CPU 5800x3D

Runamok81

Runner of Moks
Jul 27, 2015
394
464
troywitthoeft.com
@Draco25240 - Thank you, thank you, thank you for posting this detailed info! This is why this forum rocks! I am in a very similar situation. Same motherboard, same chip, and first time undervolting Ryzen. So much to learn! This helps.

We've got the same setup, but I am going to dig a bit deeper because I've got a noctua L9a which can't cool as quickly as the black ridge.

For comparison, I followed your PPT, TDC, and EDC settings and on an open bench, I got scores a touch lower while running about 2-4C hotter. So I can understand those PPT, TDC, and EDC settings, do you have a link or guide you used?

Have you tried the desktop PBO tool from the overclock forum thread? That makes things interesting!

Piggy-backing off your settings, I started undervolting using the PBO tool. I set all cores at -20 and ran cinebench. Was cooler and scores jumped. Took it -30 all core, and scores jumped and cooler still. I've tested as low as -50 and cinebench still completed fine (14.5K and 85C).

Right now, I just need to know how low is too low? From Google, I learned that UV errors (locks, bluescreens) only show on low loads over long periods of time, not stress testing. I understand that -30 is a deep cut. So, -50 can't be stable, right?

Currently searching for a tool that lets me know how deep I can go. I have seen a few recommendations for OCCT for stability testing UVs. https://www.ocbase.com/

With that tool, I have completed 1hr stress tests at -50 and no errors. I want to think I've got golden silicon, but more likely, I am naively missing something. How to know UV is stable? Will keep testing, and report here ...
 
Last edited:

Draco25240

Minimal Tinkerer
New User
Apr 26, 2022
3
10
Messed a bit around with the PBO2 Tool today, results seem promising. All runs done at 120 PPT, 75 TDC and 110 EDC (the setting that provided the highest score without throttling for me)
  • Highest without: 14,388 with a max temp of 87.6 (CCD1 at 88.5)
  • All-core -10: 14,626 with a max temp of 85.5 (CCD1 at 87.0)
  • All-core -20: 14,808 (1.3% from stock) with a max temp of 82.4 (CCD1 at 84.0)
  • All-core -30: 14,919 with a max temp of 80.4 (CCD1 at 81.0), just 0.5% from stock
Might be able to push this above 15,000 if I raise some power limits again; might try some day but I don't feel a huge need for it, scores are 1% from stock and thermals are great.

I also did a bit of 3DMark TimeSpy. Baseline score seems to be 11,250-ish, and it increased by 100 per -10 in CO, using the same PBO settings.

Overall pretty happy about getting this close to the stock score with good temps despite a 47mm cooler.
I'm somewhat doubting that all -30 is truly stable though, so gonna have to run some long duration tests with OCCT or/and CoreCycler to verify how high I can go and remain stable.

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

@Runamok81
Thank you, thank you, thank you for posting this detailed info!
Happy to help :) Glad to hear it's working out for you too

So I can understand those PPT, TDC, and EDC settings, do you have a link or guide you used?
Not on hand, mostly just random googling for what were good PBO values for the 5800X, I figured they'd apply well to the 5800X3D too.
88/60/90 was just googling for what the stock values for the 5600X were.
120/75/110 I got from here, and a few other random results suggested it as a sweet spot for the 5800X.
From there I just did a few +5s and -5s to the different values to how it impacted the result in Cinebench, temperatures, and % of PBO limit readings from hwinfo. Did maybe around 10 different tests in that area, and ended up with 120/70/105 as the best balance between thermals and performance (most other things in that area were +5 C hotter but only 0.1% better)

Have you tried the desktop PBO tool from the overclock forum thread? That makes things interesting!
Did so earlier today, just before you posted the comment. Results are above in this comment.

As for -50 and that, no idea. It's my understanding that the PBO2 Tool doesn't go below -30 no matter what you put into it, but not 100% sure on that, I just vaguely recall that I might have read it somewhere, so take with a grain of salt.
As for testing stability, I'm no expert, but I've seen CoreCycler referred to a fair bit for testing long-term stability, so feel free to try that as well.
 
  • Like
Reactions: scatterforce

Revenant

SFF Guru
The Gamer Way
Revenant Tech
Editorial Staff
Gold Supporter
Apr 21, 2017
1,114
1,774
I’ve got one coming in soon. I’ll be pairing it with a Mjolner case and 120mm AIO.
 

Revenant

SFF Guru
The Gamer Way
Revenant Tech
Editorial Staff
Gold Supporter
Apr 21, 2017
1,114
1,774
Yikes, that will be quite the combo to manage.

It might be, but I’m a bit more tolerant of hotter temps than most. AMD rates their chips to sit at 90C or 95C all day long. It feels wrong, but chip design has changed a lot.
 
  • Like
Reactions: REVOCCASES

Runamok81

Runner of Moks
Jul 27, 2015
394
464
troywitthoeft.com
I've got my 5800X3D dialed in and installed. I'm a bit crazy trying to cool it in a 4L case with with the Noctua L9A, but so far it is working out.
Some images from my build log. Check it out if you are curious!

 
  • Like
Reactions: rfarmer