- Joined
- Mar 16, 2015
- Messages
- 218 (0.07/day)
- Location
- Virginia, United States
| Processor | Intel i7 4790K |
|---|---|
| Motherboard | Gigabyte Z97X-UD5H-BK |
| Cooling | LEPA AquaChanger 240 |
| Memory | G.Skill Ares Series 16GB (2 x 8GB) DDR3-2133 |
| Video Card(s) | EVGA GTX 980 Superclocked ACX 2.0 |
| Storage | Seagate St1000dx001 |
| Case | Cooler Master HAF X ATX Full Tower |
| Audio Device(s) | Onboard |
| Power Supply | EVGA 850W ATX12V / EPS12V |
Hello all!
Been playing around with my GF's 980 Ti Classified recently. Specifically I've been investigating if I can lower the heat output a bit. Running both of our rigs at the same time in a small room in this summer heat is killer!
Through my testing, I had some very interesting results that I thought I'd share. I don't know, I'm an engineering major so maybe this is more interesting to me than the average joe, but here it is anyways!
So firstly, how the data was collected.
In EVGA precision, with a classified card, you can view the actual current voltage (So you can see the vdroop happening)
GPU Z tells me TDP, but there is a special circumstance. GPU Z reports a % of TDP, and this card I've BIOS modded for a TDP of 600W. Therefore it's a bit of backwards calculating, multiplying the new limit by the % to get the actual power usage.
I set the voltage using the classified voltage controller, observed it using EVGA, and calcualted the TDP using GPU-Z.
Specifically, here are two data rows:
So in the top test I set my voltage to 1.11875 (Excel shows a rounded value). I left the clock speed alone and ran a benchmark in Heaven 4.0.
I used GPU-Z to note the MAX Temperature, Fan speed %, TDP, as well as the eventual Boost and VRAM clock speeds.
During the benchmark I watched precision for the LOWEST voltage I saw, to me for a rock solid OC that's whats important. I marked this as "Actual Voltage"
The next test everything remained the same, except for clock speed. I DECREASED it by 90 MHz. Common intuintion would assume the same, or less heat and power usage...but actually...
Power and heat went UP. The reason being, the decresed clock speed meant the card wasn't working as hard, so during the benchmark the voltage stayed higher throughout, including during the heaviest parts, meaning a higher current flow (even with the decreased clock speed) and therefore a higher TDP.
Honestly, it's actually a fairly significant TDP increase as well, as far as overclocking goes (Which is basically nickel and diming anyways
)
Now one note, this was the TDP AT MAX. I did not record the average TDP during the runs, I imagine it too was higher, but I do not have the data. I might re-run just for fun to see.
In case anyone was interested, the heaven score with a 90 MHz underclock, and the lowest stable voltage (1.05 set, 1.02 after vdroop) ended up dropping 7%(from 2397 stock, to 2211 underclocked).
This was with a TDP drop of a little over 60 Watts (or 10%) from the stock card settings.
Max temperature dropped 5 degrees C, with a 17% lower max fan speed.
Been playing around with my GF's 980 Ti Classified recently. Specifically I've been investigating if I can lower the heat output a bit. Running both of our rigs at the same time in a small room in this summer heat is killer!
Through my testing, I had some very interesting results that I thought I'd share. I don't know, I'm an engineering major so maybe this is more interesting to me than the average joe, but here it is anyways!
So firstly, how the data was collected.
In EVGA precision, with a classified card, you can view the actual current voltage (So you can see the vdroop happening)
GPU Z tells me TDP, but there is a special circumstance. GPU Z reports a % of TDP, and this card I've BIOS modded for a TDP of 600W. Therefore it's a bit of backwards calculating, multiplying the new limit by the % to get the actual power usage.
I set the voltage using the classified voltage controller, observed it using EVGA, and calcualted the TDP using GPU-Z.
Specifically, here are two data rows:
So in the top test I set my voltage to 1.11875 (Excel shows a rounded value). I left the clock speed alone and ran a benchmark in Heaven 4.0.
I used GPU-Z to note the MAX Temperature, Fan speed %, TDP, as well as the eventual Boost and VRAM clock speeds.
During the benchmark I watched precision for the LOWEST voltage I saw, to me for a rock solid OC that's whats important. I marked this as "Actual Voltage"
The next test everything remained the same, except for clock speed. I DECREASED it by 90 MHz. Common intuintion would assume the same, or less heat and power usage...but actually...
Power and heat went UP. The reason being, the decresed clock speed meant the card wasn't working as hard, so during the benchmark the voltage stayed higher throughout, including during the heaviest parts, meaning a higher current flow (even with the decreased clock speed) and therefore a higher TDP.
Honestly, it's actually a fairly significant TDP increase as well, as far as overclocking goes (Which is basically nickel and diming anyways
)Now one note, this was the TDP AT MAX. I did not record the average TDP during the runs, I imagine it too was higher, but I do not have the data. I might re-run just for fun to see.
In case anyone was interested, the heaven score with a 90 MHz underclock, and the lowest stable voltage (1.05 set, 1.02 after vdroop) ended up dropping 7%(from 2397 stock, to 2211 underclocked).
This was with a TDP drop of a little over 60 Watts (or 10%) from the stock card settings.
Max temperature dropped 5 degrees C, with a 17% lower max fan speed.
Last edited:
