Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
GPU Power Confusion
#1
I have noticed that when testing different gfx cards, some of them report more than one GPU power result in HWinfo64.  This example that I posted below shows GPU power and GPU Core Power.  Which one is the total power consumed by the gfx card?  What is the difference between the two phrases; gpu power and gpu core power? 

http://imgur.com/a/S1ymi
Reply
#2
This has been explained a few times here on forum.
GPU Core Power covers only the core rail of the GPU (measured via VRM telemetry). GPU chip power should cover entire GPU chip power (core + aux + additional logic), which is calculated by the GPU and partially measured (core and sometimes aux too) and partially estimated based on activity counters (additional logic) or fixed constants. Note, that the core power might sometimes seem higher, due to instantaneous peaks, which are averaged in the total (chip) power calculation.
Reply
#3
(05-07-2017, 03:46 PM)Martin Wrote: This has been explained a few times here on forum.
GPU Core Power covers only the core rail of the GPU (measured via VRM telemetry). GPU chip power should cover entire GPU chip power (core + aux + additional logic), which is calculated by the GPU and partially measured (core and sometimes aux too) and partially estimated based on activity counters (additional logic) or fixed constants. Note, that the core power might sometimes seem higher, due to instantaneous peaks, which are averaged in the total (chip) power calculation.

Thank you for replying to my query.  But I'm afraid I still don't understand which of the numbers in my example above represented the power consumed by my gfx card.  101W or 88W?  I assumed the card had a method to determine VOLTS x AMPS at the +12V input stage to reach the power consumed.  Apparently it needs to be more complex than that?
Reply
#4
Yes it's more complex, because there are multiple voltage rails accounted and not all of them are capable of providing true measurements.
"GPU Power" should be more accurate, you might also better look at the average numbers sampled over the desired interval you want to measure.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)