INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: Power usage using video hardware acceleration  (Read 6712 times)

JohnT

  • Citizen of the Universe
  • *****
  • Posts: 4627
Power usage using video hardware acceleration
« on: February 13, 2012, 12:15:12 pm »

I put a new Radeon HD 6870 card in my work computer today.  I was interested in seeing what the difference in power usage would be using the DXVA hardware acceleration vs. letting my i7 2600K do the work.  I plugged the computer into a Kill-a-watt meter and then watched the watt numbers as I played the same section of movie with the video hardware acceleration checkbox checked and unchecked in Media Center (Red October Standard).  Here are the results:

Watching a blu-ray movie action scene:
  128 watts using DXVA hardware acceleration
  149 watts without hardware acceleration

Feel free to post results here from other video cards using Nvidia CUDA or Intel QuickSync.

Logged
John Thompson, JRiver Media Center

jmone

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 14266
  • I won! I won!
Re: Power usage using video hardware acceleration
« Reply #1 on: February 14, 2012, 11:35:14 pm »

John - pretty good results with the ROHQ settings you have chosen.  Here is the results from my HTPC (i7 2600k using the IGP HD 3000) playing a 1080/50i x264 movie over the NW
- HW Accell ON (Intel Quick Sync) + Deinterlacing done by madVR (in the GPU) = 137watts
- HW Accell ON (Intel Quick Sync) + Deinterlacing done by SW (YADIF) = 146watts
- HW Accell OFF (all CPU) = 155watts (tended to jump around alot in the 150 to 160 watt range and it was the same sort of reading regardless of if the deinterlacing was YADIF or madVR).

By comparison the HTPC pulls around 92 watts just setting there in TheaterView mode doing nothing.

So....taking full advantage of the IGP (both decoding and deinterlacing) saves almost 20 watts & you don't need to buy an GPU.  However turning it off when not in use saves alot more power again!
Logged
JRiver CEO Elect

jmone

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 14266
  • I won! I won!
Re: Power usage using video hardware acceleration
« Reply #2 on: February 15, 2012, 12:41:29 am »

Take these ones with more of a grain of salt as it is my "main" PC (read Server with 6 discs) and we don't use it for watching media but it does have a nvidia (450) GPU in it so good for testing how CUVID mode goes.  The watt of metre jumped around heaps on this PC (i7 920 with an nvidia 450) playing a 1080/50i x264 movie from it's HDD
- HW Accell ON (CUVID) + Deinterlacing done by CUVID (in the GPU) = 400watts
- HW Accell OFF (CPU) + Deinterlacing done by madVR (in the GPU) = 445watts
- HW Accell OFF (CPU) + Deinterlacing done by SW (YADIF) = 460watts

So on this one, taking advantage of full HW Accell (CUVID) can save up to 60watts
Logged
JRiver CEO Elect

JohnT

  • Citizen of the Universe
  • *****
  • Posts: 4627
Re: Power usage using video hardware acceleration
« Reply #3 on: February 15, 2012, 08:20:22 am »

Thanks for posting your results Nathan.   I suppose your server numbers were with all the disks spinning?  That's a lot of wattage :)
Logged
John Thompson, JRiver Media Center

jmone

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 14266
  • I won! I won!
Re: Power usage using video hardware acceleration
« Reply #4 on: February 15, 2012, 01:48:12 pm »

Yeah it is high and I really don't know why (and I've seen it peak during boot etc at much higher figures). 
- HDD (say up to 100watts): One disc is an SSD (Main OS), then 5 x WD20EADS that have a power draw between 5-15 watts each.
- GPU (say up to 100watts): the 450 pulls up to 110watts on load
- i920 (another 100 watts):  Pending on the sub model it can pull 105ish watts under load

The other one that gets me is my HTPC.  It is a shuttle SH45H7 running the same i7 2600K, a small SSD, 1 x optical drive (was not being used for the test) and no GPU yet I was still pulling more power than your Desktop.  As Watts = VA it should not matter I'm on 230v....  it could be different power metres I guess..
Logged
JRiver CEO Elect
Pages: [1]   Go Up