bitxan

Hello, after all this testing time, what is the best combination you have gotten of frequency and voltage for undervolting in 1080 ti?

Hello, who has been testing and who has this data for you?

bitxan

at the moment, at 1898 Mhz with 0,931V I get the same performance as at stock OC speeds with a 1080 ti ftw3 with a consumption reduction of 30w and 6 degrees of temperature, the benefit of undervolting is very interesting

Same performance at what? It is well known that using a voltage curve with either Precision or Afterburner for example, you can gain some efficiency over the stock BIOS.  I assume you're asking for mining purposes.

bitxan
I do not use it for mining, I was surprised that 1898 mhz and 0.931volt more performance in, time spy,fire strike,superposition,valley, Heaven,and consumes 30w less cold and therefore more silent synod, I only want to know how they have the rest of peers with respect to the stock velocities.

With a 1080 Ti very few are interested in efficiency so your pool is probably going to be limited to volunteers and miners.  I use the voltage curve and I'm running 2037mhz at 1.000v versus the typical 1.031-1.042 It is usually at. It shaves a few degrees off of the temp under load.

mark_thaddeus

HeavyHemi
With a 1080 Ti very few are interested in efficiency so your pool is probably going to be limited to volunteers and miners.  I use the voltage curve and I'm running 2037mhz at 1.000v versus the typical 1.031-1.042 It is usually at. It shaves a few degrees off of the temp under load.

What software are you using for the voltage curve, MSI AB?

Its gotta be.. its broke and has been on Vegas precision x oc for a long time now.

Never considered under volt on a performance card like the 1080ti. I do over volt though and temps are fine. (Carefully monitored when creating a profile with over volt!!)
 
A media center GPU may be good to under volt. But EVGA cards have decent warranty so if you are doing this to prolong the cards life, I suspect you have bought a new GPU any
hows. Could last a decade! I OC and expect 3 good years min - up to 5 years. But kind of moot as I upgrade almost every gen. Is that your reason for under volt?  Nothing wrong with it at all, but most do the opposite. Kind of curious why you under volt?

bitxan

when doing undervolt, I achieve lower temperatures and therefore a lower noise and at the same time the performance is higher and more stable.

Gawg36
Never considered under volt on a performance card like the 1080ti. I do over volt though and temps are fine. (Carefully monitored when creating a profile with over volt!!)
 
A media center GPU may be good to under volt. But EVGA cards have decent warranty so if you are doing this to prolong the cards life, I suspect you have bought a new GPU any
hows. Could last a decade! I OC and expect 3 good years min - up to 5 years. But kind of moot as I upgrade almost every gen. Is that your reason for under volt?  Nothing wrong with it at all, but most do the opposite. Kind of curious why you under volt?

when doing undervolt, I achieve lower temperatures and therefore a lower noise and at the same time the performance is higher and more stable.

Thanks for the answer. Totally logical. I think you have some good silicon there! If I did that on my card (ftw3 1080ti icx) It would be unstable. Even at stock really heavy games
like Doom maxed out run perfectly with the voltage slider up by 25% and a small OC. Had just a few crashes without adding extra voltage. Not complaining, you just did much better that me in the old silicon lottery. I guess I'd do the same if I could!

Sajin
From my testing undervolting at the same frequency causes a reduction in performance.

 

with the 5 tests that I have commented? what speed marks you when you start the tests and after heating which is constant at 4k?

Sajin
From my testing undervolting at the same frequency causes a reduction in performance.

That wouldn't be at the same frequency though. Because as you say, the frequency dropped causing the performance drop. So it would be user error in not using a voltage high enough to maintain the clock they wanted IMO.

Sajin

HeavyHemi

Sajin
From my testing undervolting at the same frequency causes a reduction in performance.

That wouldn't be at the same frequency though. Because as you say, the frequency dropped causing the performance drop. So it would be user error in not using a voltage high enough to maintain the clock they wanted IMO.

I started at the same frequency of 2012 MHz. Undervolting to 1.0v caused gpu boost 3.0 to pull mhz from the card because it was undervolted. That isn't user error. That is a result of undervolting. Using an even lower value than 1.0v will cause gpu boost 3.0 to pull even more mhz.

Yes, you're repeating what I said.  Who is arguing that if you keep decreasing the voltage your clocks will not drop? What is being discussed here is you can, using the the voltage curve, run the GPU at a lower voltage at the same clocks than GPU Boost uses automatically. That's all. Everyone knows if you decrease the voltage the core clocks will drop the same for a CPU. You have to find the right one for a particular GPU. In my case, I can run at 2037 at 1.0v  100% stable using the voltage curve.  GPU Boost, uses either 1.037 or higher depending upon temp. As I said, it drops the temp a couple of C and I run fans a bit slower and quieter. What's the drawback?

Sajin

HeavyHemi

Sajin

HeavyHemi

Sajin
From my testing undervolting at the same frequency causes a reduction in performance.

That wouldn't be at the same frequency though. Because as you say, the frequency dropped causing the performance drop. So it would be user error in not using a voltage high enough to maintain the clock they wanted IMO.

I started at the same frequency of 2012 MHz. Undervolting to 1.0v caused gpu boost 3.0 to pull mhz from the card because it was undervolted. That isn't user error. That is a result of undervolting. Using an even lower value than 1.0v will cause gpu boost 3.0 to pull even more mhz.

Yes, you're repeating what I said.  Who is arguing that if you keep decreasing the voltage your clocks will not drop? What is being discussed here is you can, using the the voltage curve, run the GPU at a lower voltage at the same clocks than GPU Boost uses automatically. That's all. Everyone knows if you decrease the voltage the core clocks will drop the same for a CPU. You have to find the right one for a particular GPU. In my case, I can run at 2037 at 1.0v  100% stable using the voltage curve.  GPU Boost, uses either 1.037 or higher depending upon temp. As I said, it drops the temp a couple of C and I run fans a bit slower and quieter. What's the drawback?

The OP is arguing it...

bitxan

at the moment, at 1898 Mhz with 0,931V I get the same performance as at stock OC speeds with a 1080 ti ftw3 with a consumption reduction of 30w and 6 degrees of temperature, the benefit of undervolting is very interesting

The drawback is you're losing performance because you've chose to run a lower voltage.

No he's not arguing that. He's saying that he can get an even further reduction of the voltage versus what the GPU boosts too at stock using a lower voltage. He's not trying for a higher clock. He's trying for the same stock boost clocks at a lower voltage. That's all.

HeavyHemi
That wouldn't be at the same frequency though.

No problem. Here are some true undervolts/overvolts using the classified tool @ the same frequency...
 
Stock:

 
Undervolt to 1.0v using classified tool:

 
Overvolt to 1.093v using classified tool:

 

 
 
 
 

My stock boost clock is 2012 MHz. Running 2037 @ 1.0v isn't going to give you the same performance as running 2037 @ 1.062v. You may not be losing any mhz, but you are definitely losing performance from undervolting.

HeavyHemi
BTW, if we accept your logic, then running at 2.0 volts at 2037 would autogmagically give us more performance right?

Pretty sure you already know the answer to that question. So I'll just hold my comment.

Sajin

HeavyHemi
BTW, if we accept your logic, then running at 2.0 volts at 2037 would autogmagically give us more performance right?

Pretty sure you already know the answer to that question. So I'll just hold my comment.

 
 Well you have no choice considering the flaw in your logic. This is extremely simple: Using the voltage curve you can cause the GPU to use a lower voltage at the same frequency that GPU Boost selects automatically. This does not cause a performance 'loss' for OUR PURPOSES. We don't care about your reasons of pure performance. We're looking for the same performance with a bit lower temps by altering how the GPU clocks and at what voltage. Running 2ghz at 1.v is exactly the same as running 2ghz at 1.093 volts. I get what you're arguing: if we are limiting the voltage via the curve OF COURSE WE ARE ALSO LIMITING THE BOOST CLOCK. This is ON PURPOSE. The purpose is to use the voltage curve to find a voltage and frequency more efficient than the stock on.
Man...this WCCF style infiltration is getting toxic.
 

HeavyHemi
Running 2ghz at 1.v is exactly the same as running 2ghz at 1.093 volts.

So you're saying my examples above are fake?