Question Should I raise my min frequency to a 100 below max or leave it at 500 MHz?

rbogomolec

Honorable
Nov 16, 2017
97
10
10,545
I've recently got me a RX 6800 XT. My max frequency in Adrenalin is 2494MHz. My min frequency is set to the default 500MHz. My undervolt setting is currently at 1050mVa. Everything runs fine and smooth, playing Shadow of Tomb Raider on 4k, high settings, with constant 60 fps, no framedrops or anything. I don't intend to play on ultra settings nor to go above 60fps. Should I then care about setting my min frequency to 100MHz below my max frequency? Or is all that I will get by doing so more power consumption, higher temperatures and louder fans? Which option is better for my GPU's life span? Is it more stressful for the GPU to work with occasional MHz drops, or is it more stressful to keep a constant frequency but at the cost of drawing more power and running hotter? I'd say that more power and higher temps are worse than oscillating frequencies. Am I right?
 

Ralston18

Titan
Moderator
Remember:

"If it ain't broke, then don't fix it."

Leave well enough alone until there is some specific requirement to make some change(s).

And then, as warranted, make only one change at a time.

Continue to monitor your build so you are familar with temperatures, wattage used, etc..

Lastly, as always, remember to regularly backup all important data at least 2 x to locations off of the current host. Verify that the backups are both recoverable and readable.
 

Karadjgne

Titan
Ambassador
Minimum speed is essentially 'idle' speeds, which is low power consumption. Raising that speed to just below maximum means the card will always be in high power consumption, and higher power = higher temps.

There's no need to be constantly pushing a gpu to max or close, 2d apps like windows, Office, mail or anything else other than games or other 3d apps do not use but a fraction of a gpus potential. That's why many cards have adopted the 'fanless' option for temps below @ 65°C, no need for fan use, not enough power use to drive temps beyond what the cooler can do passively.
 
  • Like
Reactions: rbogomolec
I've recently got me a RX 6800 XT. My max frequency in Adrenalin is 2494MHz. My min frequency is set to the default 500MHz. My undervolt setting is currently at 1050mVa. Everything runs fine and smooth, playing Shadow of Tomb Raider on 4k, high settings, with constant 60 fps, no framedrops or anything. I don't intend to play on ultra settings nor to go above 60fps. Should I then care about setting my min frequency to 100MHz below my max frequency? Or is all that I will get by doing so more power consumption, higher temperatures and louder fans? Which option is better for my GPU's life span? Is it more stressful for the GPU to work with occasional MHz drops, or is it more stressful to keep a constant frequency but at the cost of drawing more power and running hotter? I'd say that more power and higher temps are worse than oscillating frequencies. Am I right?
You might benefit from watching GamersNexus' overclocking sessions with RX 6800XT's. But if you don't want to watch one point they made is a minimum frequency setting of 250-150MHz under the maximum frequency setting usually returned the highest overclocks in their setups. They were obviously using drivers available at the time and AMD's made significant driver improvements so it may work a bit differently now.

Also, don't get excited about what they achieved and assume you should do the same. They were pushing things to demonstrate "ultimate limits" which are never practical for 24/7. In particular, they were looking for maximum performance 3DMark runs which are relative short duration and with a cool-down between. And they were also working with fans maxed on an open-air bench.

What I've found with my RX 6800 XT (an MSI Gaming Z Trio) is under-volting with, at most, a mild overclock gets similar or better results than a much higher overclock which also requires a higher voltage for stability in actual use for long gaming sessions. That's pretty typical of modern thermally constrained GPU's. And as a bonus it's probably best for assuring a long useful life since it not only returns a decent performance uplift but runs cooler with quieter fans and, obviously, at a lower voltage. Both temp and voltage are primary stressors affecting semiconductor useful life.
 
Last edited:

rbogomolec

Honorable
Nov 16, 2017
97
10
10,545
You might benefit from watching GamersNexus' overclocking sessions with RX 6800XT's. But if you don't want to watch one point they made is a minimum frequency setting of 250-150MHz under the maximum frequency setting usually returned the highest overclocks in their setups. They were obviously using drivers available at the time and AMD's made significant driver improvements so it may work a bit differently now.

Also, don't get excited about what they achieved and assume you should do the same. They were pushing things to demonstrate "ultimate limits" which are never practical for 24/7. In particular, they were looking for maximum performance 3DMark runs which are relative short duration and with a cool-down between. And they were also working with fans maxed on an open-air bench.

What I've found with my RX 6800 XT (an MSI Gaming Z Trio) is under-volting with, at most, a mild overclock gets similar or better results than a much higher overclock which also requires a higher voltage for stability in actual use for long gaming sessions. That's pretty typical of modern thermally constrained GPU's. And as a bonus it's probably best for assuring a long useful life since it not only returns a decent performance uplift but runs cooler with quieter fans and, obviously, at a lower voltage. Both temp and voltage are primary stressors affecting semiconductor useful life.
Tnx man! Yeah I also went for udervolt and slight overclock in the end. I don't want my GPU to run an 90-100% all the time just to prevent some random 3-4fps framedrops that happen maybe once per hour. Also, I never play on ultra settings, just on high, and I don't really need to overclock this beast of a GPU for that.
 
Tnx man! Yeah I also went for udervolt and slight overclock in the end. I don't want my GPU to run an 90-100% all the time just to prevent some random 3-4fps framedrops that happen maybe once per hour. Also, I never play on ultra settings, just on high, and I don't really need to overclock this beast of a GPU for that.
Undervolting is something you can push pretty much as far as it will take it since it is safe for GPU. And these things seem like you can push it down forever: I've run through Port Royal benchmarks at 1.056 volts and gotten terrific results while doing nothing else.

But I also found that it can't be left with such low voltage without some strange problems creeping in, although it may be a quirk of my motherboard or CPU to be honest. If I try to run 24/7 with any voltage setting below about 1.087V I'll get some serious latency issues (stuttering videos and gaming, audio pops, clicks and crackling) after a cold start-up. Strangely, a warm re-start generally clears it up. It may be peculiarities in my motherboard and CPU, but I'm running with a setting of 1.093V so none of that happens and it still makes a significant improvement in both performance and temperature during a TimeSpy stress test so I'm happy with that.

Also, something I've found is pushing up the power limit has as much negative as positive results. In the near term, it can help with a 3DMark BM run but in the long term it will heat up the board a lot more. Since these things are temperature constrained that leads to occasional performance dips to cool down in long gaming sessions. So tuning power limits sort of goes hand in hand with GPU cooling: a well ventilated case and/or high (noisy) fan speeds allow a higher limit.
 
Last edited:
  • Like
Reactions: rbogomolec

rbogomolec

Honorable
Nov 16, 2017
97
10
10,545
Undervolting is something you can push pretty much as far as it will take it since it is safe for GPU. And these things seem like you can push it down forever: I've run through Port Royal benchmarks at 1.056 volts and gotten terrific results while doing nothing else.

But I also found that it can't be left with such low voltage without some strange problems creeping in, although it may be a quirk of my motherboard or CPU to be honest. If I try to run 24/7 with any voltage setting below about 1.087V I'll get some serious latency issues (stuttering videos and gaming, audio pops, clicks and crackling) after a cold start-up. Strangely, a warm re-start generally clears it up. It may be peculiarities in my motherboard and CPU, but I'm running with a setting of 1.093V so none of that happens and it still makes a significant improvement in both performance and temperature during a TimeSpy stress test so I'm happy with that.

Also, something I've found is pushing up the power limit has as much negative as positive results. In the near term, it can help with a 3DMark BM run but in the long term it can let the board heat up a lot more. Since these things are temperature constrained that can lead to occasional performance dips in long gaming sessions. So tuning power limits sort of goes hand in hand with GPU cooling: a well ventilated case and/or high (noisy) fan speeds allow a higher limit.
I'm currently at 1.020V undervolt and it still runs smooth. Mine is a Sapphire Nitro+. I did some other tweaks though as well and capped my fps at 60 since I don't need more. It could be that that's also a reason why it can run on lower voltages.
 
I'm currently at 1.020V undervolt and it still runs smooth. Mine is a Sapphire Nitro+. I did some other tweaks though as well and capped my fps at 60 since I don't need more. It could be that that's also a reason why it can run on lower voltages.
I thought AMD had limited these things to 1.050V at the lowest. I wonder if Sapphire did something to allow more. Maybe I should try it on mine just to see...it could have been a driver limitation that's been removed.

Have you done anything with the power limit setting?

Games generally worked well at very low voltages, but if I also ran with an overclock a 3DMark BM would likely crash. It's the cold boots that throw it into latency issues where things don't work well.

I also try to cap games at around 60FPS. Triple A titles are all that I play, it's the e-sports titles where super high fps supposedly helps and I just don't like that kind of game.