[SOLVED] RX 580 using up to 1200 mV in auto voltage and poor OC

SoldierWNoName

Commendable
Jan 4, 2020
23
1
1,515
I got a XFX RX 580 GTS XXX (RX-580P8DFD6, base clock of 1366, guaranteed OC 1386 Hz) recently and I tried to OC it, but I wasn't getting good results, then I tried to undervolt, as is a hot card and I'm in the summer. I couldn't get bellow 1130 mV in Heaven benchmark, while playing Borderlands 3 or The outer worlds it would glitch after some time, normally more than 1 hour, even with 1142 mV for the 1366 (consuming ~140 W). With 1137 it'd glitch sooner, but the power draw dropped to ~135 W.
If i just let in auto setting the power limit to +50%, the GPU would average around 1330 Hz consuming ~150 W, peaking at 160 W. If I don't touch the power limit the difference isn't big.
Then I had the idea of overclock using the auto voltage, it can get the 1386 but will reach 1200 mW (almost all the time stays above 1150) and use ~160 W.
While if I set the state 7 to 1366 and let the voltages in auto, the card would consume also around 150 W, but it'd pump its frequencies to 1366 basically all the time, but also using voltages above 1150, up to 1200.
With the graphics card consuming around 140 W, with room temperature around 20 °C, it'll get ~74 °C with 2400 rpm (almost 70%). These results are with my tower side panel open or if the 2 side fans are set at 7 V.
Obs.: I used GPU-Z for the results of the voltages, and the Radeon software while gaming.

Now the questions. Is it normal for Polaris cards to use considerably more voltage in auto mode?
Am I doing something wrong, is this bad luck in the silicon lottery, or it can be something else, like the mobo or PSU?
Can I be undervolting above the 1150 mV, as in auto is already using more than that?
Can be dangerous in the long run if I pump the voltage to 1200 mV, or at least 1160 mV, how about auto mode using 1200?
Is this power usage normal? In Heaven btw it draw around 15 less watts.
Is this a normal temperature?

Sorry for the long post.
 
Solution
That's fine if you don't expect the card to drop below state 7 frequency ever, but lowering voltage for the lower power states can help the cooler "bank up" some cooling capacity during dips into the lower frequency states (FWIW). Also if you're using FRTC or Chill for lighter weight games (Warframe, LoL, CSGO, etc) where the card doesn't need to be hitting >1300MHz to produce astronomical frame rates, having lower voltages for the lower frequencies will allow the card to run cooler and quieter (for instance, my card doesn't even need to spin its fans in those listed games)

Of course that also requires you to test those lower frequency/ voltage settings for stability also, so more time required. I throw my test settings in a scatter...
Yes, Polaris cards use a large amount of voltage in auto voltage. Generally 50-75mV more than what's necessary to maintain stability.

I've been able to hit 1400MHz @ 1150mV typically. Of course every chip is different (silicon lottery). Did you work with manual voltage entries for each performance state?
 

SoldierWNoName

Commendable
Jan 4, 2020
23
1
1,515
Yes, Polaris cards use a large amount of voltage in auto voltage. Generally 50-75mV more than what's necessary to maintain stability.

I've been able to hit 1400MHz @ 1150mV typically. Of course every chip is different (silicon lottery). Did you work with manual voltage entries for each performance state?
I usually set all the 4 last states with the same voltage, as the original values are 1150 for the last four, the first 4 I don't touch.
 
That's fine if you don't expect the card to drop below state 7 frequency ever, but lowering voltage for the lower power states can help the cooler "bank up" some cooling capacity during dips into the lower frequency states (FWIW). Also if you're using FRTC or Chill for lighter weight games (Warframe, LoL, CSGO, etc) where the card doesn't need to be hitting >1300MHz to produce astronomical frame rates, having lower voltages for the lower frequencies will allow the card to run cooler and quieter (for instance, my card doesn't even need to spin its fans in those listed games)

Of course that also requires you to test those lower frequency/ voltage settings for stability also, so more time required. I throw my test settings in a scatter plot so I can visually interpolate between tested values.

As an FYI, here's some points along the curve that I've generally found stable for most cards.

1000MHz = 890mV
1100MHz = 910mV
1200MHz = 940mV (this 940mV is typically an inflection point on the curve, it gets steeper above this value. Notice that values below this are ~25mV/100MHz and values above are 100mV/100MHz)
1300MHz = 1030mV
1400MHz = 1150mV

VRAM voltage is usually fine at 930-950mV. (that sets the lowest voltage that will be applied to the core when not idle).

If you want to learn more, here's a guide I put together.
 
Last edited:
Solution

SoldierWNoName

Commendable
Jan 4, 2020
23
1
1,515
I do use Chill, and I make separate profiles for each games, specially because it seems not possible to have a custom fan curve and zero rpm mode both enable at the same time.
But I have a 1440p monitor, so for most 2017 and newer games (some older too) the GPU will easily bottleneck the system.
I will try undervolting the VRAM, thanks
Do you think is normal for the card not being able to run (glitches over time, not driver crashes) at stock speeds and 1150 mV?
 

SoldierWNoName

Commendable
Jan 4, 2020
23
1
1,515
Glitches may be a CPU/RAM issue. What's your full system spec? What slots are your RAM in? Is your RAM running at advertised speeds
4 sticks of HyperX 2400 MHz, running at 2400, a 3600x also in stock speeds, and an Asus B450M gaming Br, same as the B450M-A.
When the voltages for the GPU are in auto mode a have no glitches, just when I try to undervolt, even at stocks speeds and 1142 mV, with 1150 mV seems to be stable, but if I pump just to 1386 MHz (guaranteed overclock spec) the 1150 mV aren't enought