Question When does graphics card demand power from PCI-E?

Oct 20, 2021
8
0
10
0
Ok...so this thing happened today. I was changing the fluid in my water cooled pc and while I was leak checking.....I found a leak. One of my rotary fittings started leaking onto my RTX2070. I got quite a bit of fluid on the card. I have been unable to test it because I am waiting on new fittings to arrive. I had the pump only powered by the PSU with the motherboard unpowered during the test (jumper wire). I did however have PCI-E power plugged into the card and it was receiving power from the PSU. My question is "when does the Graphics card power on in this situation?" I am hoping it was unpowered since the Mobo was unpowered. I don't know the inner workings of the power switching that goes on between the Mobo, GPU and PSU(PCI-E power). I'm hoping it was unpowered and therefore less likely it was damaged.
 

Eximo

Titan
Ambassador
Not sure I have ever thought of that scenario. 12V would be supplied to the card, presumably those are all ganged together with the PCIe 12V pins on the slot connector.

I don't believe there is any active switching to prevent the flow of 12V back through the motherboard, but certainly wouldn't have provided much, if any, power to the GPU itself without the vBIOS telling it to turn on.

Mostly guessing though.
 
Oct 20, 2021
8
0
10
0
Not sure I have ever thought of that scenario. 12V would be supplied to the card, presumably those are all ganged together with the PCIe 12V pins on the slot connector.

I don't believe there is any active switching to prevent the flow of 12V back through the motherboard, but certainly wouldn't have provided much, if any, power to the GPU itself without the vBIOS telling it to turn on.

Mostly guessing though.
Thanks for the reply. Kinda what I'm hoping. Its driving me crazy not being able to test it. Just trying to do my best to clean the card and dry it out before the fittings arrive.
 

Eximo

Titan
Ambassador
To add, pretty sure that is what the 3.3v supplied to the slot is for. Without it, the GPU might not ever try to power up the VRMs to deliver the low voltages to memory and GPU. Doesn't mean something else didn't go wrong.

Still, if you saw no magic smoke, it is probably alright.

I have computers survive AIO leaking directly onto the back of a running GPU, even with no backplate, and survived.
 
Oct 20, 2021
8
0
10
0
To add, pretty sure that is what the 3.3v supplied to the slot is for. Without it, the GPU might not ever try to power up the VRMs to deliver the low voltages to memory and GPU. Doesn't mean something else didn't go wrong.

Still, if you saw no magic smoke, it is probably alright.

I have computers survive AIO leaking directly onto the back of a running GPU, even with no backplate, and survived.
Gives me hope! Thanks.
 
Oct 20, 2021
8
0
10
0
Gives me hope! Thanks.
The price of GPU's these days is crazy.
I'm not aware exactly how PCIe devices are set up, but I'm pretty sure that on boot up, the GPU is going to be in a low power state because the PCIe spec states that unless the expansion card announces itself as a high-power one, it's limited to 25W @ 12V (https://en.wikipedia.org/wiki/PCI_Express#Power)

It's probably after getting that set up does the graphics card attempt to draw more power.
Thats great. Thanks for the reference. Sounds like if anything, it was running in a low power state. Hopefully, all is well. Last thing I want to do is have to buy a new GPU right now. The price of this same card (if you can get it) has more than tripled due to the chip shortage and shipping issues I assume.
 

InvalidError

Titan
Moderator
Not sure I have ever thought of that scenario. 12V would be supplied to the card, presumably those are all ganged together with the PCIe 12V pins on the slot connector.
From the couple of Buildzoid GPU VRM nerd-outs I have watched, it seems like most GPUs simply power 0-4 VRM phases (often includes two for VRAM) and fans from slot 12V, everything else from AUX 12V.

There usually is no electrical connection between slot 12V and AUX 12V.
 

Karadjgne

Titan
Ambassador
If you are using distilled water for coolant, I'd not be overly worried about the leak. Distilled water is as neutral as it gets, it's not electrically conductive, no free ions, no minerals or dissolved metals or really anything that will pass electrons and allow for conductivity. It's just a mess to clean up, not so much a mess to worry over.

The propylene glycol used in AIO's is somewhat different. It's very conductive, as well as being a mess to clean up.

Either way, a good soak from a can of CLC contact cleaner on the gpu where the water was won't hurt your chances, only help with any residue from biocides etc.
 
Oct 20, 2021
8
0
10
0
I'm using Mayhems X1 Clear. I'm not sure how conductive it might be. I've cleaned everything as well as I can with rubbing alcohol and reassembled the water block. I'll get new fittings in tomorrow and reinstall everything. Fingers Crossed!!!
 

InvalidError

Titan
Moderator
If you are using distilled water for coolant, I'd not be overly worried about the leak. Distilled water is as neutral as it gets, it's not electrically conductive, no free ions, no minerals or dissolved metals or really anything that will pass electrons and allow for conductivity.
Distilled water may not be conductive initially but it will readily leach contaminants from everything it touches and become conductive from that. It may take a minute or two for the process to get going at a readily visible rate, then it quickly becomes one of the worst things you could spill on a powered board.

Distilled water buys you one, maybe two minutes to turn the thing off. Distilled water that has been in a loop for a while isn't distilled anymore: it will have picked up some aluminum, copper and other metals from the loop's metallic components over time if you didn't add an adequate anti-corrosion/leeching additives package.
 

Karadjgne

Titan
Ambassador
Description
Low Electrical Conductivity.
Prevents Metallic Corrosion.
Prevents Algae Build Up.
2 Year Shelf Life.
Colour: Clear.

Unless that was a serious leak, and you used the gpu while it was leaking, I'd think your chances are pretty good it'll be ok. If there was a short, you'd have seen it arc or puff smoke or something, even as far as popping the psu protections.
 

InvalidError

Titan
Moderator
If there was a short, you'd have seen it arc or puff smoke
You only get arcs when there either is sufficiently high voltage to jump a gap (about 300V between points at standard atmospheric pressure and temperature) or there is enough current passing through contaminants to turn them into plasma and strike the arc like scratch-start welder does.

The biggest danger would be enough leakage between 12V and the control signal to turn the high-side switch on and blast the core with a voltage much higher than normal.
 

Karadjgne

Titan
Ambassador
I see arcs with 12/24v, 120v with less than 15A on a daily basis. Doesn't require 300v or anything close. All it requires is close enough proximity to a pathway and enough current to support it. A single microscopic drop of water with high surfactant, contaminate or ion count between 2 traces or solder points would be enough if it was 12v and ground, or as you say 12v and the control signal. Even a wimpy 6pin powered gpu has a @ 18A potential on the 12v plane. More than enough support.

Which would, as you say, turn that gpu to a pile of smoking silicon.
 

InvalidError

Titan
Moderator
I see arcs with 12/24v, 120v with less than 15A on a daily basis. Doesn't require 300v or anything close. All it requires is close enough proximity to a pathway and enough current to support it. A single microscopic drop of water with high surfactant, contaminate or ion count between 2 traces or solder points would be enough
Current passing between two points through contaminants and possibly burning itself a permanent path over time is called tracking, not arcing.
 

Karadjgne

Titan
Ambassador
Cool. Didn't know it had a specific name. To me, I see a blue-yellow-white colored light flash between 2 different conductors, it's an arc. But thats layman, not technician perspective. Nice to learn something new 😁
 

InvalidError

Titan
Moderator
Cool. Didn't know it had a specific name. To me, I see a blue-yellow-white colored light flash between 2 different conductors, it's an arc.
To be fair, tracking is a specific case of arc. An arc can happen and be sustained in any media as long as you have sufficient voltage to break through it and sufficient current to sustain it while tracking is specific to surfaces it can leave tracks on. The yellow-white flashes on a PCB that has a tracking failure are tiny arcs passing through the flame and soot of burning contaminants and PCB material when contaminants get locally hot enough to either ionize or ignite - a microscopic flame is plasma too.
 
Reactions: Karadjgne
Oct 20, 2021
8
0
10
0
Great news!!! Card is working. No artifacting or screen flickering. Just ran Microsoft Flight Simulator 2000 with no issues whatsoever. I installed new fittings and had to bend a little more tubing to keep potential leak points further away from the expensive stuff. I think I'll make changing o-rings in fittings part of the future maintenance schedule. The fitting that leaked was actually a rotating 90-degree. I'm thinking in the future I'm going to design my builds to avoid using these. I think they are a weak point in the cooling system. Thanks for all the info and encouragement. I was really sweating this the last couple days.
 
Oct 20, 2021
8
0
10
0
Hmm, I have 6 of those in my rig lol. Kinda impossible not to. MITX doesn't lend to nice big straight loops
Yeah, I guess you just can't always avoid it. I still have several in my current build but am thinking of re-designing the loop a bit on my next fluid change. I've never built a MITX loop but I'm sure your pretty limited on options. Just keep an eye on those rotary joints!!!
 
Ok...so this thing happened today. I was changing the fluid in my water cooled pc and while I was leak checking.....I found a leak. One of my rotary fittings started leaking onto my RTX2070. I got quite a bit of fluid on the card. I have been unable to test it because I am waiting on new fittings to arrive. I had the pump only powered by the PSU with the motherboard unpowered during the test (jumper wire). I did however have PCI-E power plugged into the card and it was receiving power from the PSU. My question is "when does the Graphics card power on in this situation?" I am hoping it was unpowered since the Mobo was unpowered. I don't know the inner workings of the power switching that goes on between the Mobo, GPU and PSU(PCI-E power). I'm hoping it was unpowered and therefore less likely it was damaged.
It's always powered, but not the entire card. You'll have to do a plug and pray I'm afraid to find out.
 

ASK THE COMMUNITY