How accurate is GPU power draw (GPU-Z data in Furmark)?

Mas3nk0

Honorable
Jul 31, 2013
53
0
10,630
Situation: Purchased an R9 280X Sapphire Vapor-X however my PSU is the minimum requirement namely 550W (Antec Truepower with 42A on the 12V rails). I want to burn-in my card using Furmark but with caution as reviews indicate that maximum power really peaks on AMD 7970/R9 280X on benchmark tools. Figures are around 330W (board only) at Techpowerup and 460W (system level) at Eteknix under Furmark load (not clear which profiles were used).

Questions:

1) Which Furmark profile should I use to "safely" burn-in the R9 280X?
In other words which resolution/MSAA should I use and how long should I run it?

2) When using GPU-Z in combination with Furmark it will display a GPU power draw based on the VDDC and IDDC measurements from GPU-Z. How accurate is this GPU power draw?
I tested it on my current HD5850 and received only 65W during Furmark load which seems low.
Is this because it is only working correctly (software) on newer graphics cards (or only Nvidia)?

3) The R9 280X Sapphire Vapor-X has a overclock button which increases boost clock from 1050MHz to 1070Mhz and memory clock from 1500MHz to 1550MHz. Should I run the burn-in on the lower frequencies?

Thanks in advance for your feedback.

 

Mas3nk0

Honorable
Jul 31, 2013
53
0
10,630
Additional question: "If I switch the overclock button out from Sapphire the R9 280X should revert back to 1050MHz core clock and 1500MHz (6GHz effective) memory clock which is the exact same specs as the ASUS R9 280X. Would my assumption be correct that the SApphire version would then have similar power draw as the ASUS since the GPU core is the same?