Ryzen 5900x undervolt

[Wadzinsky]

I have a Ryzen 5900x paired with a Asus B550M board and I am trying to do an undervolt to keep temps and power draw down.  They aren't dangerous, but I am energy conscious and just want to get the most out of my CPU without the exponential power draw for diminishing returns in performance at stock voltages.

 

My board has a profile in the overclocking utility for Eco mode, but it only has an option for 65 watts.  It was rather unstable at the voltages it provided but I increased the top number from 87 to 90 and it became stable, however, at nearly half the power it was designed to run at, the CPU is a little sluggish and would like to increase it from 65 to 95 watts.  At stock settings, the CPU would draw upwards of 150-160 watts under load and run at 80-85c.  Currently at the 65w setting it maxes out around 70c

 

Honestly, I don't know much what the numbers I'm messing with represent except the top number is the power limit to the CPU socket.  Can anyone tell me what numbers I'd want to place in the 3 boxes to achieve my goal of 95watts and be stable?  (or close to stable with minor adjustments)?  Or perhaps tell me how adjusting the other numbers works and what it's doing?  

 

I'm kind of a noob, so thanks in advance.

 

The pics are my current stable adjustment for the 65w profile

[freeagent]

Get the newest AGESA, as it will take your max voltage down from 1.5 to 1.425. Then run it at full stock, don't touch PBO and CO. Or just get a better cooler. I ran mine at 235/160/190 with it boosting to the full 5150 on multiple cores, and no problems with temps.

[GuiltySpark_]

You could do that, but if you wanted to get a bit of a primer on tweaking, this is a nice place to start. 

 

https://www.reddit.com/r/overclocking/comments/qikcbd/zen_3_pbo_curve_optimiser_tweaking_guide/

 

[freeagent]

The only reason why I mentioned a new cooler is because he has his limits super low, well under stock and is still showing 70c.

[Wadzinsky]

16 minutes ago, freeagent said:

Get the newest AGESA, as it will take your max voltage down from 1.5 to 1.425. Then run it at full stock, don't touch PBO and CO. Or just get a better cooler. I ran mine at 235/160/190 with it boosting to the full 5150 on multiple cores, and no problems with temps.

I updated to the latest bios yesterday which should have put me on the latest AGESA, but my voltage still reports 1.5.  I'll try manually adjusting it in bios and see if its stable.

I stumbled across a redit that showed the default PBO settings for the 95w eco profile (which my board doesn't have)  I just set them to

PPT = 120, TDC = 90, EDC = 125a, then did a power curve adjustment on all cores of -10.  My Performance Test score went from 39890 at stock settings to 39470.

 

My power draw went for 162w to 126w and temp from 85c to 74c.  That's close enough to stock performance and a considerable decrease in power draw.  But I haven't run this profile long enough to know it's stable yet.  So far, it's good tho.  

 

Edit-- I also disabled the ASUS performance enhancement which would push 200w to the CPU under heavy load and cause the temps to skyrocket.  Also, considering my board only has an 8-pin CPU power connector which I found was rated for 150watts max, I wanted to keep the wattage well under the max range for safety

[Somerandomtechyboi]

18 minutes ago, Wadzinsky said:

Also, considering my board only has an 8-pin CPU power connector which I found was rated for 150watts max, I wanted to keep the wattage well under the max range for safety

Cpu eps power ratings ive found to be basically irrelevant, if they actually mattered then why the hell is This possible without the damn 8 pin just burning? Btw x58 scales with both volt and freq for powerdraw so that 1.8v isnt the only concern for powerdraw, not sure on exact powerdraw but these are ancient furnaces from 2008/9 so when overvolted and clocked that high id expect quite abit over double that 150w 8pin eps rating. Oh and the board designers of back then didnt seem to bother with dual 8 pins even on the xoc boards, idk why but if it works then it works ig. I think you should be more concerned with the vrms on that board, i suspect prime b550m-a or something similar but you didnt specify board model, and yea vrms are pretty eh on the lower end prime boards

[Wadzinsky]

6 minutes ago, Somerandomtechyboi said:

Cpu eps power ratings ive found to be basically irrelevant, if they actually mattered then why the hell is This possible without the damn 8 pin just burning? Btw x58 scales with both volt and freq for powerdraw so that 1.8v isnt the only concern for powerdraw, not sure on exact powerdraw but these are ancient furnaces from 2008/9 so when overvolted and clocked that high id expect quite abit over double that 150w 8pin eps rating. Oh and the board designers of back then didnt seem to bother with dual 8 pins even on the xoc boards, idk why but if it works then it works ig. I think you should be more concerned with the vrms on that board, i suspect prime b550m-a or something similar but you didnt specify board model, and yea vrms are pretty eh on the lower end prime boards

Back in my super-noob days, I had a board with a 4-pin cpu power connector and didn't realize it had one.  I only had the mobo power connector plugged into the board and ran the computer for years that way.  It wasn't until I upgraded that I realized it had this connector and the CPU got its power from somewhere.  While I realize that it's 'probably' fine, having real-world experience in almost starting fires by drawing too much power through wires that weren't designed to carry it, I am very conscious of this now.

[Somerandomtechyboi]

4 minutes ago, Wadzinsky said:

Back in my super-noob days, I had a board with a 4-pin cpu power connector and didn't realize it had one.  I only had the mobo power connector plugged into the board and ran the computer for years that way.  It wasn't until I upgraded that I realized it had this connector and the CPU got its power from somewhere.  While I realize that it's 'probably' fine, having real-world experience in almost starting fires by drawing too much power through wires that weren't designed to carry it, I am very conscious of this now.

The hell happened to that 20 pin?

 

You prob had a really low power cpu or an odd board design, i mean my 775 stuff will just not work without the eps plugged in

 

And as far as im aware mobo 20/24 pin atx connectors are definitely not designed to draw a crapload of power so having it handle the 80w+ ancient cpu would prob not end well

[Wadzinsky]

20 minutes ago, Somerandomtechyboi said:

The hell happened to that 20 pin?

Haha!!  It was so many years ago.  I think this board had a Athlon XP 3000+ on it.  I know the connector was melted on the board and took a little effort to get it out.  As far as I know, the power supply was well within  the power rating for the whole system.  

[Somerandomtechyboi]

4 minutes ago, Wadzinsky said:

Haha!!  It was so many years ago.  I think this board had a Athlon XP 3000+ on it.  I know the connector was melted on the board and took a little effort to get it out.  As far as I know, the power supply was well within  the power rating for the whole system.  

Apparently tdp 68w, also werent these old systems still mostly reliant on 5v or have they already switched to being reliant on 12v instead? If its still a 5v reliant platform then that would explain the connector melting at a measly 68w, 68 : 5 is like 13 and a half amps which is enough to melt those psu cables. for all i know its the amperage that causes wires to melt and not voltage so no sht i can just casually hook up a 72w tec to my psu on a random molex rated for 50w (connector) and have no issues at all because its only 6a and not a connector or wire melting 13a

[Wadzinsky]

6 minutes ago, Somerandomtechyboi said:

for all i know its the amperage that causes wires to melt and not voltage

correct.  you can put a million volts through a single strand of copper as long as the amperage is low enough