Rtx 3000 series undervolt discussion

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
8,006
I used the curve in Afterburner, so the higher boost was still available, just at lower voltage :) Used GPU-Z to monitor, the voltage was lower, power was still pegged at 350w though.
yes, you are then still power limited but at a higher clock since you are undervolted. Set power limit to where you want and as long as you have the curve set, the GPU will provide max performance for that level of power (which is a variable based on task/game/engine/res)
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,217
yes, you are then still power limited but at a higher clock since you are undervolted. Set power limit to where you want and as long as you have the curve set, the GPU will provide max performance for that level of power (which is a variable based on task/game/engine/res)

So in this case, how come at lower voltage, we are not getting less wattage? is it just pulling more amps to compensate?
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
8,006
So in this case, how come at lower voltage, we are not getting less wattage? is it just pulling more amps to compensate?
Because while it is pulling less watts, it is also doing higher clocks, which results in more work being done for the same power draw. That's why you want to set a lower power limit so you can get close to stock performance for less power, rather than more performance for same power. I'm going to revise the guide to make it a bit simpler, but as long as you have a curve set, regardless of power limit, you are running more efficient than stock. If you leave stock power limit, that equates to more work getting done rather than power savings.
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,217
Because while it is pulling less watts, it is also doing higher clocks, which results in more work being done for the same power draw. That's why you want to set a lower power limit so you can get close to stock performance for less power, rather than more performance for same power. I'm going to revise the guide to make it a bit simpler, but as long as you have a curve set, regardless of power limit, you are running more efficient than stock. If you leave stock power limit, that equates to more work getting done rather than power savings.

but it's not clocking any higher, it's about the same clocks ( I didn't raise the clocks at all), so at the same clocks, with less volts.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
8,006
but it's not clocking any higher, it's about the same clocks ( I didn't raise the clocks at all), so at the same clocks, with less volts.
Did you do your offset in a multiple of 15MHz? Sounds like whatever game you are running is just simply power limited regardless. I don't claim to be an expert, but I've spent a lot of time tweaking this stuff since I got my 3090. In my experience, even without a power limit adjustment, you should see a slight increase in clocks, unless you are in your highest boost table.
 
  • Like
Reactions: DPI
like this

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,217
Did you do your offset in a multiple of 15MHz? Sounds like whatever game you are running is just simply power limited regardless. I don't claim to be an expert, but I've spent a lot of time tweaking this stuff since I got my 3090. In my experience, even without a power limit adjustment, you should see a slight increase in clocks, unless you are in your highest boost table.
I'll post pics when I get on the rig again :)
 

Furious Nerd

2[H]4U
Joined
Sep 14, 2006
Messages
3,041
Hmm, is it possible for the card to override Afterburner curve and set its own voltage?
I have it set at 875mV and in the past when I would check HWInfo64, the max voltage would be reported at 875mV during gaming sessions.
But I just checked HWInfo64 now randomly after a gaming session and it reported a max 9xx mV
I relaunched MSI Afterburner, double checked that it's set to apply the values at Windows startup (it was), shrugged, closed Afterburner, launched Control and ran around a couple minutes to put some load on the GPU, exited and checked HWInfo64 - 875mv was the max reported voltage.
I reset HWInfo64 and went back to using the desktop and some hours later checked again and hwinfo64 reported a max of 900mv
Does anyone else have their voltage spike above the voltage they set in Afterburner?
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,217
Did you do your offset in a multiple of 15MHz? Sounds like whatever game you are running is just simply power limited regardless. I don't claim to be an expert, but I've spent a lot of time tweaking this stuff since I got my 3090. In my experience, even without a power limit adjustment, you should see a slight increase in clocks, unless you are in your highest boost table.

Here you go

1610174461685.png
 

Hresna

n00b
Joined
Jan 9, 2021
Messages
1
So I used a variety of methods to undervolt on my STRIX 3080. The OP, locking on VF points, the one with the manual flat line above a certain voltage (which I was more familiar with from my 2080Ti) and ran Timespy Extreme more times than I care to admit.

In the end I was finding things that would pass tiemspy but with noticeable glitches (which would persist even after returning to stock), and then trying them in game (AC Odyssey) only to have it crash. So I walked back a bit from timespying and now I just run AC Odyssey at 4K Ultra for a bit with a conservative undervolt.

WIth +105 to a cold-stock curve up to 981mV (and then flat from there, which changes frequency depending on what temp is), I had a sustained boost of 1995 in AC Odyssey at a cool 986mV and 66 degrees, with a custom fan curve that stayed under 55% up til 70 degrees. (55 is inaudible to me in my setup, but I do hear them a bit at 65).
I was pretty happy with that, and probably could try 120 or 135 but I would definitely try AC Odyssey gaming over TSE in the future. More fun.

I tried to reproduce that result in GPU Tweak II using a 2005 boost and an 88% power limit, but it didn’t work. Clocks were jumping all over between 1850 and 1935, and temps were at 75+ (in Control, not AC Oydssey, which I have to try... in case the RTX was the driving factor there).
 

Attachments

  • 5C42AA28-A98E-436E-A7B0-72FEBFDCEE0C.png
    5C42AA28-A98E-436E-A7B0-72FEBFDCEE0C.png
    2 MB · Views: 0

markzolla

n00b
Joined
Dec 20, 2020
Messages
14
This doesn't make any sense... I switched a 3070 Gaming X Trio for a 3070 SUPRIM, and the results are really weird. I get lower score with 2070MHz @ 981mV on the SUPRIM compared to the 2040MHz @ 981mV on the Gaming X Trio. I tryed to increase the voltage to 1000mV, and it's still slightly behind...

Now, I can understand the chip may not be as lucky, but how is it even possible that a 3070 performs worse than another one at lower clock? This really doesn't make any sense... Also this graphics card always boosts +15MHz above what I set, while the other one had the tendency to do the opposite and boost -15MHz what I set (not always like this one tho).
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
8,006
This doesn't make any sense... I switched a 3070 Gaming X Trio for a 3070 SUPRIM, and the results are really weird. I get lower score with 2070MHz @ 981mV on the SUPRIM compared to the 2040MHz @ 981mV on the Gaming X Trio. I tryed to increase the voltage to 1000mV, and it's still slightly behind...

Now, I can understand the chip may not be as lucky, but how is it even possible that a 3070 performs worse than another one at lower clock? This really doesn't make any sense... Also this graphics card always boosts +15MHz above what I set, while the other one had the tendency to do the opposite and boost -15MHz what I set (not always like this one tho).
are you setting your overclock/offset in a multiple of 15?
 

markzolla

n00b
Joined
Dec 20, 2020
Messages
14
are you setting your overclock/offset in a multiple of 15?
Yes, it was +150 on the Gaming X Trio, and it's +240 on the SUPRIM. I can't wrap my head around it... Unless there are multiple versions of the same chip, like A and non-A chips for the 20 Series?
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
8,006
Yes, it was +150 on the Gaming X Trio, and it's +240 on the SUPRIM. I can't wrap my head around it... Unless there are multiple versions of the same chip, like A and non-A chips for the 20 Series?
are you talking about time spy? If so, could you share the links to the results?
 

markzolla

n00b
Joined
Dec 20, 2020
Messages
14
are you talking about time spy? If so, could you share the links to the results?
Does TimeSpy save locally to a file? If not, I have nothing aside from the Unigine Heaven scores.

P.S. Did one more run at 2055MHz @ 981mV and this time I got a better score than the Gaming X Trio at 2040MHz, finally. Probably I did too many runs in a row before, or I don't know honestly... This graphics card is impressive for sure, being able to run 30MHz higher with the same voltage... Even tho it starts to get artifacts with TimeSpy, so not totally stable, that's why I lowered it. I could also run 2040MHz @ 950mV without any issues, even though I wouldn't know if it's because of the better components or if it's a higher binned GPU.
 
Last edited:
Top