Page 2 of 2 FirstFirst 12
Results 11 to 17 of 17

Thread: Looking for 750W+ PSU for VEGA 56, Ryzen 1600

  1. #11
    Join Date
    Oct 2017
    Location
    Grand Junction, CO
    Posts
    1,655
    Thanks Thanks Given 
    70
    Thanks Thanks Received 
    14
    Thanked in
    13 Posts

    Default

    Quote Originally Posted by Jon Gerow View Post
    Whine on the GPU is a whole other thing that has nothing to do with AC ripple.

    I've even had USB keyboards with "coil whine".

    I stand corrected; I got confused by

    Quote Originally Posted by Jon Gerow View Post
    It's just the frequency of the power going through the magnetics of the PSU being at an audible level.

  2. #12
    Join Date
    Oct 2017
    Location
    Grand Junction, CO
    Posts
    1,655
    Thanks Thanks Given 
    70
    Thanks Thanks Received 
    14
    Thanked in
    13 Posts

    Default

    Quote Originally Posted by Grixor View Post
    As you said, I really don't like believing that the PSU could be the cause

    Don't forget the speed of light limit; it takes time (albeit small) for the power supply to even 'know' the voltage has dropped at the graphics card, that is why one needs a decoupling (bypass) capacitor on the card itself, and a large capacitor may be exactly what one does not want as its stray inductance slows the reaction speed.

    1 GHz has a half-wavelength of 15 cm (for the speed of light) and one needs to be far below this length, so one can see that the capacitor needs to be on the graphics board as the power supply is too far away.
    Last edited by ashiekh; 1 Week Ago at 08:55 PM.

  3. #13
    Join Date
    Oct 2006
    Location
    jonnyGURU forums, of course!
    Posts
    15,582
    Thanks Thanks Given 
    506
    Thanks Thanks Received 
    58
    Thanked in
    48 Posts

    Default

    Quote Originally Posted by ashiekh View Post
    I stand corrected; I got confused by
    No, I get confused too. Since DC voltage has a frequency of 0HZ and there's no AC power, what is causing these frequencies in magnetics on the DC side? I'm not going to pretend to know. If I did, I could solve the world's graphics card coil whine issues.

  4. #14
    Join Date
    Jun 2007
    Posts
    6
    Thanks Thanks Given 
    0
    Thanks Thanks Received 
    0
    Thanked in
    0 Posts

    Default

    I think you proved that the issues are due to the power used by your Vega 56. The Vega 56 could be triggering over current protection in that CX750.

    The reason I think this is that I have observed multi-millisecond peaks of 60 Amps (~700 Watts) consumed by a Radeon VII while running Furmark and configured to automatically overclock in the Radeon 19.5.1 driver. This current is derived from sub-millisecond voltage measurements across a high power resistor that is in series with only the 12 Volt inputs to the GPU (PCIe, AUX1, AUX2) . Add in ~4 A (~50 W) or more drawn by the CPU when playing games and some 750 W power supplies might turn off or "hiccup."

    Also, air temperatures inside a PC over 40 C will drop the capacity of most desktop power supplies. Check for a Derating Curve when shopping for a power supply.

    Quote Originally Posted by Grixor View Post
    I tried some more GPU stress tools/benchmarks and using furmark with a +50% powerlimit, the GPU was at 80% usage, drawing 210W and sometimes the result was another kernel power error.
    I guess I really would need to try out a different PSU as I can't recreate this behavior with -50% powerlimit for example (which strangely enough puts my GPU to 99% usage in furmark).
    As you said, I really don't like believing that the PSU could be the cause, but shouldn't those stress tests actually confirm that the GPU is capable and stable?
    Last edited by atlr; 6 Hours Ago at 08:49 PM.

  5. #15
    Join Date
    Oct 2017
    Location
    Grand Junction, CO
    Posts
    1,655
    Thanks Thanks Given 
    70
    Thanks Thanks Received 
    14
    Thanked in
    13 Posts

    Default

    Quote Originally Posted by atlr View Post
    I have observed multi-millisecond peaks of 60 Amps

    Should not the over-current protection be 'slow blow'?

  6. #16
    Join Date
    Oct 2006
    Location
    jonnyGURU forums, of course!
    Posts
    15,582
    Thanks Thanks Given 
    506
    Thanks Thanks Received 
    58
    Thanked in
    48 Posts

    Default

    Quote Originally Posted by ashiekh View Post
    Should not the over-current protection be 'slow blow'?
    It can be. Depends on the PSU design. It is a cost adder.

  7. #17
    Join Date
    Jun 2007
    Posts
    6
    Thanks Thanks Given 
    0
    Thanks Thanks Received 
    0
    Thanked in
    0 Posts

    Default

    ashiekh, I do not know.

    I am hoping that relaying a direct experience with a Radeon VII which is similar in power usage to a Vega 56 is helpful to the OP. For me, the ability to measure periods of time when a Radeon VII used ~60 A @ 12 V was only possible using a power supply with a 67 A @ 12 V rail supplying the GPU. (A custom PCIe connector was used to isolate PCIe connector power for measurement purposes along with AUX1 and AUX2). I also ran Furmark with a power supply with a 12 V rail of less than 60 A and the result was an OCP reset by the power supply.

    In case someone is thinking of asking, I don't have any recommendation for any particular make or model of power supply.

    Quote Originally Posted by ashiekh View Post
    ?
    Last edited by atlr; 6 Hours Ago at 08:45 PM.

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •