top of page

03 Dec - Performance vs Power consumption

Updated: Apr 26, 2021

This article will be updated every time new components will be “deeply” tested into our Laboratory. This will be our baseline to compare performance and power consumption of any configuration.


In this article, we focus on chip performance, chip power and system power consumption in both full load and idle.

Special products need special care. Enjoy The First and The Beast at their full potential.


Last update


Some facts regarding RTX 3080 GPU or any other +300W TDP cards:

  • Can I use them in fanless mode? What’s the cost?

  • What’s the win compare to previous generation?

  • What about CPU performance? Can you tell me more?


Configurations tested


- 3800XT - 2080Super - 32Gb 3200mhz - SSD 500Go WD - 700W Fanless

- 3800XT - 3080 Trinity - 32Gb 3200mhz - SSD 500Go WD - 700W Fanless

- 6700K - 1080FTW - 32Gb 3200mhz - SSD 1To WD - 400W Fanless - 200mm Fan (Nicolas' workstation)

- 8600K - 1060 6Gb - 16Gb 3000mhz - SST 1To WD - SGX 650 - 140mm Fan (Nicolas' TV/Gaming station)


Process


For any configuration, we will test them using several benchmark software. The cooling itself is already qualified regarding loads and is treated through dedicated posts. Performance tests will be performed in the optic to optimize the component load and achieve best efficiency possible.

For all of these tests, we will use the following software:

  • Furmark, for maximal GPU load and thermal stability endurance,

  • AIDA64 for CPU load and thermal stability endurance,

  • 3DMark (Time Spy) for gaming performance score,

  • UserBenchmark for CPU & GPU performance score.

Finally we will suggest parameters you could use for your own PC.


Results


UserBenchmark: GPU performance in DX10



UserBenchmark: CPU performance in multi core application



Analysis – step 1


UserBenchmark: GPU efficiency (FPS per watt)



UserBenchmark: CPU efficiency (Points per watt)



UserBenchmark: Global efficiency regarding outlet power



Iddle power


- 3800XT - 2080Super: 47W

- 3800XT - 3080 Trinity: 52W

- 6700K - 1080FTW: 28W

- 8600K - 1060 6Gb: 26W


Analysis – step 2


UserBenchmark: GPU performance in DX10



The Beast comparison 2080Super VS 3080



  • Performance gap between generation RTX 2080 & RTX 3080 is 36% which is impressive. This gap is almost constant from 220W to 320W

  • Performance gap between GPU generation GTX 1080 & RTX 3080 is close to 130%!! Wow! Nicolas' workstation might need an upgrade 😊

  • You can enjoy your RTX 3080 (when you find one) in fanless mode. - Power consumption of RTX 3080 can be decreased by a little more than 20% to be used in fanless mode. This costs less than 5% only of calculation output. - Please note tests are made in benchmarks at extreme constant load. You will be able to fix power higher when using classic software or gaming.

  • GTX 20XX series so as RTX 3070 (&Ti??) and 3060Ti will find a sweet spot into The Beast.

  • Increasing frequency and power output have no interest for a fanless application. Overclocking the RTX 2080 Super up to 290W only increase the power calculation to +5%.

  • CPU performance gap is close. We will soon test Ryzen 5XXX series so as Intel 10XXX series.

  • The last configuration tested, 3800XT & RTX 3080 is almost 2x more efficient than the configuration 8600K & GTX 1060 at same power!


Rough data & Components voltage configuration



bottom of page