Rtx 3000 series undervolt discussion

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
Let's talk undervolting. Creating a custom voltage curve should both reduce power usage and possibly increase performance (if it reduces Temps enough to get you into higher boost bins). Here is the guide I have created, thanks to everyone who gave feedback and helped get this tested and revised.

Prerequisites: Install MSI Afterburner, 3DMark Time Spy, and BMI Benchmark.
*Changes to the curve should be done when COLD, and frequency adjustments should be made in 15MHz increments, as that is the amount between boost tables. Afterburner will maintain your curve once you set it, but whenever you try to revise it, it will be based on the reading of the CURRENT boost table the card reports, based on temperature. The curve you set is not based on the actual frequency you set, but rather the OFFSET. So +150 on a hot curve may give the same result as only +105 on a cold curve, so we want to be as consistent as possible when making changes to the curve by having the card cold each time. You can accomplish this more quickly by running the GPU fan at 100% when you are waiting for it to come down in temp. You should be able to get it under 35C, the closer to ambient the better.



  1. On a cold system (idle at least 10 minutes), open MSI afterburner and reset to stock profile. The default boost curve changes with temperature so being cold is important.
    1.jpg
  2. Press CNTL + F to open the voltage and frequency scale.
  3. Note the rightmost slider’s frequency. This is going to be our target, but at a lower voltage. In this case, it is 2040MHz.
    3.jpg
  4. Now note the leftmost slider that this frequency appears, usually about 125mV less. In this case, it is 1118mV.
    4.jpg
  5. Increase the frequency of the 800mv slider through the slider found in step 4, in 25mv increments, to +105MHz, but never exceeding the frequency found in step 3. You can see in this case, we achieve the 2040MHz goal at 1000mV, and don't need to go any further.
    5.jpg
  6. Apply the curve using the apply button on the main afterburner window. Your curve should change to something like this:
    6.jpg
  7. Run Time Spy.
  8. If you encounter a crash, error, or if Time Spy gives a score of 0, the curve is unstable. Allow the system to cool for at least 15 minutes, then start again at step 1, but with 15MHz less, in this case, +90 instead of +105. Do this until you are Time Spy stable, +105, +90, +75, +60, etc.
  9. Conversely, if you are stable at +105 and want to try to achieve more performance, you can allow the system to cool for at least 15 minutes, then return to step 1, but with 15MHz MORE, +120 instead of +105, and so on, until you encounter instability. Once you do, I recommend removing 30MHz from that point of first instability, to ensure it is stable in various scenarios other than Time Spy. Again, you must wait until the card is cool before making changes to the curve, or they will not apply correctly.
  10. Once you are Time Spy stable, run the BMI benchmark at the highest resolution your monitor supports, on a loop for 30 minutes. If you exhibit a crash, allow the system to cool for at least 15 minutes, then start again at step one with 15MHz less, but you can skip Time Spy since it has already been tested. Run BMI for 30 minutes and repeat until you do not crash in this test. DFenz has informed me through his testing that Hitman 2 seems to reveal instability better than many other games, so it may be useful for you to test using it as well. The free starter pack is available on steam here.
  11. Now that you have your curve dialed in, we have found the frequencies at each voltage to have the card work as efficiently as possible. You can adjust the power limit target to whatever you desire, and the card will operate much more efficiently vs stock. 90% power limit with the curve should offer about 100% of stock performance, while 80% power limit should offer about 95%. Going below 80%, performance will start to fall off. Remember, this is the power LIMIT. When not power limited in a game, your card will still boost up to the factory boost clock noted in step 3! This is useful in lower resolution, high FPS, or poorly optimized games, and offers an improvement over setting a single frequency to a single voltage.
  12. Once you have found settings you are happy with, enjoy the card and play some games. You want to note any crashes so you can correct the curve. Again, changes to the curve should be done when COLD, and frequency adjustments should be made in 15MHz increments, as that is the amount between boost tables. Afterburner will maintain your curve once you set it, but whenever you try to revise it, it will be based on the reading of the CURRENT boost table the card reports, based on temperature. The curve you set is not based on the actual frequency you set, but rather the OFFSET. So +150 on a hot curve might give the same result as only +105 on a cold curve, so we want to be as consistent as possible when making changes to the curve by having the card cold each time.


The end result is this in my case, with an Nvidia RTX 3090:

397721_Untitled.jpg



Obviously this guide exists to help you find more ideal settings in power draw and performance. It is up to you if you want to set your power limit lower or leave it at 100% and benefit from somewhat higher clock speeds in the mid range. I hope you share with us the results you find, and any feedback will be taken in order to revise and improve this guide.
 
Last edited:

Zorachus

[H]F Junkie
Joined
Dec 17, 2006
Messages
9,150
I have the eVga XC3 Ultra RT 3080, no idea how to undervolt? I have the Precision X1 software, just not sure what settings to adjust without screwing it up.
 

Zorachus

[H]F Junkie
Joined
Dec 17, 2006
Messages
9,150
Seemed pretty straightforward after I downloaded that MSI afterburner program.

So now what? it'll just stay like that even after restarting and shutting down the system?

That's will help make the card run cooler I assume? With little performance loss?
 

BassTek

Supreme [H]ardness
Joined
Jul 13, 2002
Messages
5,779
My first card (EVGA XC3 3080) wasn't totally stable at 1905@850, seemed fine at 1890@850 but I didn't test a whole bunch although it did score better at that setting in Time Spy than it did at stock.

Haven't tested my Gigabyte Aorus Master 3080 yet, it's pretty cool and quiet at stock but would still probably preform better than stock by optimizing voltages.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
you should play some games to check stability. You may get crashes if you took it too far. Just like overclocking, undervolting can push hardware to its limit. Afterburner has an option to run with windows startup, you have to enable that, and then click the windows logo icon in afterburner, which will apply your profile (don't forget to save the profile by clicking the save button [floppy logo] and then one of the numbers, which are your profiles.)
Seemed pretty straightforward after I downloaded that MSI afterburner program.

So now what? it'll just stay like that even after restarting and shutting down the system?

That's will help make the card run cooler I assume? With little performance loss?
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
I'm actually not thrilled with that tutorial. It does the job, but there is more too it. I'm working on my own, should have something for you guys tomorrow that both maximizes clock speed and minimizes voltage. It will be an easy to follow process.
 

Zorachus

[H]F Junkie
Joined
Dec 17, 2006
Messages
9,150
I'm actually not thrilled with that tutorial. It does the job, but there is more too it. I'm working on my own, should have something for you guys tomorrow that both maximizes clock speed and minimizes voltage. It will be an easy to follow process.

Looking forward to your guide
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
Take with a grain of salt as always, but apparently a 3080 Ti is back in the running with the following specs:

RTX 3080 Ti FE:
PG133-SKU15,
GA102-250-KD-A1,
20GB GD6X,
the same FP32 count as 3090,
10496FP32,
the same MEM speed and TGP as 3080,
no NVLINK.

https://twitter.com/kopite7kimi/status/1323785556417863680?s=20
RTX 3090 for sale. Well maintained, oil changes every 3000mhz. :p

But really, thats cool, but what's it have to do with undervolting (this thread)?
 

spine

2[H]4U
Joined
Feb 4, 2003
Messages
2,650
This thread is dumb, under-volting just so you can theoretically see a high boost frequency?!

Less power = less performance. Period.
 

VirtualMirage

Limp Gawd
Joined
Nov 29, 2011
Messages
384
This thread is dumb, under-volting just so you can theoretically see a high boost frequency?!

Less power = less performance. Period.
Less power can mean less potential peak performance, but there is a tipping point where it provides diminishing returns. Added power creates more heat and depending on the clock/temp curve set on the GPU, the added power could give you a spike in peak performance but the heat generated will cause it to settle down at a lower average core clock than a GPU that has found the right power/temp ratio via undervolting and adjusting the clock/temp curve.

So, yes, undervolting may keep you from hitting higher peak clock speeds, but the idea is that the average clock speed once temperature starts to settle in will be higher since there is not as much voltage creating heat.

Just adding more power is the brute force method and you need to have the cooling capability to maintain it. Also keep in mind that some GPUs bin better than others and are capable of performing at the same level or better than another GPU but with less power.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
Less power can mean less potential peak performance, but there is a tipping point where it provides diminishing returns. Added power creates more heat and depending on the clock/temp curve set on the GPU, the added power could give you a spike in peak performance but the heat generated will cause it to settle down at a lower average core clock than a GPU that has found the right power/temp ratio via undervolting and adjusting the clock/temp curve.

So, yes, undervolting may keep you from hitting higher peak clock speeds, but the idea is that the average clock speed once temperature starts to settle in will be higher since there is not as much voltage creating heat.

Just adding more power is the brute force method and you need to have the cooling capability to maintain it. Also keep in mind that some GPUs bin better than others and are capable of performing at the same level or better than another GPU but with less power.
Would you be willing to try some afterburner settings I'm testing for an underclocking guide? I'm testing on my FE 3090 currently but want to make sure it scales across a few cards before posting it.
 

Mylex

Gawd
Joined
Aug 30, 2018
Messages
839
This thread is dumb, under-volting just so you can theoretically see a high boost frequency?!

Less power = less performance. Period.
If you don't know what you are talking about its best to just listen. The 3080 series cards have a heavy handed power curve that in alot of cases are unecessary. This is one of my cards in the picture below the score on the right is from "undervolting", its actually setting the proper curve for my chip. It runs 70 watts less, 5-7 degress C cooler and performs the same or better overall. If this design wasn't such shit on Samsung's process this wouldn't be happening but Nvidia shoved turds out the door. This was something AMD cards needed before this gen.

undervolt.JPG
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
If you don't know what you are talking about its best to just listen. The 3080 series cards have a heavy handed power curve that in alot of cases are unecessary. This is one of my cards in the picture below the score on the right is from "undervolting", its actually setting the proper curve for my chip. It runs 70 watts less, 5-7 degress C cooler and performs the same or better overall. If this design wasn't such shit on Samsung's process this wouldn't be happening but Nvidia shoved turds out the door. This was something AMD cards needed before this gen.

View attachment 295861
could you sent me a screenshot of the curve you set?
 

Mylex

Gawd
Joined
Aug 30, 2018
Messages
839
could you sent me a screenshot of the curve you set?
Before it would boost to 2000+ for extremely short burst @1.075 but the performance difference was terrible and not worth using the out of the box curve.
power curve.JPG
 
Last edited:

Micas99

Weaksauce
Joined
Aug 5, 2017
Messages
70
The curve I'm using now on a 3080; 1800mhz, .806v I've setup a few more higher performance under-volt curves, but the power increases, so additional heat and so on, for a very marginal performance gain. Quite unnoticeable when actually playing a game with the lower clock speed/voltage. It does appear that at default, Nvidia pushed it really hard. I imagine that most non-hardcore gamers just plug the thing in and go. Good thing winter is coming.


3080-3.jpg
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
as I work on a guide to undervolt without sacrificing performance, I could use some people to try my method for a week or so and report back if they have problems.
 
Last edited:
  • Like
Reactions: DFenz
like this

DFenz

Gawd
Joined
Apr 3, 2014
Messages
1,022
After testing yesterday on my card atleast it seems to run most efficient at about 320w, 812mv 1785mhz, +500 on memory (tried +700 and scores went down, probably would need to increase power slightly).
At these settings I get a 19772 graphics score in Time Spy, which is a lot better than stock while putting off less heat. Stock was 18106 according to my first run. I don't use the wccftech method.
 
Last edited:

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
After testing yesterday on my card atleast it seems to run most efficient at about 320w, 812mv 1785mhz, +500 on memory (tried +700 and scores went down, probably would need to increase power slightly).
At these settings I get a 19772 graphics score in Time Spy, which is a lot better than stock while putting off less heat. Stock was 18106 according to my first run.
Nice. I'm going to do some testing at 85% (297.5w) to see how it fairs. At 315w it literally left no performance on the table vs stock. As far as memory goes, I haven't really looked into it. For my gaming profile I've been using +250mhz. I know with this gen you aren't supposed to get visible memory errors or crashes, but scores will go down, so I need to find a consistent enough test for that.

For your time spy score, post it in the time spy thread! https://hardforum.com/threads/rtx-3090-h-owners-official-3dmark-time-spy-leaderboard.2003111/
 
  • Like
Reactions: DFenz
like this

Recipe7

Limp Gawd
Joined
Jul 25, 2011
Messages
437
I did some undervolting the past few days and managed to come up with some useful numbers in regards to stability.

My setup is a 9900k running 5ghz at 1.28V and an air cooled 3080 FTW3 running at 100% fan. Ran a lap of Port Royal for all voltages. Ambient is roughly 25C.
I tested from 0.850v to 0.975v. Stopped at 0.975v due to throttling.

Stock voltage results at +118% PT, ambient 25C:
1.081v (throttled down to 1.068) 1980core (throttled down to 1950) +0mem 70C (Core boosts to 1980 with updated BIOS at 118% PT)
1.081v (throttled down to 1.043) 2130core (+150 core OC, then throttled down to 2070) +500 mem 74C

Undervolting results at 118% PT, ambient 25C:
0.850v 1950core +500mem 58C
0.875v 1995core +500mem 58C
0.900v 2010core +500mem 59C
0.925v 2015core +500mem 60C
0.950v 2025core +500mem 62C
0.975v 2055core +500mem 68C, drops to 2044core due to throttling


Bonus:
Running SOTR at stock 1.081v I was pushing 80C with 1.081v 2130core 500mem.
Running SOTR with 0.900v I was pushing 64C with 2010core and 500mem.

Bottom line:
  • I love the drop in temps for gaming, +15C. I didn't even bother checking the drop in value of FPS because I was so keen on the temps (I'm guessing 4-5fps?)
  • Looking forward to picking up a waterblock soon. I'm hoping to hit 2100core 500mem at 1.025v while keeping the GPU temps in the 40s.
  • I much prefer Afterburner for overclocking/undervolting, but need X1 due to fan control. I can use both but it is a huge hassle. Bypassing the fans and RGB on the FTW3 should net a couple more watts for the GPU to utilize.
 
Last edited:
  • Like
Reactions: DFenz
like this

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
  • I love the drop in temps for gaming, +15C. I didn't even bother checking the drop in value of FPS because I was so keen on the temps (I'm guessing 4-5fps?)
  • Looking forward to picking up a waterblock soon. I'm hoping to hit 2100core 500mem at 1.025v while keeping the GPU temps in the 40s.
I doubt you are losing 4-5FPS. You can see in my test results I lost 6FPS in shadow of the tomb raider going from 350w to 245w, albeit with a 3090, and based on your results I'd guess you are in the 275w range, probably down 2-3fps if that.
 

Recipe7

Limp Gawd
Joined
Jul 25, 2011
Messages
437
I doubt you are losing 4-5FPS. You can see in my test results I lost 6FPS in shadow of the tomb raider going from 350w to 245w, albeit with a 3090, and based on your results I'd guess you are in the 275w range, probably down 2-3fps if that.
Good to know! Thank you
 

Micas99

Weaksauce
Joined
Aug 5, 2017
Messages
70
  • I much prefer Afterburner for overclocking/undervolting, but need X1 due to fan control. I can use both but it is a huge hassle. Bypassing the fans and RGB on the FTW3 should net a couple more watts for the GPU to utilize.
I'm using a 3080 XC3 Ultra Gaming. With afterburner, the 3 fans appear to be in sync. The fan is set to auto and I ticked the "synchronize fan speeds" button. I added "fan speed" and "fan speed 2" to the monitor, and it looks like it's working. I could be totally wrong since I'm just looking at the fans and eyeballing it, but it looks like all 3 fans are changing speed properly.
 

Recipe7

Limp Gawd
Joined
Jul 25, 2011
Messages
437
I'm using a 3080 XC3 Ultra Gaming. With afterburner, the 3 fans appear to be in sync. The fan is set to auto and I ticked the "synchronize fan speeds" button. I added "fan speed" and "fan speed 2" to the monitor, and it looks like it's working. I could be totally wrong since I'm just looking at the fans and eyeballing it, but it looks like all 3 fans are changing speed properly.

There is a slight delay on the 3rd fan when using Afterburner, but you are right, they seem to work. For me, this only happens when running both X1 and Afterburner.

However, when fiddling with X1, and then afterburner, and then going back and fourth between programs, it seems that the fan system in afterburner eventually (if not immediately) does not control the 3rd fan.

Also, the third fan does not run period when just using Afterburner alone.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
After further testing on my rig, I was able to find the card is very efficient around 285w, may actually switch it it full time and take the slight performance hit of 2% from my current profile.

700-725mV clocks between 1585-1605 (can be made constant, but *card is instable closer to 700mV, driver crashes)
725-755mV clocks between 1700-1740 (can be made constant) (very stable)
812mV clocks at 1785 (consistant clock) (my current 24/7 profile)
875mV clocks between 1800-1860, averages 1830 (can't get a stable consistant clock)
925mV clocks between 1800-1935, (least stable clock range so far in terms of consistancy, I've done the least amount of testing here)

Shadow of the Tomb Raider (4K/High). GPU Speed is according to what I have it set in afterburner.
Stock EVGA XC3 Ultra: 109 FPS
725mV - 1605 Core: 108FPS
755mV - 1725 Core: 112FPS
812mV - 1785 Core: 116FPS
875mV - 1840 Core: 119FPS
925mV - 1860 Core: 121FPS


TimeSpy Free (Graphics Score)
:
Stock EVGA XC3 Ultra: 18391
725mV - 1605 Core: 18385
755mV - 1725 Core: 18813
812mV - 1785 Core: 19538
875mV - 1840 Core: 20379
925mV - 1860 Core: didn't run, went to bed, don't really care much honestly

Hopefully this helps someone who is testing ranges on their card.
could you screenshot your stock curve, taken after a period of idling, not gaming? It would help me :)
 

spine

2[H]4U
Joined
Feb 4, 2003
Messages
2,650
If you don't know what you are talking about its best to just listen. The 3080 series cards have a heavy handed power curve that in alot of cases are unecessary. This is one of my cards in the picture below the score on the right is from "undervolting", its actually setting the proper curve for my chip. It runs 70 watts less, 5-7 degress C cooler and performs the same or better overall. If this design wasn't such shit on Samsung's process this wouldn't be happening but Nvidia shoved turds out the door. This was something AMD cards needed before this gen.

View attachment 295861
You're just temp limited. Ok, so this is just for the non-watercooling peeps, fair enough. If you can't stand to turn your fans up, then watercool. Don't shy away from the heat! Control it! And then add more fuel and more cooling and don't stop till it explodes. :smuggrin:
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
Original post has been updated with the guide I put together. Thanks to DFenz for the data he provided and everyone else who contributed information. If you have questions or want to discuss results, please share!
 
  • Like
Reactions: DFenz
like this

exlink

Supreme [H]ardness
Joined
Dec 16, 2006
Messages
5,431
Interesting...so it seems like a 3090 can essentially run around 300w with minimal to no performance loss. That’s the same as the 6900XT.

Either Nvidia rushed these cards out so quickly that their power curve testing is basically unrefined. Or they have the cards run at 350w because the binning is so poor on some chips that they need all that extra power to run stable.

Great job btw OP!
 
Last edited:
  • Like
Reactions: DFenz
like this

VirtualMirage

Limp Gawd
Joined
Nov 29, 2011
Messages
384
Original post has been updated with the guide I put together. Thanks to DFenz for the data he provided and everyone else who contributed information. If you have questions or want to discuss results, please share!
Nice write up and a lot easier to understand than some of the other write ups I have seen.

Did you happen to keep track of what your average temps and fan speeds were too when collecting those numbers? It might be helpful for those that might be putting a higher priority on temperature or noise versus power consumption. Of course, ambient temps and case environment will vary from person to person, but it might still offer a good comparison.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
Interesting...so it seems like a 3090 can essentially run around 300w with minimal to no performance loss. That’s the same as the 6900XT.

Either Nvidia rushed these cards out so quickly that their power curve testing is basically unrefined. Or they have the cards run at 350w because the binning is so poor on some chips that they need all that extra power to run stable.

Great job btw OP!
if you set 85% power limit, you would be right around 300w and probably 99% of stock across the board.
Nice write up and a lot easier to understand than some of the other write ups I have seen.

Did you happen to keep track of what your average temps and fan speeds were too when collecting those numbers? It might be helpful for those that might be putting a higher priority on temperature or noise versus power consumption. Of course, ambient temps and case environment will vary from person to person, but it might still offer a good comparison.
I didn't track temps or fan speed, I let the card manage fan speed (not afterburner). That being said, I don't think I saw over 69C, ever. It hung out in the low 60s for most of my testing, 20C ambient.
 

VirtualMirage

Limp Gawd
Joined
Nov 29, 2011
Messages
384
if you set 85% power limit, you would be right around 300w and probably 99% of stock across the board.

I didn't track temps or fan speed, I let the card manage fan speed (not afterburner). That being said, I don't think I saw over 69C, ever. It hung out in the low 60s for most of my testing, 20C ambient.
20C? Jeez, do you have that thing in an ice box? :ROFLMAO:

My room ambient is typically between 22-25C. I MIGHT get down to 20-21C at night when everyone is in bed during mid-winter, but then nobody is using the PC.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
7,620
20C? Jeez, do you have that thing in an ice box? :ROFLMAO:

My room ambient is typically between 22-25C. I MIGHT get down to 20-21C at night when everyone is in bed during mid-winter, but then nobody is using the PC.
20C = 68F, which is pretty comfortable for me in the fall/winter. It's gotten up to about 72F ~22C this week. I sleep better when it's cold :)
 
Top