AMD Curve Optimizer any guides / experience

[juanyunis]

@Zach_01 for some reason I read 5800X lol. So what do you think about my screenshots? Should I keep it like that? Or try to tune it more? Like I said before I'm using the pc for VMs and some heavy gaming. Do you think moving to a 360mm radiator from a 240mm would help?

 

[Zach_01]

5800X has also the same limits with 5900X/5950X.

As for the AIO, not too much I would say, on the same ambient/case conditions. It would probably improve temp by a little but not so much that can justify the cost of it and the potential performance gains, if there is any.

It’s not like going from conventional cooling (air/water) that always rely on ambient temp, to a sub-zero cooler, like a chiller.

 

[bubbleawsome]

@bubbleawsome does your reach more than 1.5V? Or close like 1.498V? What is your motherboard? If you are running everything stock at the moment you can try enabling PBO in advanced mode and put some initial values, if your processor is 65W TDP then maybe you can try running PPT at 220, TDC/EDC at 120 and the curve -15, scalar auto and platform thermal auto, max boost at 0.

Edit: I though you had a 5800X, still those settings apply to you, you can increase the PPT to 240.

The motherboard is the MSI MEG Unify, and ryzen master and HWinfo say it's running right about 1.485v but HWinfo CPU Core Current (Effective) goes up to 1.500v sometimes. I tend to get errors going over about -10 in the voltage curve.

The second is the one with max load, on the last was already finished.


Sounds about right... As for what is the best EDC value has no straight answer. It depends, mostly on temp.
From my understanding, using the curve optimizer with custom PBO limits you are by passing some of the protection of the chip. Mostly on current. It should be used with care and awareness of what is what. Temperature is an indication of stress but doesn’t tell the whole story. EDC too.

For example (and I’m going to use random numbers) having EDC, let’s say 140A at 70C is ok, but having 140A at 90C is not. The speed or voltage at those 2 cases is less significant but still play a role.
Same applies to every EDC value. It’s the combination of current(A), voltage and temp.

Keep temperature as far as possible from max operating limit (95C), under 80C I would suggest, EDC in reasonable levels (120~130) and all other (voltage, speed, watt) will fall were they should.

I’m not afraid of the 1.45-1.5V on single thread, nor the 200W PPT on multi.
It’s the high temp/EDC combination that can tear the chip’s internal traces rather quickly.


The 5900X has the exact same stock PBO limits with 5950X.

I believe those are

PPT: 142W
EDC: 140A
TDC: 90A

I'll keep that in mind and try out those PBO limits. I'd like to say it never ever goes over 80c with the dark rock pro 4 I've got on there, but once the case heats up CCD1 can reach towards 85c as is. I don't think I've ever seen it pass 85c though no matter what I'm doing.

 

[mtcn77]

The motherboard is the MSI MEG Unify, and ryzen master and HWinfo say it's running right about 1.485v

What is up with the voltage? MSI takes the high load as primary, so you might want to consider 1.25v as the maximum.

 

[bubbleawsome]

What is up with the voltage? MSI takes the high load as primary, so you might want to consider 1.25v as the maximum.

That is with a single core load, under 24-thread loads it's down as low as 1.114v. I'll do a quick average while gaming and see where it's trending.

 

[mtcn77]

I'll do a quick average while gaming and see where it's trending.

It would be better if you checked how much ST heats up the working core. Don't encroach upon TjMax too much.

 

[bubbleawsome]

It would be better if you checked how much ST heats up the working core. Don't encroach upon TjMax too much.

The highest I've ever seen on either CCD even with the case and room as warm as they get is 87c on CCD1. Since then I've done some tuning and never seen it that hot again. Under the gaming I've just done the hottest it got was 80.0c, with the measured TDC and EDC maxing out in the high 50A range, voltage maxing at 1.49 effective but averaging 1.43 effective. Not sure what the difference is between voltage and voltage effective

 

[mtcn77]

Not sure what the difference is between voltage and voltage effective

Probably vdroop.
What I recon is, this power optimizer does not sit well with overvolting.
I thought I could explain it using the boost bins if it were applying overvoltage at the highest bin, therefore dropping the lower bins to an undervolted setting, but there is no precedence AMD could setup such a novelty. If it had been, that would be great, overclocking just the highest bin setting, that is.

 

[Zach_01]

"Core 0, 1, 2... VID" is the requested voltage of each core.

The power supply system cannot supply individual core voltages, only one voltage can be supplied to the CCDs (1 or 2)
So...

"CPU Core VID (effective)" is the one and only voltage request by the CPU. But that is not what the CPU is getting. Its only a request.
"Vcore" is what the board reports for CPU voltage by its own(board) sensor.
"VR OUT" is what the VRM reports for CPU voltage by its own(VRM) sensor.
"CPU Core Voltage (SVI2 TFN)" is what actually, or as closest as possible, the CPU is getting and it is read internally (in CPU).

According to HWiNFO author the SVI2 TFN is the most accurate and close to the true value that the CPU is getting. I've write it before a few posts back.

I have red box all these voltage sensors for easy location


Also, Ryzen CPUs do not have dedicated temp report of each core like Intels. They do have a few dozen sensors per CCD though. Instead of reporting dedicated (static) core sensors, they report the "Tctl/Tdie" which is the hotest spot inside all package (CCD1, CCD2, IOD). This one is switching between all package sensors (could be over 100) almost instantly and report always the highest one.
The "Die (average)" temp is the average of all package sensors or only of all CCD1/2 sensors.
The "CCD1, CCD2 (Tdie)" is from a dedicated(static) sensor located on one side of each CCD.

When the CPU is fully loaded all 4 values should be around the same as heat is destributed almost equally to all die area. When load is light or medium (like gaming) these values are different as not all parts of the CPU is loaded equally.

 

[bubbleawsome]

Ah, very useful. In that case I'm seeing SVI2 at 1.49 is the highest it's ever been, usually closer to 1.47. Tctl/Tdie is maxing right under 80, CCD1 tends about 2-4c hotter than CCD2 even under multicore. I figure that's due to the leakier cores being on that one

 

[Zach_01]

Ah, very useful. In that case I'm seeing SVI2 at 1.49 is the highest it's ever been, usually closer to 1.47. Tctl/Tdie is maxing right under 80, CCD1 tends about 2-4c hotter than CCD2 even under multicore. I figure that's due to the leakier cores being on that one

Not exactly... It may indicate that CCD1 is doing more work and running higher speeds than CCD2.
See the screenshot the "Core 0, 1, 2... Clock (perf #x/x)"? Those numbers are the order of the performance cores. The second number represents the order hardcoded in the CPU and first is the choice of Win scheduler for core loading order. #1 is the highest perf.
Depending on scheduler, PowerPlan and Chipset drivers, Windows may be "smart" enough to load the proper cores for best performance and low possible latency. This is important for single threaded or reduced threaded applications like gaming.
You can see it by the average values of...
1. The "Core 0, 1, 2... T0, T1 Effective Clock"
2. The "Core 0, 1, 2... T0, T1 Usage"
3. The "Core 0, 1, 2... Ratio"
4. The "Core 0, 1, 2... C0 Residency"

The higher perf #x/x cores should have higher average values on the above 4 multiple readings.
Check all those under various loads and see if the high performance cores are getting the most load.

 

[jesdals]

Did some testing with
PPT: 130
TDC: 125
EDC: 90
Curve: - 5 on highperformance cores 0, 3, 12 and 15 and - 10 on the rest

But here the multicore performance in the CPU-Z benchmark is stil beneath 4125MHz - but tempc stay under 60c

Now changeing those settings to these below seems to raise multicore performance above 4300MHz and give 5100MHz singlecore boost - but it cost on temps that get above 60c

will do some testing on above settings and try to tweak them - curve settings is stil as shown above

 

[Zach_01]

Did some testing with
PPT: 130
TDC: 125
EDC: 90
Curve: - 5 on highperformance cores 0, 3, 12 and 15 and - 10 on the rest

But here the multicore performance in the CPU-Z benchmark is stil beneath 4125MHz - but tempc stay under 60c

Did you really wanted those PBO settings?
Stock are this

PPT: 142
TDC: 90
EDC: 140

Did you accidentally confuse EDC with TDC and wanted them like this?

PPT: 130
TDC: 90
EDC: 125

 

[jesdals]

Setting
PPT: 250
TDC: 90
EDC: 125

Actually gave lower multi performance below 4100MHz - but temps under 60c

Will do some testing with
PPT: 250
TDC: 125
EDC: 110

These settings give me above 4375MHz multicore - but above 75c on temps single core is still high

Wil test stability here and see if I can find a sweat spot

 

[juanyunis]

@jesdals i think you found a pretty sweet spot already, you still reach 5ghz boost, no clock stretching voltage lower than 1.5V and temperatures are less than 80C., . It makes me want to try it now.

What scores you got for cpu-z ST/MT. Have you changed any other bios settings? Are you still using turbo for LLC and Auto voltage for cpu core?

 

[jesdals]

Still stability testing but got this CPU-Z bench result - not to shabby

 

[juanyunis]

Still stability testing but got this CPU-Z bench result - not to shabby

View attachment 183369

Is this with your latest settings? Also, I asked you if you wanted any other bios settings like voltage etc. It looks pretty good, do you have cinebench?

 

[mtcn77]

The thing is, you won't lose performance should you decrease these settings. We still have the temperature going for us. So, if you were to decrease EDC somewhat, MT scores would start to pick up since it would redirect power to more stable workloads(PPT). You should definitely chart it out starting from a very throttled setting up to where you want to be. Keep PPT there and just advance EDC in increments, maybe 5-10A each time. If you get to a point where MT core average crests over, do the other curve optimizer stuff alongside it. Just don't leave it to doubt.

 

[jesdals]

Is this with your latest settings? Also, I asked you if you wanted any other bios settings like voltage etc. It looks pretty good, do you have cinebench?

Here is my current settings (did try to lower the target to 300MHz to see if it makes a difference)

My cpu settings is basic - usally I would disable AMD Cool&Quiet

My Loadline settings is the same above in previous test

My Samsung B-die voltage and XMP is the only special setting in general

Did start with SoC adjustment to 1150 but have lowered in this and above settings

My current VDDG settings - may raise them again for a new try at higher Infinity setting with some of the new tweaks

My PBO settings - juanyunis did manage above 5GHz with a low setting of the target overide so if It does anything is not sure - my previous settings was 500MHz

These settings give 4450MHz and above multicore performanze during CPU-z benchmark and single core above 4925MHz but temps above 75c

Will test this some more and try setting scalar to auto next

 

[juanyunis]

@jesdals thank you for the settings. I was able to tweak a little bit more to get more performance.

I'm currently running a stability test using Aida with a -30 on the curve using cpu core voltage normal with dynamic +0.050, cpu llc in auto, 250/125/125 for PBO, and +50Mhz for max boost.

I get 685/12980 in CPU-Z, Cinebench R20 11220, R23 28700 and Temps are below 80.

 

[Zach_01]

I believe there is a reason why TDC by default is lower than EDC. Take a look the descriptions of each PB limit

Package Power Tracking (“PPT”): The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and/or “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.

1. Default for Socket AM4 is at least 142W on motherboards rated for 105W TDP processors.
2. Default for Socket AM4 is at least 88W on motherboards rated for 65W TDP processors.

Thermal Design Current (“TDC”): The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in thermally-constrained scenarios.

1. Default for socket AM4 is at least 95A on motherboards rated for 105W TDP processors.
2. Default for socket AM4 is at least 60A on motherboards rated for 65W TDP processors.

Electrical Design Current (“EDC”): The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in a peak (“spike”) condition for a short period of time.

1. Default for socket AM4 is 140A on motherboards rated for 105W TDP processors.
2. Default for socket AM4 is 90A on motherboards rated for 65W TDP processors.

From my understanding by having TDC=EDC you remove the thermally constrained scenario protection. I dont really know if this thermally-constrained scenario would be reflected to CPU temp readings directly or the CPU "sees" and monitor something else internally that we as users dont have access to.
Also the EDC limit as it states its for peak and short timed boosting.

Just some food for thoughts

This article by GN is over a year old and was for Zen2 but I dont think anything has changed to Zen3 for the general principles of the chips.

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

Edit: typo

 

[juanyunis]

@Zach_01 thank you for your thoughts. They are very helpful. I will try playing with different bues for TDC and EDC. It does make sense.

 

[harm9963]

195 195 125 -30 from AMD OC menu, not PBO =195 watts / 73c for my system .
Added 10 min throttle test , watts stay the same , temp went to 77c .

 

[juanyunis]

Well, I think I have something nice, I know you guys won't like it @mtcn77 and @Zach_01 but haha this is the best I have ever achieved, I ran AIDA stability test for about 6 hours without any issue.

195 195 125 -30 from AMD OC menu, not PBO =195 watts / 73c for my system .
Added 10 min throttle test , watts stay the same , temp went to 77c .

wow, this is incredible, could you provide your PBO settings (scalar, max boost override, etc.), also voltage/offset, LLC

 

[harm9963]

Well, I think I have something nice, I know you guys won't like it @mtcn77 and @Zach_01 but haha this is the best I have ever achieved, I ran AIDA stability test for about 6 hours without any issue.

View attachment 183487View attachment 183488View attachment 183489


wow, this is incredible, could you provide your PBO settings (scalar, max boost override, etc.), also voltage/offset, LLC

Give it a go

http://imgur.com/a/hdElP39