• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

3080 Undervolting

Caporegime
Joined
21 Jun 2006
Posts
38,372

1900 at 887mV is the sweet spot.

I found I used 80W less power than stock. Yet had more power.

I also tried 2000 at 987mV and I gained 3% in benchmarks yet used 80W more power.

290W vs 370W.

Definitely undervolt your card guys. Makes zero sense in trying to get the absolute best clocks apart from benchmark score e-peens.
 
Associate
Joined
19 Feb 2020
Posts
324
I'm also a fan of undervolting. My 3080ti is at 1800mhz @ 800mv. I don't recall my exact wattage but it's under 300w (350w stock). My undervolted TimeSpy Vs stock TimeSpy was margin of error better and with 5C average temperature lower.
 
Soldato
Joined
26 May 2014
Posts
2,955
I've found 950mV to be a great spot if you want to lose essentially zero performance versus a maximum overclock. All the Ampere cards I've touched have been able to do ~2GHz at that voltage (most slightly over, with the best hitting a locked 2040MHz I believe). A 3070 will pull in the 180-190W range at that voltage, which is pretty incredible perf/watt and performs outright better than stock on the FE at least, since it keeps it well away from the power limit. Of course, you can go even lower if you want the absolute best efficiency and don't mind trading a little performance.
 
Soldato
Joined
27 Mar 2009
Posts
3,301
Also underclocked here, 1800 @ 850mv, could run faster in benchmarks but would crash in full RTX/DLSS games after an hour or 2.

Plus a little headroom on the voltage as it boost 1815-30 until the temps get up high.
 
Associate
Joined
10 Feb 2021
Posts
608
My 3090 I run at 1815 @ 850mv, keeps power usage at or below 300W. Mostly it helps tame the coil whine.
I am easily able to game / complete benchmarks at 1850@850mv too, but had the odd crash on older games (GTA V mostly) so dropped a little to 1815. Been solid for months now.
 
Soldato
Joined
28 Oct 2009
Posts
5,294
Location
Earth
been testing 1905mhz at 887mV does seem the sweet spot passes stock performance with added benefit of lower temps/power usage some reason I cant lock it to 1900mhz as soon I apply it locks to 1905

I did try 1950 and ran some benchmarks only slightly improved was expecting more
 
Caporegime
OP
Joined
21 Jun 2006
Posts
38,372
been testing 1905mhz at 887mV does seem the sweet spot passes stock performance with added benefit of lower temps/power usage some reason I cant lock it to 1900mhz as soon I apply it locks to 1905

I did try 1950 and ran some benchmarks only slightly improved was expecting more

Yeah I tried 2000 and like I said 80w more power usage for 3% gain in benchmarks.

Absolutely pointless trying to squeeze more out of it. The gains are so small yet the power usage goes up exponentially.

25% more power for 3% performance. Makes zero sense to try higher other than benchmark e-peen.
 
Soldato
Joined
7 Oct 2003
Posts
3,888
Location
York
1920/1905 at 900mv here. lower volts are fine for most normal gaming and benchmarks but RT stuff will randomly crash. At 900mv it seems stable for everything. clocks start at 1920 and then drop to and stay at 1905 more-or-less permanently (for games that arent hitting the top of the power budget) once the card get up to temperature.
 
Soldato
Joined
12 Jun 2007
Posts
8,737
Location
Ireland
been testing 1905mhz at 887mV does seem the sweet spot passes stock performance with added benefit of lower temps/power usage some reason I cant lock it to 1900mhz as soon I apply it locks to 1905

I did try 1950 and ran some benchmarks only slightly improved was expecting more

same I tried 2000 with 980mv ,but was using to much power , 1905mhz at 887mV good sweetspot, i think it uses 270w on that
 
Soldato
Joined
6 May 2009
Posts
19,923
I find running my 3080 at stock is easiest and works best.

Max temp 64c after a few hours of Warzone with FPS limited to 97. Core max 1995, mem 9502. Voltage max 1.081, max 320w
This is using a custom fancurve starting 30% @ 45c, ramping to 100% @ 80c
 
Soldato
Joined
28 Oct 2009
Posts
5,294
Location
Earth
I find running my 3080 at stock is easiest and works best.

Max temp 64c after a few hours of Warzone with FPS limited to 97. Core max 1995, mem 9502. Voltage max 1.081, max 320w
This is using a custom fancurve starting 30% @ 45c, ramping to 100% @ 80c

depends on the game /res and how taxing it cant imagine it being when locked 97fps

my card at stock would run upto 74c+ using 300w+ , more power used / higher fan speed, now at 1905 / 887mV now in the 60s and low 70s with lower fan speed mostly now below 300w and beats stock benchmarks .. really is worth doing

also my case isnt the best for airflow hoping to upgrade soon
 
Last edited:
Caporegime
OP
Joined
21 Jun 2006
Posts
38,372
I find running my 3080 at stock is easiest and works best.

Max temp 64c after a few hours of Warzone with FPS limited to 97. Core max 1995, mem 9502. Voltage max 1.081, max 320w
This is using a custom fancurve starting 30% @ 45c, ramping to 100% @ 80c

easiest - obviously, you aren't doing anything bar plugging it in and turning it on

best? - I don't think using considerably more power and getting less scores in benchmarks is best. I can eek more power but then as I've already stated it's 3% performance gain yet using 30% more power. That isn't a good ratio and law of diminishing returns has kicked in. I don't even need to set a custom fan curve. Using 80W less power means it runs as cool as ice. The ram is the only issue and there is nothing you can do to fix that bar modding the card.
 
Soldato
Joined
26 May 2014
Posts
2,955
You do know overclocking ram on the 3080 past 250 nets you less fps right?

Jays 2 Cents made a video on it
That's going to be very sample-specific. From my own experience and reports I've read, memory overclocking varies hugely between each card with Ampere. The 3070 Ti I have does +1000 with performance increases all the way, verified and tested with a bunch of benchmarks and games (including Heaven, which is what Jay used). Of the two 3060s I've had my hands on, one could do +1150 without issue, whilst the other just hard locked up at anything above +700. Silicon lottery, innit.
 
Caporegime
OP
Joined
21 Jun 2006
Posts
38,372
That's going to be very sample-specific. From my own experience and reports I've read, memory overclocking varies hugely between each card with Ampere. The 3070 Ti I have does +1000 with performance increases all the way, verified and tested with a bunch of benchmarks and games (including Heaven, which is what Jay used). Of the two 3060s I've had my hands on, one could do +1150 without issue, whilst the other just hard locked up at anything above +700. Silicon lottery, innit.

it's specific to the new gddr6x ram so doesn't apply to 3060
 
Back
Top Bottom