Discussion Undervolting Ryzen 7950X - massive power and temperature drop for tiny performance drop

And this was a thing to recommend on previous iterations of Zen.

AMD's marketing must really love being on TSMC's process, driving the engineers to go do things like saying the thermal limit is the primary boost wall.
 
Last edited:

Math Geek

Titan
Ambassador
saw one review where they tested the eco mode and some other power limiting mode as well.

the 7600x limited to 65w barely lost any performance vs the full 130w or so it used in a torture test.

stands to reason the other chips will do as well including the 7950x. really is good news since that makes the whole system cheaper with less cooling and mobo power delivery needs.

looks like the 7600x will use about the same as the last couple 6 core ryzens which is pretty good considering what extra it brings to the table. i know it's fun to look at the top end cpu but that's not what most will be buying so i tend to focus more on the mid range and cheaper since this is likely what we'll be suggesting to most folks looking to build/upgrade
 
Why would anybody get a 7950x for gaming...
all other 7xxx cpus have faster clocks.
Undervolting is ok if you know what you are doing but it always opens up the door for crashes when running something that needs the juice.
AMD can't release a CPU that has even a tiny possibility of crashing when running some software so they have to make sure by using the max vcore the CPU can handle.
 
Personally since I've got the 5900x, I'm probably sitting with this, but I'd likely want to be at a 7700x or the 7900x. But for now I'm probably sitting tight. However, you hate to buy a shiny new cpu to have to hamper performance a bit. But I don't want my cpu's running 95 degrees either.
 
This works on pretty much all CPUs. You can drop temps on Alder Lake with little to no performance drops, too, it's not hard. Always worth a consideration.
Performance loss for watts saved seems to be much better for Ryzen's new chips. Efficency even beats the Ryzen 5000 chips on their watt-constrained platform.
Course, the Raptor is right around the corner. We'll see how Intel's new chips stack up to AMD's in a month.
 
Performance loss for watts saved seems to be much better for Ryzen's new chips. Efficency even beats the Ryzen 5000 chips on their watt-constrained platform.
Course, the Raptor is right around the corner. We'll see how Intel's new chips stack up to AMD's in a month.
The 5950x uses 120W the 12900ks is locked at 125W
intel score=24770 amd score=24838
So while the 5950x is technically more efficient it is extremely close to where a normal person wouldn't even consider it different.
It's about 1% difference in performance for about 5W difference in power.

https://www.hardwareluxx.de/index.php/artikel/hardware/prozessoren/58452-intel-holt-wieder-die-brechstange-raus-der-core-i9-12900ks-im-test-update.html?start=8
 
Reactions: KyaraM

Phaaze88

Titan
Ambassador
View: https://www.youtube.com/watch?v=y_jaS_FZcjI


That and the TechYesCity video:

Reduction in power consumption: Neat!
Performance loss: Negligible(save for nitpicking in benchmarks). Also neat.
Performance headroom: Looks to be negligible too. Ehh...

But one can do the same with an Intel cpu.
Really, this is overreaction to the full load temperature behavior. The engineers say it's safe, then it's safe.
If I had one, I'd do a small Vcore offset, and wouldn't touch the core clocks at all. It bugs me that some users still set static boost clocks when it's a double-edged sword.


Why would anybody get a 7950x for gaming...
I'd imagine for the same reasons an i9 is picked up, whether we find said reasons practical or not. Someone is going to do it.
 
View: https://www.youtube.com/watch?v=y_jaS_FZcjI


That and the TechYesCity video:

Reduction in power consumption: Neat!
Performance loss: Negligible(save for nitpicking in benchmarks). Also neat.
Performance headroom: Looks to be negligible too. Ehh...

But one can do the same with an Intel cpu.
Really, this is overreaction to the full load temperature behavior. The engineers say it's safe, then it's safe.
If I had one, I'd do a small Vcore offset, and wouldn't touch the core clocks at all. It bugs me that some users still set static boost clocks when it's a double-edged sword.



I'd imagine for the same reasons an i9 is picked up, whether we find said reasons practical or not. Someone is going to do it.

Friend of mine yesterday sent me a copy of stuff from Microcenter. Apparently he picked up a new 7900x and AsRock Steel Legend x670E, ram etc.
 

fybyfyby

Commendable
Dec 25, 2019
17
9
1,515
0
Why would anybody get a 7950x for gaming...
all other 7xxx cpus have faster clocks.
Undervolting is ok if you know what you are doing but it always opens up the door for crashes when running something that needs the juice.
AMD can't release a CPU that has even a tiny possibility of crashing when running some software so they have to make sure by using the max vcore the CPU can handle.
I really dont know how AMD sets voltage but it seems to me, they choose right V/F curve for right CCD. Maybe even cores. They definitely leave some room for undervolt. But sometimes I think it is dumb to leave so much room. Because AMD could sell that cpu in similar way like Intel is with their KS series. On my 7950x I can instantly shift V/F curve and see, how easily can cpu boost higher with lower TDP.
Maybe simply AMD doesnt have so much possibilities to fine tune every CCD so they do basic testing during binning and set V/F curve with some buffer to be on the safe side. Im really curious, how this is done.
 

ASK THE COMMUNITY