Trying to understand Ryzen 3000 series boost speed variations

[tabascosauz]

Efffective clock is not for single core boost/clock. It contains the sleeping states of the cores, so the reported clock has such a low value. Think of it as an average between active and sleeping state.

FEATURE - Effective clock vs instant (discrete) clock

It has become a common practice for several years to report instant (discrete) clock values for CPUs. This method is based on knowledge of the actual bus clock (BCLK) and sampling of core ratios at specific time points. The resulting clock is then a simple result of ratio * BCLK. Such approach...

www.hwinfo.com

Another topic...
Have some premature (not alot testing) findings while playing around with EDC from UEFI that may interest you.
Remember a few posts back I was giving a (negative) offset to vcore so that overall power draw (PPT) and temp of CPU would drop and thus having (in theory) more headroom for clocks. That was not the case for my CPU as this was decreasing eventually clocks and performance, due to poor silicon quality that cannot sustain max clocks with direct under voltage. This offset is applied across the entire range of the clock (single and all core). Tho... the power effectiveness of the CPU was increased. I had decreased voltage by ~2% and lost clock/performance by ~1%.
This could work better with a higher silicon quality CPU like the 3900X.

So I decided to take another route by decreasing the CPU's current (EDC/Amps) and not voltage directly to control max power draw.
My default max values goes like this:
PPT: 88 Watt
TDC: 60 Amps
EDC: 90 Amps

When running R20 all core, those limits was never reached except PPT, so:
PPT: 87.7 Watt
TDC: ~49 Amps
EDC: ~79 Amps
Max temp: 63C

Into UEFI I set the EDC limit to 76A and run R20:
PPT: 87.8 Watt
TDC: ~49 Amps
EDC: 76 Amps
Max temp: 63C

Into UEFI I set the EDC limit to 73A and run R20:
PPT: 88 Watt
TDC: ~49 Amps
EDC: 73 Amps
Max temp: 63C

This indicates that in order to still having the same PPT and temp, while power current (EDC) has dropped... something else must have gotten up! The only thing that could have gotten up is the clock...
Even though I didnt see a clock increase visually, I did see some increase in R20 scores for the few runs I did. The "drawback" in the EDC adjustment is that only affects max power draw, so only max all core clocks/boost... unlike the voltage offset that affects the entire range of clocks.
I will come back with this, but feel free to do some testing yourself (voltage offset or EDC). I will even test with a combination of the two (negative voltage offset + limited EDC).

I had the most success with EDC 80 in combination with the -0.075V offset. Without combining the two, reducing EDC at stock voltage just makes for a poorly performing version of more aggressive PBO settings. Performance is comparable with the offset, at EDC 80 and 90, but the reduction makes for more consistent scoring. I'm pretty happy with how it's doing right now.

@lorry sounds like your P95 is borked somehow or an old version. Large is able to occupy close to 30GB out of 32GB of RAM on mine. Maybe use Custom and copy my Large settings? We both have 32GB kits.

 

[lorry]

I had the most success with EDC 80 in combination with the -0.075V offset. Without combining the two, reducing EDC at stock voltage just makes for a poorly performing version of more aggressive PBO settings. Performance is comparable with the offset, at EDC 80 and 90, but the reduction makes for more consistent scoring. I'm pretty happy with how it's doing right now.

@lorry sounds like your P95 is borked somehow or an old version. Large is able to occupy close to 30GB out of 32GB of RAM on mine. Maybe use Custom and copy my Large settings? We both have 32GB kits.

View attachment 141134

Eh? why do you think my P95 is borked? When I talked about Maximum not running properly that was talking about Intel Burn, not P95
is this the latest version of P95 (just to be safe)



--------------------------------------------------------------------------------------------------------------------

Don't know if this is of interest to any of you?

Ryzen Above: Best Memory Settings for AMD's 3000 CPUs, Tested

Is Ryzen 3000 optimized for DDR4-3200? What if we want more? We examine everything from frequency to rank count to nail down optimal settings.

www.tomshardware.com

 

[tabascosauz]

Eh? why do you think my P95 is borked? When I talked about Maximum not running properly that was talking about Intel Burn, not P95
is this the latest version of P95 (just to be safe)

View attachment 141135

You mentioned earlier that Large doesn't fill your RAM.

That does look to be interesting but I feel that I would need to know how far to go in undervolting and/or EDC - for instance do you reduce or increase EDC?

I just ran CB20 and here are my max limits, you can see that I do reach 100% on EDC and close to the limits on the other two

View attachment 141133

Definitive guide to configuring the Ryzen 3900X

It has taken me this long to write this guide because I don't get free stuff sent to me so I had to wait until I was able to get hands on experience with this CPU and its two chiplet design. So I am sorry if you have been having a bad experience trying to make sense of what "influencers" who do...

community.amd.com

We're talking about reducing EDC. EDC is the maximum short-term current that can be delivered. Stock for 65W parts is 90A, stock for 105W is 140A. For all the talk about reducing EDC slightly, I think it's all been for single-chiplet CPUs so far. No harm in trying.

I'm not sure how I feel about that nec_v20 guy's "guide". There's recommendations of disabling SMT and running a 3900X as a 12c/12t, arguing that with fixed freq results they can tweak to achieve lower stable Vcore than with boost (no shit??), and running fixed 1.3v Vcore while not running fixed clocks. The references to FIT are referencing the Stilt thread from before where he claimed 1.325v max voltage when running boost, which seems about right

 

[lorry]

View attachment 141136

You mentioned earlier that Large doesn't fill your RAM.



We're talking about reducing EDC. EDC is the maximum short-term current that can be delivered. Stock for 65W parts is 90A, stock for 105W is 140A. For all the talk about reducing EDC slightly, I think it's all been for single-chiplet CPUs so far. No harm in trying.

I'm not sure how I feel about that nec_v20 guy's "guide". There's recommendations of disabling SMT and running a 3900X as a 12c/12t, arguing that with fixed freq results they can tweak to achieve lower stable Vcore than with boost (no shit??), and running fixed 1.3v Vcore while not running fixed clocks. The references to FIT are referencing the Stilt thread from before where he claimed 1.325v max voltage when running boost, which seems about right

ah! hmm, okay I'll redownload it and see if it was borked

View attachment 141136

You mentioned earlier that Large doesn't fill your RAM.



We're talking about reducing EDC. EDC is the maximum short-term current that can be delivered. Stock for 65W parts is 90A, stock for 105W is 140A. For all the talk about reducing EDC slightly, I think it's all been for single-chiplet CPUs so far. No harm in trying.

I'm not sure how I feel about that nec_v20 guy's "guide". There's recommendations of disabling SMT and running a 3900X as a 12c/12t, arguing that with fixed freq results they can tweak to achieve lower stable Vcore than with boost (no shit??), and running fixed 1.3v Vcore while not running fixed clocks. The references to FIT are referencing the Stilt thread from before where he claimed 1.325v max voltage when running boost, which seems about right

Okay, latest version downloaded from techowerup, 29.8 build 6, same thing happens, screenshots follow

 

[tabascosauz]

View attachment 141138

View attachment 141139

View attachment 141140

View attachment 141141

I mean, the version is exactly the same, but I've always gone to the actual Prime95 site for download.

P95 does detect and optimize for your CPU so maybe it's a dual chiplet thing? Which doesn't seem to make any sense, since the FFTs sit in RAM for Large. I notice that in the greyed out torture test settings (you can select each different test and see the values change), the minimum FFT size used for your Large are significantly longer in length. See mine:

What happens if you use my settings to run a Custom test (which allows you to change the values)? RAM load in HWInfo should be way up near 100% in Large, that's what makes it a memory test.

 

[lorry]

I mean, the version is exactly the same, but I've always gone to the actual Prime95 site for download.

P95 does detect and optimize for your CPU so maybe it's a dual chiplet thing? Which doesn't seem to make any sense, since the FFTs sit in RAM for Large. I notice that in the greyed out torture test settings (you can select each different test and see the values change), the minimum FFT size used for your Large are significantly longer in length. See mine:

View attachment 141142

What happens if you use my settings to run a Custom test (which allows you to change the values)? RAM load in HWInfo should be way up near 100% in Large, that's what makes it a memory test.

You mean From https://www.mersenne.org/download/

That Was where I got it from the first time. TBH I don't feel that changing the settings is going to change anything really. As you say it's likely the dual chiplet that makes the difference, and/or the virtual memory maybe?
Anyway it completed the Intel Burn tests on standard, high, very high (see my comment above where I pasted the results) and I will have another try later on with the maximum.
I stopped it before it had even done the first run as I thought it might be faulty, but on looking online it seems that each run takes in excess of 15 minutes and I stopped it after about 11, doh!
Oddly though whilst even on Max (when I saw the available ram drop to 2MB !) the temp was 12c lower at 71 compared to a max of 83 on p95 (maybe the Intel Burn can't handle this particular AMD setup of CPU?) and the ram remained in the 30s (33.5 - 34.5c).

I did try a search to see if anyone else had reported this but maybe I used the wrong serach words?

-------------------------------------------------------------------------------------------------------

Oh, and @Fry178 got my refund from that seller on eBay, plus, so far anyway, he hasn't sent me any return details, so I could well end up keeping it!
Have ordered a specifically detailed backlit model (the site on Amazon details 3 models, backlit, non backlit and a newer version that shows usage over 7 days on the display all at one)

Hmm, watch this space

Report: Nvidia's next-gen Ampere cards 50% faster than Turing with half the power draw

As reported by the Taipei Times (via Tom's Hardware), Yuanta Securities Investment Consulting Co said in a client note that Nvidia would launch its Ampere-based GPU in...

www.techspot.com

------------------------------------------------------------------------------------------------------------------------------------

----------------------------
IntelBurnTest v2.54
Created by AgentGOD
----------------------------

Processor: AMD Ryzen 9 3900X 12-Core Processor
Clock Speed: 2.92 GHz
Active Physical Cores: 24
Total System Memory: 32713 MB

Stress Level: Maximum (25070 MB)
Testing started on 03/01/2020 8:37:11 PM
Time (s) Speed (GFlops) Result
[20:53:51] 861.881 141.4827 3.195339e-002
[21:12:26] 977.931 124.6931 3.195339e-002
[21:28:15] 820.179 148.6763 3.195339e-002
[21:43:59] 822.284 148.2958 3.195339e-002
[21:59:41] 819.897 148.7275 3.195339e-002
Testing ended on 03/01/2020 9:59:49 PM
Test Result: Stopped by user.


I had to stop it as it would have taken another 90 minutes to complete but the results look good to me all the same
-------------------------------------------------------------------------------------------------------------------------------------

Temps reached a max of 73c for cpu, with it dropping back down to 52c after each run had finished.
Ram was between 34.5 - 37c

Interestingly the ram usage for this Was 100%, in that there were times when the available ram was 0

 

[Zach_01]

That does look to be interesting but I feel that I would need to know how far to go in undervolting and/or EDC - for instance do you reduce or increase EDC?

I just ran CB20 and here are my max limits, you can see that I do reach 100% on EDC and close to the limits on the other two

View attachment 141133

I pointed out what I did but I'll write again.

These are the limits of the 3600

My default max values goes like this:
PPT: 88 Watt
TDC: 60 Amps
EDC: 90 Amps

----------------------------------------
These are the actual numbers the CPU was hitting during R20 without touching anything

When running R20 all core, those limits was never reached except PPT, so:
PPT: 87.7 Watt (99.7%)
TDC: ~49 Amps (82.5%)
EDC: ~79 Amps (88%)
Max temp: 63C

----------------------------------------

I wanted to reduce the power current (EDC), so that PPT would have headroom and see if that would raise clocks. My CPU was previous hitting 79A so I had to set it below that value. I tried it first with 76A and then 73A

Into UEFI I set the EDC limit to 76A and run R20:
PPT: 87.8 Watt (99.7%)
TDC: ~49 Amps (82.5%)
EDC: 76 Amps
Max temp: 63C

Into UEFI I set the EDC limit to 73A and run R20:
PPT: 88 Watt
TDC: ~49 Amps
EDC: 73 Amps
Max temp: 63C

It seems that it worked.
Without touching anything when the CPU run R20 the clock was 3975~4000MHz.
After reducing the limit to 73 the CPU run R20 at 3975~4050MHz and the score went up about 50~60points.

 

[lorry]

My PPT seems to be the least used, at least in P95 & IB, can't remember offhand for R20. EDC & TDC often run at 100% and 98% respectively, with PPT I think about 96%

 

[mtcn77]

I had the most success with EDC 80 in combination with the -0.075V offset. Without combining the two, reducing EDC at stock voltage just makes for a poorly performing version of more aggressive PBO settings. Performance is comparable with the offset, at EDC 80 and 90, but the reduction makes for more consistent scoring. I'm pretty happy with how it's doing right now.

@lorry sounds like your P95 is borked somehow or an old version. Large is able to occupy close to 30GB out of 32GB of RAM on mine. Maybe use Custom and copy my Large settings? We both have 32GB kits.

View attachment 141134

If you limit EDC, you will likely curtail LLC need and thereby improve FIT-temperature response of your CPU. Driving faster pwm, or double the phase count motherboards should not be beneficial in this regard where the beef is between your cpu cooler and temperature throttle curve.
EDC+voltage offset shouldn't be possible, unless Stilt hasn't raised enough flags and motherboard OEM's are still applying higher than default voltage settings in order to one-up their ever important rivals.

Also, at a decently low EDC throttling point, you should be able to raise TDC up to that of EDC safely for continuous performance - PBO without turbo taper.

 

[Zach_01]

My PPT seems to be the least used, at least in P95 & IB, can't remember offhand for R20. EDC & TDC often run at 100% and 98% respectively, with PPT I think about 96%

You will see different numbers at different benchmarks, as all of them are loading the CPU slightly different. Its normal.
I would say to try to raise PPT (from 142W), and reduce EDC (from 140A).

Like this...
PPT: 145 (+3)
TDC: Default (0)
EDC: 135 (-5)

First run a CB R20 or any other benchmark to know you starting point.

 

[tabascosauz]

If you limit EDC, you will likely curtail LLC need and thereby improve FIT-temperature response of your CPU. Driving faster pwm, or double the phase count motherboards should not be beneficial in this regard where the beef is between your cpu cooler and temperature throttle curve.
EDC+voltage offset shouldn't be possible, unless Stilt hasn't raised enough flags and motherboard OEM's are still applying higher than default voltage settings in order to one-up their ever important rivals.

Also, at a decently low EDC throttling point, you should be able to raise TDC up to that of EDC safely for continuous performance - PBO without turbo taper.

See, that's what I thought, too. But I'm just not seeing the performance increase, not even slightly. Tried seeing if changing other PBO parameters helps. TDC tops out at 52A; no sense in trying to go past stock. That's why I set the offset too, as since I get the same performance, I may as well go and reduce the heat and noise.

Interestingly enough, I occasionally boost past 4400MHz to the 44.3x multiplier on my best two cores when I don't have the offset on, but it doesn't matter, as CB scores don't reflect that. There is a little bit of a ST benefit going to an EDC of 80, ST is up to 507 now, and consistently so, since CB is smart enough to have the load sit on the best two cores for the entire test (~43.8x multiplier on effective clocks).

I'm starting to think that connecting EDC and multi-core score is a bit misguided. And unless the original graph from the EDC-related post with the same exact setup as mine shows consistent averages over at least 20 runs over 20 boots, it's really hard to disprove that a handful of points gain in CB R15 isn't due to regular variances. Hell, I sometimes gain 10-20pts in R20 from tweaking these settings, but instantly lose a guaranteed 60-80pts in R20 from forgetting to close Dropbox. Single-thread gains, sure.

And as to factory overvolting, they most definitely still do that.

 

[lorry]

I pointed out what I did but I'll write again.

These are the limits of the 3600


----------------------------------------
These are the actual numbers the CPU was hitting during R20 without touching anything


----------------------------------------

I wanted to reduce the power current (EDC), so that PPT would have headroom and see if that would raise clocks. My CPU was previous hitting 79A so I had to set it below that value. I tried it first with 76A and then 73A


It seems that it worked.
Without touching anything when the CPU run R20 the clock was 3975~4000MHz.
After reducing the limit to 73 the CPU run R20 at 3975~4050MHz and the score went up about 50~60points.

View attachment 141191

well that seems to have fecked just about everything up ????
Look at my memory clock speeds etc?

 

[tabascosauz]

If you limit EDC, you will likely curtail LLC need and thereby improve FIT-temperature response of your CPU. Driving faster pwm, or double the phase count motherboards should not be beneficial in this regard where the beef is between your cpu cooler and temperature throttle curve.
EDC+voltage offset shouldn't be possible, unless Stilt hasn't raised enough flags and motherboard OEM's are still applying higher than default voltage settings in order to one-up their ever important rivals.

Also, at a decently low EDC throttling point, you should be able to raise TDC up to that of EDC safely for continuous performance - PBO without turbo taper.

well that seems to have fecked just about everything up ????
Look at my memory clock speeds etc?

View attachment 141220

Did you only change EDC? Something surely hit the fan, because you're back at JEDEC on RAM, RAM is uncoupled, Vcore goes up to 1.9V which is clearly a bug, and all your TDC/EDC/PPT values appear to be exactly stock.

Clear your CMOS and restore your previous settings, if you've saved them as a BIOS profile?

-----

@mtcn77 see these results at 83A EDC and no offset. I put LLC back on Auto for more droop at load (I had a nasty habit of leaving it on Turbo, a remnant of when I was testing fixed freq).



That's the best I can do. That's with no HWInfo monitoring, internet adapters disabled, no unnecessary background tasks. I had previous beat my former best by 1pt at 4959 on my first run at 83A, then failed to hit within 20pts of that best score until I changed to Auto LLC. Single core always tops out at 507, which is pretty much the best it can do, and it's always running on the best cores. The problem with the single core benchmark is that it favours people who have as minimal a Windows installation as physically possible, as it's clearly audible when the CPU sometimes does other tasks with inactive cores in the middle of the benchmark; the test takes so long that there's definitely going to be background tasks going on that fuck with the score.

To the ST scores, case in point: CPU-Z in safemode, because CPU-Z is pretty much the only bench that runs in Safemode. Reaching 525 single thread is literally impossible in normal Windows, regardless of how much extraneous tasks are disabled. Clearly something, possibly the fact that Windows scheduler in normal Windows always locks the bench to Core 0 to run the task, is causing the drastic difference. But, it's not like I can investigate what core is running CPU-Z in Safemode; HWInfo doesn't work in Safemode.



I think I'm going to call it quits here. 83A empirically works better than 80A, as is evident by the best score, but 85A starts losing score again. It still makes a bit of a racket and unncessary heat at these settings due to no offset, so I don't know how long I'll keep putting up with this all in the name of gaining 30pts in Cinebench. Tired of having all the control taken out of my hands by how unpredictable Ryzen is. Just gonna enjoy what I have.

Hats off to you for bringing the EDC tweaking to our attention. It clearly works - the tweaks are different for everybody's chips. I'm just not sure the gains are going to be worth it. From what I see, the SFF crowd loves EDC tweaking, but I've been part of the SFF crowd since it first started to gain traction, and the problem with Ryzen with the small coolers were stuck with is not peak consumption, but how it sounds and behaves the other 95% of the time. Not that EDC tweaking alone even comes close to the thermal and acoustic benefits of undervolting, but hey, at least now I know there are people out there who are getting something more substantial out of it.

 

[Zach_01]

well that seems to have fecked just about everything up ????
Look at my memory clock speeds etc?

Did you only change EDC? Something surely hit the fan, because you're back at JEDEC on RAM, RAM is uncoupled, Vcore goes up to 1.9V which is clearly a bug, and all your TDC/EDC/PPT values appear to be exactly stock.

Clear your CMOS and restore your previous settings, if you've saved them as a BIOS profile?

Did you use the numbers I suggested?
And where did you make the changes? AMD CBS or AMD Overclocking?
Pretty much it cleared/reset it self at boot

 

[tabascosauz]

Well, looks like I jumped the gun a little bit. It had a bit left in it, probably because it was long enough after boot that background tasks had long since settled. I have a feeling 4985 is givin' her all she's got, capt'n. Clocks are starting at 41.5x and ending at 41.0x.

Did you use the numbers I suggested?
And where did you make the changes? AMD CBS or AMD Overclocking?
Pretty much it cleared/reset it self at boot

I'm not sure if you've noticed this on your X570 board tweaking these PBO settings, but the manual PPT/EDC/TDC settings in my PBO menu are ridiculously difficult to save. Like, they have a mind of their own. I'll save them very clearly at 88/60/83, and I'll reboot back into BIOS to check, and they'll be at 105/60/83, even though I haven't touched that PPT value for a week. Or I'll change 88/60/83 to 88/60/85, and reboot to find that the 85 didn't stick. A couple of times, I didn't even leave the BIOS; I went to the Save and Quit tab and right back to the PBO page, only to find that the values had changed on their own.

Man, Gigabyte sucks at making BIOSes.

 

[lorry]

Did you use the numbers I suggested?
And where did you make the changes? AMD CBS or AMD Overclocking?
Pretty much it cleared/reset it self at boot

I did them in PBO, was that correct?
yes I thought i had used your number, but didnt screenshot them

PPT 145
TDC 95
EDC 135 Oh feck! By the look of my writing when I took down your figures, I may have set this one at 155 !!

also any idea what the motherboard setting is?

 

[Zach_01]

I'm not sure if you've noticed this on your X570 board tweaking these PBO settings, but the manual PPT/EDC/TDC settings in my PBO menu are ridiculously difficult to save. Like, they have a mind of their own. I'll save them very clearly at 88/60/83, and I'll reboot back into BIOS to check, and they'll be at 105/60/83, even though I haven't touched that PPT value for a week. Or I'll change 88/60/83 to 88/60/85, and reboot to find that the 85 didn't stick. A couple of times, I didn't even leave the BIOS; I went to the Save and Quit tab and right back to the PBO page, only to find that the values had changed on their own.

Man, Gigabyte sucks at making BIOSes.

Not such issue on my board for PBO settings...

I did them in PBO, was that correct?
yes I thought i had used your number, but didnt screenshot them

PPT 145
TDC 95
EDC 135 Oh feck! By the look of my writing when I took down your figures, I may have set this one at 155 !!

also any idea what the motherboard setting is?

No idea, I did try it once but it make the system to run/respond like a snail...
This page is in AMD Overclocking...

Try this:
Set that "Precision Boost Overdrive" from "Advanced" to "Auto" (actually set all those settings to Auto). You may loose that +200MHz theoritical boost but bare with me...
Go to XFR Enhancement as below and set the values







Leave TDC 0

 

[lorry]

Not such issue on my board for PBO settings...


No idea, I did try it once but it make the system to run/respond like a snail...
This page is in AMD Overclocking...

Try this:
Set that "Precision Boost Overdrive" from "Advanced" to "Auto" (actually set all those settings to Auto). You may loose that +200MHz theoritical boost but bare with me...
Go to XFR Enhancement as below and set the values

View attachment 141241

View attachment 141242

View attachment 141243

Leave TDC 0
View attachment 141244

why is TDC left at 0 though?
And why reset PBO to auto again?
What is being tried and why please?

Also what is that MB option in my BIOS? Any ideas?

 

[tabascosauz]

why is TDC left at 0 though?
And why reset PBO to auto again?
What is being tried and why please?

Also what is that MB option in my BIOS? Any ideas?

View attachment 141246

He's not talking about the Peripherals>AMD Overclocking menu. There's a separate Precision Boost submenu under Peripherals>AMD CBS, which is the option located directly above AMD Overclocking.

From what I understand, people seem to think that AMD CBS is more geared towards Zen+ use, whereas AMD Overclocking is supposedly for Zen 2. Never hurts to try it out, though. Worst case, just clear CMOS again.

 

[lorry]

He's not talking about the Peripherals>AMD Overclocking menu. There's a separate Precision Boost submenu under Peripherals>AMD CBS, which is the option located directly above AMD Overclocking.

From what I understand, people seem to think that AMD CBS is more geared towards Zen+ use, whereas AMD Overclocking is supposedly for Zen 2. Never hurts to try it out, though. Worst case, just clear CMOS again.

yes I did know about that but thought that it shouldn't make much of a difference.
Why leave TDC at 0 though? what is the reason for that?

And any thoughts on this Motherboard setting under AMD CBS section?

--------------------------------------------------------------------------------------------
Interesting post here I think?

https://www.reddit.com/r/overclocking/comments/ejd5c9
https://www.reddit.com/r/Amd/comments/b8a5ft

 

[Zach_01]

why is TDC left at 0 though?
And why reset PBO to auto again?
What is being tried and why please?

Also what is that MB option in my BIOS? Any ideas?

View attachment 141246

As I said, I dont know what Motherbord option is in PBO. I did tried it once and made my system run like slug...
TDC (Thermal Design Current) is for thermaly constrained scenarios. Like thermal throttling. So if we dont have extreme thermal issues there is no point to alter it.
0 = default limit

-------------------------------------------
AMD CBS
AMD Overclocking

2 sections that do the same things. Preferably choose AMD CBS for oc and tweak in general, because there is a possibility that if the system needs a clear CMOS and there is no dual BIOS... AMD Overclocking may not clear some settings (heard by BZ). In this case reflash is the only way to get around that.
And what are we trying to do here is tweak and maybe optimize auto PBO so FIT may increase all core boost clocks. The less things interfere (like +200MHz or any other) the better.
ZEN2 is so f*cking annoyingly complex and advanced (pardon me please, but true) that you have to take things one at the time.

So thats why I suggest to have every thing in AMD Overclocking PBO on Auto and do the PPT/EDC setting/tweaking in AMD CBS PBO section.

 

[lorry]

As I said, I dont know what Motherbord option is in PBO. I did tried it once and made my system run like slug...
TDC (Thermal Design Current) is for thermaly constrained scenarios. Like thermal throttling. So if we dont have extreme thermal issues there is no point to alter it.
0 = default limit

-------------------------------------------
AMD CBS
AMD Overclocking

2 sections that do the same things. Preferably choose AMD CBS for oc and tweak in general, because there is a possibility that if the system needs a clear CMOS and there is no dual BIOS... AMD Overclocking may not clear some settings (heard by BZ). In this case reflash is the only way to get around that.
And what are we trying to do here is tweak and maybe optimize auto PBO so FIT may increase all core boost clocks. The less things interfere (like +200MHz or any other) the better.
ZEN2 is so f*cking annoyingly complex and advanced (pardon me please, but true) that you have to take things one at the time.

So thats why I suggest to have every thing in AMD Overclocking PBO on Auto and do the PPT/EDC setting/tweaking in AMD CBS PBO section.

Thanks and, no probs, i swear as much as anyone else lol (Especially at other drivers [did the ambulance car service for 21 years driving 1,200 miles every week so you can imagine some of my situations]).

One point, I Do have dual BIOS on this MB - 2 x 128Mbit flash, support for DualBIOS (one of the reasons i choose this one).

 

[Zach_01]

Hats off to you for bringing the EDC tweaking to our attention. It clearly works - the tweaks are different for everybody's chips. I'm just not sure the gains are going to be worth it. From what I see, the SFF crowd loves EDC tweaking, but I've been part of the SFF crowd since it first started to gain traction, and the problem with Ryzen with the small coolers were stuck with is not peak consumption, but how it sounds and behaves the other 95% of the time. Not that EDC tweaking alone even comes close to the thermal and acoustic benefits of undervolting, but hey, at least now I know there are people out there who are getting something more substantial out of it.

I'm not sure to whom you are refering to but the first man that brought EDC tweaking here is @Fry178 a few pages back in this thread!
Even though its not ground breaking tweaking I like it. And yes things are moving towards auto everything and the OC is dying, but thats Ok with me...

 

[lorry]

@Zach_01 when you said "set PBO from advanced to auto (actually set all those settings to auto)"

exactly which settings should be reset to auto?

 

[Zach_01]

@Zach_01 when you said "set PBO from advanced to auto (actually set all those settings to auto)"

exactly which settings should be reset to auto?

This...
View attachment 141240

And do it from here...
View attachment 141241

View attachment 141242

View attachment 141243

View attachment 141244