RX series undervolt. How much have you undervolted it ?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
Yea something doesn't sound right, I don't think it should run that hot even on stock voltages. My RX 580 would get to about the mid 70sC on the standard fan curve & voltage (although it was unnecessarily aggressive & would rev the hell out of the fan to stay below 80C!). Even when I reduced fan speeds it was easy to keep it in the high 70s to low 80s.
What kind of case cooling have you got? (ao_ika_red's ninja'd me ;)).

Btw my RX 580 is a Powercolor Radeon RX 580 8GB DDR5 Red Dragon V2 Graphics Card
 
Last edited:

Stuka87

Diamond Member
Dec 10, 2010
5,757
1,632
136
Lol before i undervolted my rx580 it used to go to 90C and shut down the PC...I'm not quite sure what the temps are now but my state 7 is at 1060 mV and I can confirm it doesn't shut down anymore and my top temperature is 80C.it doesn't go past that.Although my fans go past 2000RPM and it sounds like a jet engine,I don't really mind it because it doesn't get through my headphones...would be really nice to have my fans be silent since everyone is saying that the xfx GTS xxx black edition i bought is dead silent even at peak temps....
The GPU throttling should never shut down the computer. It will just down clock to keep the temps in check. If the PC shut down, it means you hit the wall on your power supply's over current protection.

But even at that, 90C is SUPER high for that card, and I am inclined to agree with the above poster that you have a case cooling issue. Even under full load the card should not hit 90C. My RX480 Nitro+ (without undervolting) would only hit ~78C under 100% load. With undervolting it stay mostly in the upper 60's. And this was in a Mini ITX case (Corsair 380T)
 
Last edited:
  • Like
Reactions: Assimilator1

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
I forgot to mention earlier, that since I've undervolted the GPU & put the side back on my case (which has 2 fans blowing onto the GPU), when running the GPU @~100% with MW@H it was at about 58-60C! :cool: (putting the side back on alone lowered GPU temps by about 3-4C). And that is with the less aggressive fan curve!

Btw, done about a 1000 MW@H WUs (roughly 24hrs worth) at 1050mV, not a single error :), next is Einstein@H, but that is mostly only loading the GPU to ~50% so I wonder how useful it is as a stress test....
Yea the GPU load is jumping between 50-100% so is not loading it fully. Think I'll switch back to MW@H & lower the voltage 50mV.E@H GPU load.png
 
Last edited:

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
So I started off with this -
Over several WUs of MW@H (they only take ~1.5 minutes), av GPU chip power draw is ~109w (still at stock settings atm, 1150mV). GPU temp 71-72C, ambient 22-23C.
System wall draw (also running LHC/R@H 3/7 threads) ~285w.


Then went to this :-
Dropped to 1050 mV with stock 1350 MHz, system draw is down to ~259w! (same loads as previous post) Av GPU chip power draw 88w! GPU temp ~64C! :) (ambient ~23C).
Just checking the above setting, av GPU chip power draw with current MW WUs it's averaging ~89.5w, GPU temp is 59C*, fan speed ~1750 RPM, ambient ~22C (CPU load is 10 threads of R@H this time, system wall draw is ~253w).
*side of the case is on this time, it has 2 inlet fans for the GPU which dropped it's temp by ~3C. Going against it slightly is I've reduced the fan curve ramping up below 85C).

I've now dropped it to 1000mV, GPU chip power draw with current MW WUs it's averaging ~82.3w, GPU temp is ~56C, fan speed ~1750 RPM still, ambient ~22C, system draw ~242w.
I don't know if it's stable yet though! 1st 13 WUs completed without erroring. (the previous setting had ~24hrs of error free MW@H, the last 2/3 of a F@H WU & at least a few hrs of gaming (Elite Dangerous). 1/2 the voltage drop this time, but a little less than 1/2 of the last GPU chip power draw reduction. Still worthwhile though :)

[update] Elite freaked out at 1000mV @1.35 GHz, the walls came closing in on a station! lol, seems to be ok at 1025 mV.
With that voltage, av GPU chip power draw is 85.8w for MW@H (over a longer time of about 1/2hr), GPU temp is 57-58C, fan speed ~1750 RPM, ambient 22C, system draw ~246w (10 threads of LHC@H this time, non avx).
[edit] 8/4 After ~22hrs running the GPU @1025 mV on MW@H I'll say it's good enough to move onto the acid test of F@H.....if I can get WUs!

9/4 - Damn it, didn't notice until after 3 aborted WUs that it was failing! F@H logs failures but doesn't notify you of them!
Increased GPU vcore to 1050mV.
 
Last edited:

Shmee

Memory and Storage, Graphics Cards
Super Moderator
Sep 13, 2008
6,026
1,394
136
On my Sapphire Vega 64 Nitro+ I started with a 150mV offset from each Pstate starting at P7, and the original P7 was 1200mV. Was fine at first but then I got an error on F@H after a while, so I increased each Pstate I toughed by 50mV. Testing now, so far so good. Temps and power are still a lot better.
 

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
Cool, what drop in power & temps did you get?

(I wish F@H would at least alert us to computational errors! e.g. flash the program red, &/or give a notification. And BOINC is no better! :p)

I've done a few F@H WUs now @1050mV, so far so good :).
 

Shmee

Memory and Storage, Graphics Cards
Super Moderator
Sep 13, 2008
6,026
1,394
136
You may have seen my post in DC, but I will go ahead and mention it here as well.

Temps currently are at 65C, down from over 80 C at stock. Power consumption is down right now at about 185W, down from around 220W before.
 
  • Like
Reactions: Assimilator1

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
Yea I did, but I thought it would be worthwhile to have it here too :).
That's a whooping drop in temps! And a very decent drop in power :cool:
 

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
At 1350 MHz @1050 mV I got a single error (which it recovered from) in 1 amongst several WUs in F@H, so I've dropped it to 1325 MHz.

08:46:58:WU00:FS01:0x22:Bad State detected... attempting to resume from last good checkpoint. Is your system overclocked?
08:46:58:WU00:FS01:0x22:Following exception occured: Particle coordinate is nan
08:49:27:WU00:FS01:0x22:Completed 3160000 out of 4000000 steps (79%)


Damn, got another 'nan' today (12/4), dropping clock to 1300 MHz.
 
Last edited:

Assimilator1

Elite Member
Nov 4, 1999
23,975
450
126
6 days latter and I've got 1 more :-

08:08:32:WU01:FS01:0x22:Completed 850000 out of 5000000 steps (17%)
08:10:05:WU01:FS01:0x22:Bad State detected... attempting to resume from last good checkpoint. Is your system overclocked?
08:10:05:WU01:FS01:0x22:Following exception occured: Particle coordinate is nan
08:13:12:WU01:FS01:0x22:Completed 900000 out of 5000000 steps (18%)


Getting close to stable though! ;)
Dropping to 1290 MHz.

And today (22/4) I got a failed WU with 'nan', downclocked to 1275 MHz.
03/05 a 2nd failure at 1275 MHz, dropped to 1266 MHz. Interestingly (if stable), it's the default clock rate of an RX 480.
 
Last edited:

ASK THE COMMUNITY