequences of using settings labelled A5 will be? For now, I will just go back down to default and not be too greedy?

A5B- CURVE - 12

Same minus 12 curve as above but PBO limits set to AUTO gives-
Temp CPU package
64 C, CPU 54 C
During benchmark all cores boost to 4650MHz for few seconds only and then settle on 3900 MHz.
Multi core test: C23:26287


Feels like riding a bicycle after riding a motorbike! For the low heat generated and reasonable performance, A5B setting might be be a good compromise?
 
Last edited:

Azrak

[H]ard|Gawd
Joined
Oct 4, 2015
Messages
1,068
Not to throw a monkey wrench into your benching sessions, but when I tried using All Cores Negative -5 (yeah, just -5!), I experienced an idle system reboot while browsing a webpage. So it wasn't high load that caused a problem, but rather just an idle system. No idea why it happened and have not had time to investigate further yet. My point is: Spending all of your time benching high loads to learn how hot things get may not find all of the possible issues you can have by setting curve optimizer. And for the record, I don't even know if the -5 all core setting had anything to do with it. But it has not happened again since I went back to Auto, so I assume it was a primary culprit. Maybe I lost the silicon lottery, too, who knows.
 

Light1984

Gawd
Joined
Dec 31, 2004
Messages
588
Yea to do it right you'll need to go through and find tune each core. When you get a reboot check for whea errors in event viewer to see which core it was and adjust accordingly.
 

professional loser

Limp Gawd
Joined
Mar 22, 2018
Messages
134
Not to throw a monkey wrench into your benching sessions, but when I tried using All Cores Negative -5 (yeah, just -5!), I experienced an idle system reboot while browsing a webpage. So it wasn't high load that caused a problem, but rather just an idle system. No idea why it happened and have not had time to investigate further yet. My point is: Spending all of your time benching high loads to learn how hot things get may not find all of the possible issues you can have by setting curve optimizer. And for the record, I don't even know if the -5 all core setting had anything to do with it. But it has not happened again since I went back to Auto, so I assume it was a primary culprit. Maybe I lost the silicon lottery, too, who knows.

It is rock stable with 0 WHEA Errors. I did the experimenting over 3 days. Was browsing online for hours and also left the PC idling for many hours. Never crashed!
The only time it crashed was when I tried -25 on all cores.

Anyway for now it is back on plain normal stock settings with PBO disabled. 32 to 63C, C23: 25550
 

professional loser

Limp Gawd
Joined
Mar 22, 2018
Messages
134
It is like a drug!!!! Started overclocking again, and I think I will stick with the following settings-

Best cores Minus 12
Mid level cores Minus 17
Worst cores Minus 20
PPT,TDC,EDC: 188,188 & 145
I get Cinebench 23 score of 29600
Max Temp of 77c after 1 hour of Stability test
V core stays around 1.120 (have not touched anything apart from the PBO curves). Spikes to 1.4 Max
 

soulesschild

Supreme [H]ardness
Joined
Feb 18, 2007
Messages
6,176
I've yet to really do power limit tuning. I might cut down on my EDC as that's usually the culprit for massive heat.
 

Light1984

Gawd
Joined
Dec 31, 2004
Messages
588
It is like a drug!!!! Started overclocking again, and I think I will stick with the following settings-

Best cores Minus 12
Mid level cores Minus 17
Worst cores Minus 20
PPT,TDC,EDC: 188,188 & 145
I get Cinebench 23 score of 29600
Max Temp of 77c after 1 hour of Stability test
V core stays around 1.120 (have not touched anything apart from the PBO curves). Spikes to 1.4 Max


So YMMV, but I thought the CO should be set the way you have it. However, some other users in another thread told me I had it backwards and that my good cores should get more negative offset with my "bad" cores less negative offset. I gave it a shot and my scores improved quite a bit. It also allowed me to hit higher clocks. I can typically hit 5050 MHz on 6 or 7 out of 8 cores under perf in hwinfo 64 and 4955 MHz effective now. Should be even better once I get things under water.
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,785
So YMMV, but I thought the CO should be set the way you have it. However, some other users in another thread told me I had it backwards and that my good cores should get more negative offset with my "bad" cores less negative offset. I gave it a shot and my scores improved quite a bit. It also allowed me to hit higher clocks. I can typically hit 5050 MHz on 6 or 7 out of 8 cores under perf in hwinfo 64 and 4955 MHz effective now. Should be even better once I get things under water.

The good cores should get LESS negative offset, since they will clock higher, more frequently. EG. my setup, all of my cores at -15, but my best 2 cores & second best cores are -5, to allow them more voltage for higher clocking scenarios.
 

Light1984

Gawd
Joined
Dec 31, 2004
Messages
588
The good cores should get LESS negative offset, since they will clock higher, more frequently. EG. my setup, all of my cores at -15, but my best 2 cores & second best cores are -5, to allow them more voltage for higher clocking scenarios.

That's what I originally thought too, but because they are "better" cores, they can reach the same clocks on less voltage. That's what makes them better. Hence the "better" cores can get a larger negative offset. The "bad" cores need more voltage (less negative offset) to reach higher clocks. I did preface what I said with YMMV, but for me, I was able to get much better scores in CB, everything else kept the same for control, with more negative offset on good/better cores and more voltage (less negative offset) to "bad" cores.

Check out page 3 here for our other discussion with some data points from others.

https://hardforum.com/threads/first-amd-build-in-20-years.2009575/
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,785
That's what I originally thought too, but because they are "better" cores, they can reach the same clocks on less voltage. That's what makes them better. Hence the "better" cores can get a larger negative offset. The "bad" cores need more voltage (less negative offset) to reach higher clocks. I did preface what I said with YMMV, but for me, I was able to get much better scores in CB, everything else kept the same for control, with more negative offset on good/better cores and more voltage (less negative offset) to "bad" cores.

To be clear, even if you put -50,the cores will try to reach the same clocks (defined by the CPU/mobo settings with the +xxx offset as well) , it will just cause your computer to crash :). So the reverse is true, Windows/AMD setup the CPU to use the favored cores for more demanding workloads, and also get clocked higher more often, they need the additional power. I'm 99.9% sure of this from my own testing as well.

Everything at -15 = crashes, caused by the "best" core
Everything at -15 apart from the fastest 4 cores(-5) = No crashes.
 

lobstar

Limp Gawd
Joined
Nov 24, 2004
Messages
300
I'm really confused about the curve optimizer. I have a 5950x. I reset the bios back to defaults on my dark hero. I enabled advanced PBO, set the limits to motherboard, and managed to get every core to -30 except for two (one -28, one -27, both on the second CCD). The max temps are 72C during cinebench 23 but I still can't get as good performance as a per CCD overclock of 47x/46.5x at the same temps and that lets me run FMax to hit 5.25ghz single core. I used a negative vcore offset of 0.0625v. How are people getting better performance?
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,785
I'm really confused about the curve optimizer. I have a 5950x. I reset the bios back to defaults on my dark hero. I enabled advanced PBO, set the limits to motherboard, and managed to get every core to -30 except for two (one -28, one -27, both on the second CCD). The max temps are 72C during cinebench 23 but I still can't get as good performance as a per CCD overclock of 47x/46.5x at the same temps and that lets me run FMax to hit 5.25ghz single core. I used a negative vcore offset of 0.0625v. How are people getting better performance?

When you use PBO + FMAx to get higher clocks, you don't always get better performance as you're still going to be limited by something. Use LLC 2 or 3 and your PBO scores & Temps will skyrocket :p
 

lobstar

Limp Gawd
Joined
Nov 24, 2004
Messages
300
My single core score is definitely higher with fmax when I'm not using the curve optimizer. Effective clocks match the reported clocks. I use LLC 3 every day. My water temps are 25-26C.
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,785
My single core score is definitely higher with fmax when I'm not using the curve optimizer. Effective clocks match the reported clocks. I use LLC 3 every day. My water temps are 25-26C.

That has always been the case, manual OC limits single core, but boosts multi core.

Are you using the dynamic OC switch?
 
Top