AMD Curve Optimizer any guides / experience

[motleyguts]

I'm testing my curve optimizer undervolts with CoreCycler, and it turns out my new 5600x's Core 1 returns rounding errors even with no offset (negative offset set to 0). Does this mean this core is defective (and if so, is it worth it to activate my warranty just for this)? Or should I attempt to fix this by applying a positive offset?

@blu3dragon thank you for your info!

I'm just learning about this CoreCycler and the ins and outs of PBO, etc, but I have to ask, does this core pass with XMP (DOCP) turned off? Also, how much influence does LLC have? I've run prime95 for hours at various settings and found neg30 to all but my last core at neg27 was stable for hours upon hours (Edit: That includes monitoring HWiNFO64 and verifying clock and effective clock are matching). Now I run this CoreCycler, and they're failing within a minute or so, so now I feel like I don't know up from down, left from right. Back to testing!

 

[Nantes]

I'm just learning about this CoreCycler and the ins and outs of PBO, etc, but I have to ask, does this core pass with XMP (DOCP) turned off? Also, how much influence does LLC have? I've run prime95 for hours at various settings and found neg30 to all but my last core at neg27 was stable for hours upon hours (Edit: That includes monitoring HWiNFO64 and verifying clock and effective clock are matching). Now I run this CoreCycler, and they're failing within a minute or so, so now I feel like I don't know up from down, left from right. Back to testing!

That's interesting, because CoreCycler at default settings uses none other than Prime95. What's probably happening is that you were running Prime95 on all cores, which makes the cores reach a lower clock, whereas CoreCycler runs Prime95 on one core at a time, which reaches a high clock. Compare my 5600x running 4050 MHz all cores vs. 4650 MHz on one core on CoreCycler. The higher clock exposes the instability.

It's also worth noting that it's unlikely all your cores really are stable at -27 or below. Check out the offsets I reached: -9, +3, -25, -14, -19, 7. They are all over the place. You'd have to have bought a processor hand-picked by Jesus himself for all cores to be stable with the offsets you describe (or perhaps by Beelzebub himself, since I've read that the cores that are unstable with a not-very-negative offset is because they're already receiving their optimal voltage/are good cores).

An update to my issue: In the end I just grit my teeth and put +3 offset on Core 1, it seems to be perfectly stable now.

 

[motleyguts]

That's interesting, because CoreCycler at default settings uses none other than Prime95. What's probably happening is that you were running Prime95 on all cores, which makes the cores reach a lower clock, whereas CoreCycler runs Prime95 on one core at a time, which reaches a high clock. Compare my 5600x running 4050 MHz all cores vs. 4650 MHz on one core on CoreCycler. The higher clock exposes the instability.

It's also worth noting that it's unlikely all your cores really are stable at -27 or below. Check out the offsets I reached: -9, +3, -25, -14, -19, 7. They are all over the place. You'd have to have bought a processor hand-picked by Jesus himself for all cores to be stable with the offsets you describe (or perhaps by Beelzebub himself, since I've read that the cores that are unstable with a not-very-negative offset is because they're already receiving their optimal voltage/are good cores).

An update to my issue: In the end I just grit my teeth and put +3 offset on Core 1, it seems to be perfectly stable now.

It wasn't too terrible dialing in the offsets. I just watched twitch and youtube while corecycler did its thing. I ended up at -28, -20, -28, -29, -16, -18, -27, -14 with +200Mhz boost. I run some single core benches and older games and wanted to keep that boost. 7's the dreamer wanting to kiss 5050Mhz which is probably why it needs -14 to reach it, but it was a recurring theme during reboots, cursing 4, 5, and 7 for dashing my plans.

I'll throw an Arctic Freezer II 280mm aio on it today or tomorrow, and then I'll finally be able to see the RAM again, hiding underneath a Noctua NH-D14 120mm fan. I'll be able to pull 2 sticks and maybe do some better bench runs.

 

[blu3dragon]

It wasn't too terrible dialing in the offsets. I just watched twitch and youtube while corecycler did its thing. I ended up at -28, -20, -28, -29, -16, -18, -27, -14 with +200Mhz boost. I run some single core benches and older games and wanted to keep that boost. 7's the dreamer wanting to kiss 5050Mhz which is probably why it needs -14 to reach it, but it was a recurring theme during reboots, cursing 4, 5, and 7 for dashing my plans.

I'll throw an Arctic Freezer II 280mm aio on it today or tomorrow, and then I'll finally be able to see the RAM again, hiding underneath a Noctua NH-D14 120mm fan. I'll be able to pull 2 sticks and maybe do some better bench runs.

Those are still some very impressive offsets.

You might want to try leaving corecycler running when you are not using the pc as those background tasks could be lowering your boost clocks.

 

[craxton]

Most of you running stable in everything with a CO offset but crash at some point during idle. It's called DF-CSTATES turn it OFF in bios. Also, download zanstates to turn off c6 package all together. I had stable CO on my 5600x values of, -15, -20, -15, +8, -11, -15 and passed core cycler 12 hour runs with all fft sizes, passes prime 95, y-cruncher, etc but upon idle would crash. Someone else mentioned this to me so I take no credit.
I'll also add I'm running a 4x8 3200c14 kit of tforce dark pro bdie @4000mhz 1:1 mode. (Tuned)
if you go to the ZEN overclocking spread sheet, im on there. (notes to consider, ive lowered my IOD, CCD and SOC voltages ALOT since
posting this submission on that spreadsheet. but none the less, my RAM is stable, and now finally my CO values are too.
(you can also use an exe called TOOL hard to find, but it will tell you your silicon quality, (if it works right, some have issues as do i)
stating i have a 80 sil quality. will attach zenstates download and this "tool" as thats what its called, to my gdrive.

ZenStates_2.0.0_debug_20210120 - Google Drive

drive.google.com

Tool1007 - Google Drive

drive.google.com

(using tool go to DB query, then select amd v/f select number of cores etc, then hit get sil.... (also has a monitor feature but it wont stay open long
due to its poll rate being so high.)
as for zenstates, simply open it and go to power, then turn off package C6 state, (this will need done upon every reboot) unless you allow it to
run at start up.
you can also overclock you processor etc with this. (so be careful).

 

[squallypo]

Hi everyone I'm new in this topic and i wanna make sure I'm doing something right for my system since my aim is to keep a more cooler cpu in general without damaging any component of my PC but keeping the performance if possible.
Now I found this video and i wanna make sure this is actually accurate and it's safe to do
I must add that I'm using a ryzen 5600x with an aorus elite b550m Mobo

Also I know it might be just marketing but why does it says in the warning section that undervolting may damage or shorten the life or the processor or any other system components ? So there's a risk at doing this ?
someone in another forum suggested me to put these values in the pictures i added, not sure if they are right for me so i havent applied them yet.

thanks for your time !

 

[Mussels]

undervolting wont kill the system, but that -30 is pretty aggressive and might be unstable

 

[Makaveli]

undervolting wont kill the system, but that -30 is pretty aggressive and might be unstable

yup if you are going use CO at all cores and not per core I would start at maybe -10 on all cores then slowly test and work my way up.

 

[squallypo]

So doing undervolt won't damage the cpu or any other part at all ? Long-term speaking.
And I tested what the person did on the video and it's impressive how the temps lowers very significantly while keeping the cores working as normally would, so at least under some kind of load the system seemed stable at -30 on all cores in curve optimizer, didn't got to test on idle yet so as you guys said that might be the trick to test, also say if I jusr wanna do this unfervolting through curve optimizer and keep my clocks as default without adding any more Mhz , is fine no ? Theres not some kind of tradeoff I need to perform in order to get the clocks I aim to get ?

Also why i notice that some fiddle around with the values in ppt, tdc and EDC? Is it for ocing factor ? Is it find to keep it disabled while doing curve optimizer undervolting ? Or even just leave it at default and just doing the co undervolting?

 

[Makaveli]

I would download and run corecycler to test for stability.

CoreCycler - tool for testing single core stability...

www.overclock.net

 

[squallypo]

I would download and run corecycler to test for stability.

CoreCycler - tool for testing single core stability...

www.overclock.net

Doing this as we speak and i noticed how it's running iterations but the number of iterations are set to 1000 by default, isn't that gonna take a very long time ? I did change the runtimepercore to 60 seconds

 

[jesdals]

Hi everyone I'm new in this topic and i wanna make sure I'm doing something right for my system since my aim is to keep a more cooler cpu in general without damaging any component of my PC but keeping the performance if possible.
Now I found this video and i wanna make sure this is actually accurate and it's safe to do
I must add that I'm using a ryzen 5600x with an aorus elite b550m Mobo

View attachment 209419View attachment 209420

Also I know it might be just marketing but why does it says in the warning section that undervolting may damage or shorten the life or the processor or any other system components ? So there's a risk at doing this ?
someone in another forum suggested me to put these values in the pictures i added, not sure if they are right for me so i havent applied them yet.

thanks for your time !

You can make pics by pressing F12 in that bios

 

[Makaveli]

Doing this as we speak and i noticed how it's running iterations but the number of iterations are set to 1000 by default, isn't that gonna take a very long time ? I did change the runtimepercore to 60 seconds

You don't have to wait that long if a core is going to fail it will do it fairly quickly.

 

[squallypo]

You don't have to wait that long if a core is going to fail it will do it fairly quickly.

alright im back with results:

cinebench30min test hwinfo64 log
also added the log file of the cinebench test

let me know if everything is working as intended.
what i did on bios was just disable pbo limits (not on manual nor motherboard )
and on curve optimizer set all cores at -30

 

[Zach_01]

Curve Optimizer is a little more complicated than just an undervolt mechanism. While negative offsets are indeed undervolt the CPU will increase speed and its PBO limits (PPT/EDC/TDC) way beyond its stock limits. This increases temp dramatically.
If anyone wants to keep it as safe as possible and wants just undervolt must set the desired limits of PBO or have it completely at stock.

Default limits of 5600X are:
PPT: 76W
TDC: 60A
EDC: 90A

I would consider the following as safe if cooling device can keep it at reasonable levels (80~85C max, 100% load).

PPT: 90~95W
TDC: 70~75A
EDC: 100~110A

Max operating temp for 5000 is 90C.

 

[squallypo]

Curve Optimizer is a little more complicated than just an undervolt mechanism. While negative offsets are indeed undervolt the CPU will increase speed and its PBO limits (PPT/EDC/TDC) way beyond its stock limits. This increases temp dramatically.
If anyone wants to keep it as safe as possible and wants just undervolt must set the desired limits of PBO or have it completely at stock.

Default limits of 5600X are:
PPT: 76W
TDC: 60A
EDC: 90A

I would consider the following as safe if cooling device can keep it at reasonable levels (80~85C max, 100% load).

PPT: 90~95W
TDC: 70~75A
EDC: 100~110A

Max operating temp for 5000 is 90C.

Isn't disabling pbo limits like having it at stock ? Cause when I ran cinebench i had open Ryzen master and the PPT , EDC and tdc were at the values that you mentioned as stock

 

[freeagent]

Core cycler fails for me with my current settings.. I get no hardware errors, no junk files, no bluscreens, passes every single thing, except that program. I can fold, wcg, game, internet, desktop just fine.. no errors. Not a random number generator at all.. but is a kickass calculator.. all core set to -30 +200 with custom ppt tdc edc and she rips. Falls on its face with core cycler, so I suppose it is a random number generator

The 5900X’s that are faster at superpi-32m than mine at hwbot are running at 5500MHz+

For now

It was difficult to tune the 5900X compared to the 5600X. That’s when I realized we aren’t controlling any limits, just manipulating them.. at least on the big CPU’s. Even on the small ones too I suppose, but there is way less going on and you can use pbo to brute force it to top clocks all the time, the big ones you have to cheat a little.

 

[Zach_01]

Isn't disabling pbo limits like having it at stock ? Cause when I ran cinebench i had open Ryzen master and the PPT , EDC and tdc were at the values that you mentioned as stock

Indeed it is...

 

[Mussels]

Curve Optimizer is a little more complicated than just an undervolt mechanism. While negative offsets are indeed undervolt the CPU will increase speed and its PBO limits (PPT/EDC/TDC) way beyond its stock limits. This increases temp dramatically.
If anyone wants to keep it as safe as possible and wants just undervolt must set the desired limits of PBO or have it completely at stock.

Default limits of 5600X are:
PPT: 76W
TDC: 60A
EDC: 90A

I would consider the following as safe if cooling device can keep it at reasonable levels (80~85C max, 100% load).

PPT: 90~95W
TDC: 70~75A
EDC: 100~110A

Max operating temp for 5000 is 90C.

Throwing that in the PBO link in my sig

 

[freeagent]

I would consider the following as safe if cooling device can keep it at reasonable levels (80~85C max, 100% load).

PPT: 90~95W
TDC: 70~75A
EDC: 100~110A

On my 5600X:

PPT: 200W
TDC: 140
EDC: 180

+200MHz All core set to -30 = Perfectly safe, runs cool, rips hard.

 

[squallypo]

On my 5600X:

PPT: 200W
TDC: 140
EDC: 180

+200MHz All core set to -30 = Perfectly safe, runs cool, rips hard.

but im guessing you have AIO cooling system no?

 

[Cowboystrekk]

Been running -30 allcore and +50MHz pbo for 3 months now, still stable and a bit faster and cooler than stock Currently running PPT 45W since I run monerominer with nicehash.

 

[Mussels]

I'm gunna guess 5600x get all the cherry picked cores, you lucky undervolting bastards

 

[freeagent]

but im guessing you have AIO cooling system no?

No sir, just an air cooler.

 

[squallypo]

No sir, just an air cooler.

im really tempted to try your config.
which cooler you got ?