Question Popping noise out of PCI sound card that is located in between video cards in crossfire

[badjona7an]

Hello there,

I have an interesting issue with my PC. Basically two video cards, each one been fed by a separate power supply, different single rail design PSUs, main one is 54A according to specs but it's platinum and last time it gave up at 800W drawn from AC if truly 92% efficient should've been 62A @ 12v and second one I think was somewhat 47A. Anyway, everything works perfectly fine until I turn on Crossfire and put up some load on I start getting some popping noise out of my PCI X-fi sound card which happens to be in between both GPUs. To me it seems to be something like a ground loop but it can be something else because it is present only when I enable crossfire and my knowledge definitely insufficient to figure out the problem without committing direct experiments, any ideas?

 

[AllanGH]

You are begging for problems by supplying 12V from different power supplies, quite frankly.

In general Ground / Earth leads of all PSUs should be bonded together (case 1 bonded to case 2 with something like a 6ga or 8ga THHN copper conductor, with crimped and soldered ring terminals and toothed washers); however, you will still be dealing with voltage differentials that will exist between rails that are not precisely the same output voltage. Those differentials are going to cause unpredictable issues with operation.

You would be far better-off using a single 130-ish Amp 12 VDC source, Ground / Earth commoned with any other PSU present, and supplying ALL 12VDC loads in the system. That would be any Molex or AMP SATA 12V connector, CPU connectors, and the 12VDC connections to the MB (pins 10 and 11 at the 24 pin connector).

 

[badjona7an]

Thanks for the quick response. I am aware of single 12 VDC source would be the best solution but is quite expensive. I wonder if I go through something like this adapter can help isolate the noise assuming is indeed a ground loop.

 

[AllanGH]

I can't answer the question of the adapter....it's really in the realm of "try it and see".

At minimum, though, you should bond the PSU enclosures together (using a heavy gauge conductor, or multiple sub-10ga conductors) to reduce the incidence of system-side ground loops. If you do have 12VDC differentials, you can open-up one PSU, or the other, and use the applicable trim-pots to get those voltages VERY close together, to reduce those differentials. Ideally, they should be identical, but a mV, or two, or three, would be tolerable.

REQUISITE WARNING: Use an insulated tweak to adjust VRs, and steer-clear of anything inside the PSU, that is past the transformers (usually in the middle of the PC board) on the end where the power mains connection enters the PSU.

Outside of the "Y" capacitor coupling, the "low-side" of the PSU is completely isolated (usually up to 5KV potential differences, at any rate) from the "high-side", but always behave as though it isn't.

 

[badjona7an]

I'll definitely try at first to bond both grounds with some heavier wire as they are now only connected by the tiny single wire on the 24 pin splitter and see how it goes from there, really appreciate your adequate replays, thanks again!

 

[AllanGH]

Best to do it with a direct connection from one PSU enclosure to the other....multiple wires, if you have 12 ga or lighter, and using toothed washers between the cases and ring terminals will give you the best connection. Using the PSU mounting screw holes are probably the most convenient point of attachment.

 

[badjona7an]

Well, I've tried improving common ground but it didn't fix it. I've measured both voltages on my cheap 10mV scale multimeter appear to be exact equal. As I initially thought it seems to be a bit more complicated and the fact that the noise is present only when I have crossfire enabled it has to be somewhat related to what crossfire physically does with pci-e lanes power management but even if I figure it out I doubt there will be an easy fix so I consider trying that pci-e to 32bit PCI adapter.

 

[badjona7an]

OK, since I had doubts about the accuracy of my multimeter I borrowed a professional one from a friend and turned out the issue was indeed a ground loop caused by 0.009v difference between both 12v rails. So you were exactly correct and by adjusting the voltage of one PSU to the other the issue is now fixed and yet I realize how stupid was not to ever measure that because if there was a substantial difference presumably I would've killed my motherboard long time ago

 

[AllanGH]

I am happy to see that the issue is now cleared-up for you.

Do keep in mind that, as the PSUs age, you will have output drift, and will probably have to repeat the adjustments.

....Just to keep the situation in perspective, mind you.

 

[badjona7an]

Thanks, I hope not to as I'll go for a single PSU as soon as I can afford it.

 

[AllanGH]

I'll keep your wallet in my prayers.