Overclocking My GTX 1060 - What's 'Safe' RPMs?

[jiyeon]

Hi, so I've just dipped my feet into the waters of GPU overclocking after trying it out in PC Building Simulator first, and so far I've managed to get this profile.

 

My current fan curve:

Spoiler

 

 

As I was setting my fan curve and was playing Overwatch, I happened to notice my card peaked 68°C when it would usually do around 73-76°C at max, so I was well impressed.

... That is until I saw it was doing 4.7K RPM!

 

I was wondering, is there a certain safe RPM range for my card? Without MSI Afterburner, it will max out at 2.7K RPM so seeing that 4.7K RPM was a shock. I couldn't hear the card while I was playing because I have headphones in so noise is not a concern.

 

Additionally, are there any modifications I could do to my fan curve profile? I have it relatively steep due to my aforementioned concerns with 'safe' RPM ranges.

[Hiya!]

There is no such thing as safe rpm.

 

But the faster it is the faster your fan bearing will degrade usually the lubrication..

[knightslugger]

Why didn't you crank the power limit...? max fan speed is fine anyway.

[Tristerin]

FANS are rated at their MAX RPM for the LIFE of the fan (I.E. 50,000 hours of whatever).  So running the fan faster does NOT kill the fan faster.  

 

As long as you are comfortable with the audible level, do it!

[jiyeon]

1 minute ago, knightslugger said:

Why didn't you crank the power limit...? max fan speed is fine anyway.

I wanted to dip my feet first before cranking the power limit there. I'm still not 100% sure as to what the power limit is so I'm playing it safe for now.

 

Just now, Tristerin said:

FANS are rated at their MAX RPM for the LIFE of the fan (I.E. 50,000 hours of whatever).  So running the fan faster does NOT kill the fan faster.  

Oh right. So I can in theory just ramp my fan curve up to 100% at say around 70°C (that's where I start to feel uncomfortable) and the fans will be fine in the long run?

[Tristerin]

Just now, seoz said:

I wanted to dip my feet first before cranking the power limit there. I'm still not 100% sure as to what the power limit is so I'm playing it safe for now.

 

Oh right. So I can in theory just ramp my fan curve up to 100% at say around 70°C (that's where I start to feel uncomfortable) and the fans will be fine in the long run?

Yes - my fan curve is that lol - 40% until 69c then straight to 100% @70c  (my card runs 48c average its the just in case something happens and it gets hot I want to be alerted at that temp)

[knightslugger]

1 minute ago, seoz said:

I wanted to dip my feet first before cranking the power limit there. I'm still not 100% sure as to what the power limit is so I'm playing it safe for now.

 

it wont hurt it, believe me. under hard stress, it wants more power. By pulling back the reigns, the GPU will respond by down-clocking.

 

crank it.

[jiyeon]

Just now, Tristerin said:

Yes - my fan curve is that lol - 40% until 69c then straight to 100% @70c  (my card runs 48c average its the just in case something happens and it gets hot I want to be alerted at that temp)

Ah awesome, thank you for your insight!

 

Just now, knightslugger said:

it wont hurt it, believe me. under hard stress, it wants more power. By pulling back the reigns, the GPU will respond by down-clocking.

Do you think I might incur any sort of electricity bill increase, heat output, or any other caveats by cranking it all the way up?

[Mira Yurizaki]

Just now, seoz said:

Do you think I might incur any sort of electricity bill increase, heat output, or any other caveats by cranking it all the way up?

The fan consumes a negligible amount of power compared to the video card.

[Wh0_Am_1]

7 minutes ago, Hiya! said:

There is no such thing as safe rpm.

 

But the faster it is the faster your fan bearing will degrade usually the lubrication..

Actually fans prefer higher RPMs (within rating), due to the fact that more air is moving around the motor assembly, and the electromagnets are on contact for a shorter period of time leading to a cooler motor, and less evaporation of the lubricant.

[jiyeon]

1 minute ago, Mira Yurizaki said:

The fan consumes a negligible amount of power compared to the video card.

I was referring to the power limit being cranked up, not the fan RPMs.

[Wh0_Am_1]

-Redacted-

[knightslugger]

3 minutes ago, seoz said:

Do you think I might incur any sort of electricity bill increase, heat output, or any other caveats by cranking it all the way up?

on a 1060? if i'm being honest, yes you will have a higher electricity bill but because you're using more power. How much?

 

...about an LED lightbulb's (or two) worth...

[Mira Yurizaki]

4 minutes ago, seoz said:

I was referring to the power limit being cranked up, not the fan RPMs.

Well yes, if you're allowing the card to consume more power, then it's going to consume more electricity to get to the new limit.

[Wh0_Am_1]

6 minutes ago, seoz said:

I was referring to the power limit being cranked up, not the fan RPMs.

At most $100 over the life of the card (assuming that you have it running at max power for the duration ~8 years 24/7) other wise with a normal use case ~ maybe $5 what is bigger is the wear and tear on the card, turning put the max power could reduce the life expectancy from like 10 years to 7-8 years in  more extreme cases.

[jiyeon]

2 minutes ago, knightslugger said:

on a 1060? if i'm being honest, yes you will have a higher electricity bill but because you're using more power. How much?

 

...about an LED lightbulb's (or two) worth...

Just now, Mira Yurizaki said:

Well yes, if you're allowing the card to consume more power, then it's going to consume more electricity to get to the new limit.

Oh right, interesting, I've just had an epiphany.

 

I've already reached my limit on my core clock - which is +200MHz, any further and it will cause artifacting - and so, if I do crank it up to the max that it allows, in this case, it's 116%, I should in theory be able to go higher to say +300MHz on the core clock, right?

 

I'm yet to truly push my memory clock at the moment.

[Wh0_Am_1]

10 minutes ago, seoz said:

Oh right, interesting, I've just had an epiphany.

 

I've already reached my limit on my core clock - which is +200MHz, any further and it will cause artifacting - and so, if I do crank it up to the max that it allows, in this case, it's 116%, I should in theory be able to go higher to say +300MHz on the core clock, right?

 

Not necessarily, this just removes the wattage limitations allowing for it to maintain the clock at slightly higher clocks, (and potentially higher clocks if their is a wattage limitation) to really stretch your processor, you will need to mess with voltages (which is dangerous and could reduce the life span of the card or outright kill it.) but generally this just allows for your processor to maintain higher clock for longer.

[Hiya!]

20 minutes ago, Wh0_Am_1 said:

electromagnets are on contact for a shorter period of time leading to a cooler motor, and less evaporation of the lubricant.

Huh? Yes it does but it will be more frequent like 100x or whatever depending on the speed and we know that friction creates heat so even though it on contact less it is more frequent. The type of bearing also comes to play..

[Jurrunio]

18 minutes ago, seoz said:

I've already reached my limit on my core clock - which is +200MHz, any further and it will cause artifacting - and so, if I do crank it up to the max that it allows, in this case, it's 116%, I should in theory be able to go higher to say +300MHz on the core clock, right?

Depends. If your GPU can't do higher because it's not allowed to pull more current (i.e. power limit) then you can probably do like 220-230MHz offset. If it's artifacting because of not enough voltage for this frequency then you can't do much about it, aside from swapping the cooler to push temps even lower for better stability.

[Wh0_Am_1]

Just now, Hiya! said:

Huh? Yes it does but it will be more frequent like 100x or whatever depending on the speed and we know that friction creates heat so even though it on contact less it is more frequent. The type of bearing also comes to play..

Indeed by generally on fans despite the increase in friction, with the exception of a complete stop, they run hotter on lower RPM due to the reduced efficiency on the power to cooling ratio, thereby despite the fact that you need more power for higher RPMs, you get a better return on your investment for that power.

[DildorTheDecent]

14 minutes ago, seoz said:

I've already reached my limit on my core clock - which is +200MHz, any further and it will cause artifacting - and so, if I do crank it up to the max that it allows, in this case, it's 116%, I should in theory be able to go higher to say +300MHz on the core clock, right?

You need to get the card cooler if you want to push past 2100MHz. Simply going "more power/volt" doesn't really work the with Pascal.

 

Get it down into the mid 50Cs or 60C wouldn't be so bad. Alternatively, find the temperature point where you card starts to reduce the clock and try and get underneath it. Whenever the card is that little bit cooler the extra 20Ws you'll get from increasing the power target might help. "Might" because most people just crank the slider and try to focus on the voltage for clock/heat reasons. You're still within the limits (that don't mean anything at all really) that your GPU manufacturer has set.

[Jurrunio]

24 minutes ago, DildorTheDecent said:

You need to get the card cooler if you want to push past 2100MHz. Simply going "more power/volt" doesn't really work the with Pascal.

 

Get it down into the mid 50Cs or 60C wouldn't be so bad. Alternatively, find the temperature point where you card starts to reduce the clock and try and get underneath it. Whenever the card is that little bit cooler the extra 20Ws you'll get from increasing the power target might help. "Might" because most people just crank the slider and try to focus on the voltage for clock/heat reasons. You're still within the limits (that don't mean anything at all really) that your GPU manufacturer has set.

der8auer did push a 3 phase VRM 1060 to 2200MHz with its stock cooler... of course the power limit is gone and there's manual voltage control to push past Nvidia's limits

[jiyeon]

I thank you all ever so much for the advice regarding GPU overclocking as well as fan RPM speeds.

 

I've settled for this profile for tonight, I managed to get +200MHz stable without artifacts (at least from the 20 minutes I played CSGO) and bumped up the power limit a bit.

 

I also re-adjusted my fan curve to be super aggressive and now my GPU maxes out at 65°C. I could go for an even more aggressive fan curve in the near future but this will do for now.

 

Again, thank you all so much for helping a noob overclock her GTX 1060 over the course of 2 hours.