GTX 980 ti vs 980 ti FTW power useage - stats

[xactoman]

So I know this is a very hot topic here on the forum. How much power does a 980 card draw??....

I have a mac pro hex core (3.46) with both OSX and windows on it. Recently upgraded my GTX680 to a GTX 980ti FTW. I have access to a regular 980ti that I use on a work machine and thought it would be a good comparison to see what each one draws. I used iStat monitor to watch amp draw and then converted those numbers to watts pulled. This is probably not the most accurate method, but it at least gives information as to what is going on.

First, the 980ti. This card is both a 6 pin and a 8 pin, rated at 250w max. I used an 8 pin adapter and both power leads are being pulled from the logic board. This card has been connected and running this way (on work computer) for the past year and a half with no trouble. I used Unigine Valley for the test load. Ultra settings at 2560x1600.
PCI = 50.4 watts
Boost 2 = 56.4 watts
Boost 1 = 94.8 watts.
total: 201.6 watts
As you can see boost 1 pulls a massive amount of power, well over the rated spec of 75w. The load certainly is not balanced. Card or computer does not run hot, no artifacts or problems. I have had one shutdown in windows and I attribute that to boost 1 pulling over 100w and the power supply just shut off. Card is not reaching the 250w max and not sure what you could do to reach that. maybe 4k?

980ti FTW. This card has 2, 8-pin connectors and is rated at 275w max. After seeing the results of the first card, I decided to hook this up with extra power. I used an 8-pin to SATA Y-adapter and ran that off both leads in the optical drive bay. I then used the two booster PCI leads and joined them with a 6-pin to 8-pin Y adapter. The results are interesting. Loaded up Unigine valley at the same Ultra 2560x1600 settings.
PCI = 51.12 watts
Boost 2 = 51 watts
Boost 1 = 38.4 watts
Optical = 31.92 watts
total: 172.44 watts
The FTW card has a much higher clock speed and higher max power, but pulled considerably less power. I was surprised. settings were the same, FPS results were the same. Power across the machine was being pulled in a more reasonable manner and well within the specs of the connections

It's interesting to see that the PCI slot does not pull 75w as rated and hovers around the 50w mark.

Now, iStat monitor could be full of crap and these readings way off. If that is the case, who knows what this thing is doing. Either way, both setups work, just that the first pulling that much power from the board is somewhat unreliable. It seems that the power supply protection switches the machine off before something bad happens to the board.

I wish there was an app in windows that showed amp draw like iStat. That would be a helpful test to run. So far, have not found one that reads the mac sensors.

 

[joot]

So I know this is a very hot topic here on the forum. How much power does a 980 card draw??....

I have a mac pro hex core (3.46) with both OSX and windows on it. Recently upgraded my GTX680 to a GTX 980ti FTW. I have access to a regular 980ti that I use on a work machine and thought it would be a good comparison to see what each one draws. I used iStat monitor to watch amp draw and then converted those numbers to watts pulled. This is probably not the most accurate method, but it at least gives information as to what is going on.

First, the 980ti. This card is both a 6 pin and a 8 pin, rated at 250w max. I used an 8 pin adapter and both power leads are being pulled from the logic board. This card has been connected and running this way (on work computer) for the past year and a half with no trouble. I used Unigine Valley for the test load. Ultra settings at 2560x1600.
PCI = 50.4 watts
Boost 2 = 56.4 watts
Boost 1 = 94.8 watts.
total: 201.6 watts
As you can see boost 1 pulls a massive amount of power, well over the rated spec of 75w. The load certainly is not balanced. Card or computer does not run hot, no artifacts or problems. I have had one shutdown in windows and I attribute that to boost 1 pulling over 100w and the power supply just shut off. Card is not reaching the 250w max and not sure what you could do to reach that. maybe 4k?

980ti FTW. This card has 2, 8-pin connectors and is rated at 275w max. After seeing the results of the first card, I decided to hook this up with extra power. I used an 8-pin to SATA Y-adapter and ran that off both leads in the optical drive bay. I then used the two booster PCI leads and joined them with a 6-pin to 8-pin Y adapter. The results are interesting. Loaded up Unigine valley at the same Ultra 2560x1600 settings.
PCI = 51.12 watts
Boost 2 = 51 watts
Boost 1 = 38.4 watts
Optical = 31.92 watts
total: 172.44 watts
The FTW card has a much higher clock speed and higher max power, but pulled considerably less power. I was surprised. settings were the same, FPS results were the same. Power across the machine was being pulled in a more reasonable manner and well within the specs of the connections

It's interesting to see that the PCI slot does not pull 75w as rated and hovers around the 50w mark.

Now, iStat monitor could be full of crap and these readings way off. If that is the case, who knows what this thing is doing. Either way, both setups work, just that the first pulling that much power from the board is somewhat unreliable. It seems that the power supply protection switches the machine off before something bad happens to the board.

I wish there was an app in windows that showed amp draw like iStat. That would be a helpful test to run. So far, have not found one that reads the mac sensors.

Are you using this card for gaming or Adobe products? I like to know how it performs for Premiere (CUDA)...

 

[mmomega]

I have 2 980Ti SC+ cards overclocked and my power supply measures it's own power usage, kind of.
Just the moment with both cards overclocked running at 92+% and an i7 6700k running at 98-100% I'm running at an average of 600W.
Minus the CPU It's averaging 500W.
Both GPU clocks are at 1430MHz
GPU#1 270'ish Watts. 44-48C
GPU#2 340'ish Watts. 44-48C

Both cards are exactly the same.
GPU#2 that pulls more power also can be over clocked much more than #1. 1430 just seems to be a good stable clock for both cards at the same time.

This computer isn't a MP but I am hoping I can help get some kind of information for you.
I need to check the power draw again later 2 more seperate times and take the average of those tests.
I just have a Corsair HX850i Power Supply that plugs into the motherboard via USB cable and an app reads the information. My computer has just been Folding so it keeps the CPU and GPU's pegged out rather good so I just took reading during that while stopping one card while letting another continue to fold.

 

[ActionableMango]

The FTW card has a much higher clock speed and higher max power, but pulled considerably less power. I was surprised.

MVC saw much higher power draw from the 980ti than even the TitanX. Speculation at the time was that the top end chips are binned based on power usage, with the efficient chips becoming TitanX cards and the power hungry chips downclocked to be 980ti because they required too much power at TitanX clocks.

I could also speculate that your 980ti FTW model is a newer production chip than your 980ti, and that Nvidia's manufacturing processes have improved over time as would be expected.

 

[xactoman]

Are you using this card for gaming or Adobe products? I like to know how it performs for Premiere (CUDA)...

At home I use the Ti FTW for gaming and at work I mainly use the Ti for rendering in Blender. It accelerates creative suite very nicely, but I don't know how it performs in After Effects or Premier as I dont use those. Comparatively, (using Blender) the Ti renders out twice as fast as the (CPU) 12 core (24 thread) mac pro I am using, so it's a powerhouse.

 

[Asgorath]

As you can see boost 1 pulls a massive amount of power, well over the rated spec of 75w. The load certainly is not balanced.

Just to avoid confusion, you're talking about the spec of a 6-pin connector, right? It's totally expected that the majority of the power draw in a PCIe + 6-pin + 8-pin config would be through the 8-pin connector. The load is about as balanced as I would've expected: both 75W connectors are drawing 50W (i.e. PCIe slot and 6-pin), while the 150W connector is drawing 95W (i.e. the 8-pin). So, all connectors are providing about 2/3 of their capacity.

The fact you're running this on a system that doesn't natively support the 8-pin connector is the only issue here, which is why we generally suggest to not do that (esp. for cards that really need 250W+ like this one).

 

[xactoman]

yes,

Just to avoid confusion, you're talking about the spec of a 6-pin connector, right? It's totally expected that the majority of the power draw in a PCIe + 6-pin + 8-pin config would be through the 8-pin connector. The load is about as balanced as I would've expected: both 75W connectors are drawing 50W (i.e. PCIe slot and 6-pin), while the 150W connector is drawing 95W (i.e. the 8-pin). So, all connectors are providing about 2/3 of their capacity.

The fact you're running this on a system that doesn't natively support the 8-pin connector is the only issue here, which is why we generally suggest to not do that (esp. for cards that really need 250W+ like this one).

yes, you are right. I am talking about the 6-pin spec. I knew it was going to pull over the 75w, I just didnt know how much. there was speculation that the card would pull a balanced load from the 3 sources, which it doesn't. So it is confirmed and a better idea to give it more load from the SATA connectors.

I wonder how much the numbers would change if the card was connected fully to an external power supply. Not sure it matters.