Power Supply A Guide to 12V PSU

[msystems]

Hey all,

Just wanted to chip in with my 12V experience in case it helps someone else. A few months ago, I upgraded my system (R5 1600, GTX 970, EPP-400-12), to a 3060ti. I started having several GPU issues with game and Windows crashes, most of which had some version of "GPU Not There Anymore" as the error.

I remembered people talking about instantaneous current draw on 30 series GPUs when they came out, so I figured the EPP-400 just didn't have enough output capacitance to keep it going during spikes. To fix this, I wedged a 1000uF capacitor in the EPP-400's output terminals. And it actually solved it! No more crashes.

In an effort to reduce jank, I designed a circuit board that has a bunch of capacitors on it, and two XT30s for 12V outputs. I'm aware that XT30s are only rated for 15A (180W at 12V) continuous, but my 3060ti only pulls 200W, and I'm willing to risk it for how small this solution is. It screws directly onto the EPP-400's terminals using the included screws, so mounting is pretty easy and it helps with wiring anyways.

The details and source files are available at this Github repo, so if you think this'll help you, give it a try!

Very nice. I'm surprised that such a small amount of capacitance did the trick. Maybe since you were only running a ~200w card, it doesn't take very much capacitance increase to have a big positive effect. Would need tools to really be able to measure what is going on. I tried a similar idea using 12x 470uf caps for 5,640uf on a y-splitter in parallel with the GPU 8-pin connector. And my project was copying another one which used like 16,000uf of caps.

 

[BaK]

The details and source files are available at this Github repo, so if you think this'll help you, give it a try!

Very nice!
Thanks for sharing!

Wouldn't you mind showing the whole rig of what we see in the 2nd pic?
Looks like there is lots of customized stuff!

 

[patchez]

Of course! Happy to share; this thread helped me a lot when I built the case a few years ago.


The whole thing is 3d printed, with the exception of some 10mm threaded steel cubes I bought off of @Colinreay a few years back. The case could be made without the cubes; I just had them around already and they help the whole thing stay together nice and neat.

I printed mine on an Artillery Sidewinder X1, which has a 300mm x 300mm build plate (hence the side panels are 300mm tall). The black plastic is PLA (bad idea, it warps), with the blue being ABS. I'm hoping to design a new version over the holidays & print it in PETG the whole way through sometime next year.

It's 230mm(deep) x 162mm(wide) x 305mm(tall), for a total of ~11.3L. It's not space optimized by any means due to the ducts for the PSU and CPU cooler, but I designed it for a 12V system using the hardware I had and didn't want a riser cable.

For the 12V stuff, I have a Meanwell EPP-400-12 getting power through the switched plug on the bottom of the case, then going to my little distro board, then out to the GPU with one cable and a cheap DC/DC ATX board & the processor with the other cable. I've got a 3d printed fan holder & duct for the PSU that let it get fresh air from through the left panel. There's technically space for an SFX PSU in there, but going 12V vastly reduced the wires and made it such that there are only wires coming out of the bottom rear of the case while maintaining easy printability.

 

[patchez]

Very nice. I'm surprised that such a small amount of capacitance did the trick. Maybe since you were only running a ~200w card, it doesn't take very much capacitance increase to have a big positive effect. Would need tools to really be able to measure what is going on. I tried a similar idea using 12x 470uf caps for 5,640uf on a y-splitter in parallel with the GPU 8-pin connector. And my project was copying another one which used like 16,000uf of caps.

My board even only has 517uF of caps; I wanted to go with only ceramics so it was limited a bit there. For my setup, it seems to be doing just as well as the bigger electrolytic I jammed in there.

My original plan was to make a 180-degree PCIe power adapter like some people use for really tall cards, but add caps to it. I ended up not doing that due to a lack of board-mount PCIe receptacles, but I'd love to see it done.

What kind of setup are you adding that much capacitance to? Must be big to need that much extra.

 

[msystems]

My board even only has 517uF of caps; I wanted to go with only ceramics so it was limited a bit there. For my setup, it seems to be doing just as well as the bigger electrolytic I jammed in there.

My original plan was to make a 180-degree PCIe power adapter like some people use for really tall cards, but add caps to it. I ended up not doing that due to a lack of board-mount PCIe receptacles, but I'd love to see it done.

What kind of setup are you adding that much capacitance to? Must be big to need that much extra.

I designed it,, but then I didn't have the right psu and gpu setup to really test it. Because its pretty situationally dependent. For my setup it was a solution in search of problem. But from your results, it seem promising, and with less capacitance.

GPJuice - Experimental device to improve PSU transient response

A little while back I saw the GN video regarding "the GPU transient problem" and came across people mentioning their experiences with this problem. Like (flex) PSUs tripping overcurrent protection when loading a benchmark, even though the PSU is rated above the sum of the tdp of their...

smallformfactor.net

 

[patchez]

I designed it,, but then I didn't have the right psu and gpu setup to really test it. Because its pretty situationally dependent. For my setup it was a solution in search of problem. But from your results, it seem promising, and with less capacitance.

GPJuice - Experimental device to improve PSU transient response

A little while back I saw the GN video regarding "the GPU transient problem" and came across people mentioning their experiences with this problem. Like (flex) PSUs tripping overcurrent protection when loading a benchmark, even though the PSU is rated above the sum of the tdp of their...

smallformfactor.net

Very interesting! If I had found that first, I probably would've made one. I also have no idea how to test it outside of my PC (and it was sporadic on mine), but hopefully someone can make use of our work here.