Question Undervolting Ryzen 9 5900x?

G-star93

Honorable
Oct 19, 2014
560
0
11,060
43
Hi all!
Last month I built a new system.

Lian Li Lancool II mesh RGB(upgraded stock fans to Bequiet 4 x 120mm and 1 x 140)
Ryzen 9 5900X
Bequiet Dark rock pro 4
Arctic MX-4 2019 edition thermal paste
MSI B550 Tomahawk
Crucial Ballistix RGB 3600 32gb(4x8)
MSI RTX 3080 Z Trio 10gb
PNY XLR8 1TB M.2 nvme
Crucial P1 1TB M.2 nvme
Asus ROG 850w 80+gold

Everything runs fine, even better than expect it, if I'm honest. I'm using dual monitor setup. My main monitor for gaming is Dell S271DGFA and my second one(for better productivity :D) is Dell U2412M.

The only complaint that I have is CPU voltage and temperatures. I know the only real program for monitoring temps on AMD cpu is AMD Ryzen master, all the others lie for couple degrees(3-4c).
In gaming I'm hitting avg 65c-72c celsius. I would like to lower my temps for around 5c without an impact on performance. All I did so far was playing around with PBO2 in BIOS and I set it to negative and max 30. In cinebench R23 I even got almost a 1000 points higher score, but the temps stayed the same. So I was looking around on the internet about safe lowest voltage for my cpu but I couldn't find anything useful. So if any of you have the same cpu and had problems with high temps aswell, how did you fixed your problem? With undervolting, or just PBO2? What voltage and settings etc?

Thanks in advance
 
... I know the only real program for monitoring temps on AMD cpu is AMD Ryzen master...
There's also HWInfo64, look to CPU Die (Average) readings. That's going to be very close to RyzenMaster.

IMO using HWInfo64 is preferred since RyzenMaster's service is known to affect performance while gaming. So if not in need of it's complete feature set it's best to not run it (or even have it installed) just for monitoring temperatures. RM was created as a tool for extreme overclockers to use for tweaking manual overclocks (especially when using LN2) for demonstrations and competitions. For them, actual performance in applications isn't the goal, just getting an extreme overclock even if it's going to be dynamically unstable in applications.

Slight undervolting can help with Ryzen 5000, usually when done with PBO and the curve optimizer. But be aware temps will always be hotter than you may be used to, it's normal and expected and it's designed for it. Tjmax for a 5900x is 90C.
 
Last edited:

G-star93

Honorable
Oct 19, 2014
560
0
11,060
43
There's also HWInfo64, look to CPU Die (Average) readings. That's going to be very close to RyzenMaster.

IMO using HWInfo64 is preferred since RyzenMaster's service is known to affect performance while gaming. So if not in need of it's complete feature set it's best to not run it (or even have it installed) just for monitoring temperatures. RM was created as a tool for extreme overclockers to use for tweaking manual overclocks (especially when using LN2) for demonstrations and competitions. For them, actual performance in applications isn't the goal, just getting an extreme overclock even if it's going to be dynamically unstable in applications.

Slight undervolting can help with Ryzen 5000, usually when done with PBO and the curve optimizer. But be aware temps will always be hotter than you may be used to, it's normal and expected and it's designed for it. Tjmax for a 5900x is 90C.
I know I also have HW monitor, but like I said it's not as accurate as Ryzen master. I know TJmax is 90c but it's not good for cpu lifespan that is running close to 90c for a long time.
 
I know I also have HW monitor, but like I said it's not as accurate as Ryzen master. ...
HWMonitor is notoriously bad for Ryzen and AM4 motherboards. It may have improved with some later releases but it still lacks any sort of logical naming of sensor readings to assist with understanding anything. You're often left looking at the wrong sensor, or even an unused sensor that's reporting random numbers instead of actual temperatures.

In addition to being far more complete HWInfo organizes the sensors so you know the source of the readings, and as well has very informative tool-tips to explain most of them.

.... I know TJmax is 90c but it's not good for cpu lifespan that is running close to 90c for a long time.
Not according to AMD's engineers. You can trust the FUD-spreaders of the internet if you want though.

 

G-star93

Honorable
Oct 19, 2014
560
0
11,060
43
HWMonitor is notoriously bad for Ryzen and AM4 motherboards. It may have improved with some later releases but it still lacks any sort of logical naming of sensor readings to assist with understanding anything. You're often left looking at the wrong sensor, or even an unused sensor that's reporting random numbers instead of actual temperatures.

In addition to being far more complete HWInfo organizes the sensors so you know the source of the readings, and as well has very informative tool-tips to explain most of them.



Not according to AMD's engineers. You can trust the FUD-spreaders of the internet if you want though.

Believe what you wan't but also with less voltage there's less heat to distribute. So that's the goal. So the pc can run cooler, quieter and there's less heat that comes out of it.
Yea I know HW monitor is bad with ryzen cpu's right now I'm looking at 9-10c celsius difference between RM and HW monitor.
 
Believe what you wan't but also with less voltage there's less heat to distribute. So that's the goal....
It's not a religion so it's not a matter of 'what I believe'. The designers put spec's out there for a reason, it's a matter of who I trust to be telling me the correct specs and how they relate to performance. Read that article...Robert Halleck's explaining a bit of the functioning of the CPU's boosting and how it relates to the 90C limit.

Keeping it cool is good, though, as it helps the CPU stay boosting cores higher and for longer under load. If it's not hitting the 90C limit it will push as far as it can within it's other FIT parameters in the boosting algorithm...SOME of which you can open up with PBO. But be careful about pulling back too far on voltage as it can actually hurt performance, possibly even before it starts pushing the processor into instability.
 
Temperature by itself is not exactly a good metric in a vacuum to use as a comparison. For instance, you said that you gained higher benchmark scores in Cinebench. Great! That also means the CPU worked harder. This implies the CPU boosted harder for longer because it believed it had more thermal headroom to do so. Performance doesn't simply come free just because you undervolted.

If lowering temperatures are your goal, you may need to:
  • Rein in how fast the thing turbo boosts to
  • Set a fixed voltage/clock frequency
  • Get a better cooler
 

ASK THE COMMUNITY

TRENDING THREADS