• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

AMD Curve Optimizer any guides / experience

New AMD chipset driver - on x570 it seems stable P200T125E115
1612604096322.png

1612604108396.png
 
@harm9963 wow, thank you so much for the settings I think I have a winner, Best CPU-Z score I have ever had, great Cinebench R20 and R23 scores, voltage CPU Core SVI never goes beyond 1.469V) and temperatures are okish, never reaching 90C which is OK for Zen 3 CPU (According to AMD Technicians) and not thermal throttling.

I think you might have a better cooling solution and your ambient temperature is lower, I'm thinking of getting the Artic Liquid Freezer II 360mm. Could you tell me what you have? Thanks!

1. Voltage: Auto
2. LLC: Auto
3. 200/200/145 for PPT/TDC/EDC
4. Curve -30
5. Scalar: Auto
6. Max Boost Override: +50Mhz
7. Platform Thermal: Auto

View attachment 183490 View attachment 183491
View attachment 183492 View attachment 183493


Can you pass prime95 with those settings?
 
@lynx29 at the time it did. The problem with the CO is when leaving the machine idle. I'm no longer running those settings. I found better tools/procedures to test stability while using CO. I'm running the following:

X570 Aorus MASTER rev 1.2 with BIOS F32
Voltage: Auto
CPU LLC: Auto
PPT: 210
TDC: 150
EDC: 170
Scalar: Auto
CO: -15, -20, -20, -15, +2, -15, -25, -25, -25, -20, -20, -20, -25, -25, -15, -15
Max Boost: +50Mhz
Thermal Platform: 85C

Using F33a beta bios with AGESA 1.2.0.0 I get better curve but I rolled back to F32 official bios.

These settings pass prime95, OCCT, AIDA, etc.
 
@lynx29 at the time it did. The problem with the CO is when leaving the machine idle. I'm no longer running those settings. I found better tools/procedures to test stability while using CO. I'm running the following:

X570 Aorus MASTER rev 1.2 with BIOS F32
Voltage: Auto
CPU LLC: Auto
PPT: 210
TDC: 150
EDC: 170
Scalar: Auto
CO: -15, -20, -20, -15, +2, -15, -25, -25, -25, -20, -20, -20, -25, -25, -15, -15
Max Boost: +50Mhz
Thermal Platform: 85C

Using F33a beta bios with AGESA 1.2.0.0 I get better curve but I rolled back to F32 official bios.

These settings pass prime95, OCCT, AIDA, etc.


How did you know the CO negatives for each core? Is it that new tool that just came out?

 
@lynx29 basically you put -5 on all cores CO and test each core and if all of them passes, you add another -5. If one of the cores fails then you switch to per core CO and revert -5 to the core that failed you and then keep going.

In my case my core 4 (0 to 15) wasn't able to run stable using negative values or even 0, so I had to put +2. AGESA 1.2.0.0 fixes this for me but because i reverted to F32 I had to put +2 again.
 
Would guess it all comes down to the silicon lottery with me being able to run -30 allcore along with better cooling (?)

The last time i was lucky was with a newcastle cpu running at above 3ghz on air cooling!
View attachment 187113
Yes, that would certainly help. I cannot run -30 all-core. It will run great during benchmarks, and improves my scores a good bit, but when the computer is idle it will reset / WHEA error on my highest priority cores. Also, I'm assuming you have better cooling than me. I'm running on regular old air (NH-D15) and using a quiet fan profile.
 
This is a great thread with alot of good info.

Question if you set PBO to use motherboard limits.

The settings look high compared to what i've seen you guys using.

Ryzen master.png


Precision boost overdrive.png


on CO

Core 04 is my fastest and Core 01 second fastest.

-17 on Core 04
-15 on Core 01
-10 on the rest of my cores

CO.png
 
I found even with extreme good cooling there was no benefit for settings above 125 on Edc and Tdc, but it differ from motherboard to motherboard model and cpu lottery
 
I found even with extreme good cooling there was no benefit for settings above 125 on Edc and Tdc, but it differ from motherboard to motherboard model and cpu lottery

Good to know.

With my current settings its stable and performance is good the only thing that i see is high temps on a cinebench test. I will hit 90c on that. if I test with the aida64 stress test with AVX I hit 80c and everything else is well below that.
 
Question if you set PBO to use motherboard limits.

If you are OK with your temps, then motherboard limits are fine to use and should give you the best multi-core performance in all cases.
I would do some single core stability testing with the curve offsets you are running though, and potentially reduce max cpu boost clock override.
-17 on your fastest core along with +200 Max cpu boost clock override seems aggressive but this is going to be chip dependent, but if stable that is a pretty sweet result.
Typically people are able to set more negative values on the less favored cores than on their favored (highest performance) cores as those are already more aggressively tuned out of the box.
 
If you are OK with your temps, then motherboard limits are fine to use and should give you the best multi-core performance in all cases.
I would do some single core stability testing with the curve offsets you are running though, and potentially reduce max cpu boost clock override.
-17 on your fastest core along with +200 Max cpu boost clock override seems aggressive but this is going to be chip dependent, but if stable that is a pretty sweet result.
Typically people are able to set more negative values on the less favored cores than on their favored (highest performance) cores as those are already more aggressively tuned out of the box.

I need to do more testing on the off sets these are first set of numbers I tried. And pretty much left it at that to check stability so I've been on these for a week now no crashing on idle or Whea errors and its stable.

Choose 200Mhz for the Max cpu boast and can hit 5.05Ghz and that seems stable also. I under the impression that you set the higher negative offset on your faster cores.

As for silicon quality as well we all know it varies.

The CTR app for whats is worth says I have a golden sample if that means anything.

5800X CTR Golden Sample2.png
 
Last edited:
I need to do more testing on the off sets these are first set of numbers I tried. And pretty much left it at that to check stability so I've been on these for a week now no crashing on idle or Whea errors and its stable.

Choose 200Mhz for the Max cpu boast and can hit 5.05Ghz and that seems stable also. I under the impression that you set the higher negative offset on your faster cores.

As for silicon quality as well all know it varies.

The CTR app for whats is worth says I have a golden sample if that means anything.

View attachment 187705


Nice, mine comes back as silver and with a +200 offset I actually need to set a positive curve offset on 2 cores if I want 100% stability.
 
Dynamic OC switching/ -20
still working and still learning with this feature, ASUS DARK HERO.
plus warm day ,75f room !
 

Attachments

  • Capture013835.PNG
    Capture013835.PNG
    148.9 KB · Views: 304
  • Capture5301.PNG
    Capture5301.PNG
    487.2 KB · Views: 312
  • Capturetemp.PNG
    Capturetemp.PNG
    69.3 KB · Views: 325
  • Capturedocsw.PNG
    Capturedocsw.PNG
    98.3 KB · Views: 283
  • Capture1818.PNG
    Capture1818.PNG
    46.6 KB · Views: 278

Attachments

  • golden.png
    golden.png
    134.5 KB · Views: 244
I have upgraded to EK AIO 360 and it did improve my scores. However my ambient temp is too high (not using A/C).


CPU-Z 2021-02-11 21.43.14.png
CINEBENCH R20.060 2021-02-11 21.45.17.png
CINEBENCH R23.200 2021-02-11 21.51.59.png
 
Only -20 , DOCS , BIOS auto
 

Attachments

  • Capture20a.PNG
    Capture20a.PNG
    186.4 KB · Views: 272
  • Capture1823.PNG
    Capture1823.PNG
    24.2 KB · Views: 242
  • Capturetf.PNG
    Capturetf.PNG
    83.4 KB · Views: 228
Happy to report that CTR 2.1 beta 6 (for early subscribers only, I'm a patreon supporter of @1usmus) is looking pretty good. I was able to get way more performance than using PBO+CO. Take a look at mi Cinebench R23.

Using CTR 2.1 beta 6
=====================
PPT: 200
TDC: 150
EDC: 250
Temperature: 85C
PX: 5000/4850/4800 at 1480mV (For single boosting)
P2: 4650/4600 at 1356mV
P1: 4575/4525 at 1256mV

CINEBENCH R23.200 2021-02-21 17.20.10.png


I just wanted to add Cinebench R20

CINEBENCH R20.060 2021-02-21 17.52.57.png
 
that worked - if its unstable does it identify which core/thread it was on?
Yes, although I need to do something to record a log of where it is up to in case the whole computer crashes, it does monitor the results.txt file that prime95 writes and then reports any error as well as copying that to a new file named per core and loop iteration.
 
I have my 5800X up and running, and would like to do a little undervolting, however I am fairly confused.

1) How can I tell how much of an offset to set for each core? From what I read, it's during idle when too much can cause instability, so how the hell do I test that, and how can I even know what core is malfunctioning?

2) I learned there is something like quality of cores and that Hwinfo can show that, but I can't make ANY sense out of this:
1614103731862.png

What the heck does all this mean? I've already looked at Ryzen Master and cores 1 and 7 are the best, but these numbers I don't understand.

3) When taking the cores quality into consideration, which ones are to be run with higher lor lower curve offset? I keep reading conflicting information about this.

4) Is it enough to just play with the curve optimizer in my use case? My head is already like a baloon staring into the BIOS.



I have no idea what kind of sample do I have, probably utter shit, because all core boost in Prime95 is 4300MHz on average, and highest frequency frequency shown in Hwinfo over some period of time (not constant load) is 4850MHz.
 
Back
Top