2

I would like to know, whether it's possible to use two ordinary PC/ATX power supplies safely in one machine.

I've heard about Add2Psu board, but I have a very severe objections about it. (The power on molex from 1st PSU, shorts PWR_ON on 2nd PSU via relay.)

I would like to build something like that, but maybe in a much safer manner.

I'd really appreciate if someone could check the following scenario and fix it if it's needed. I refer to pin numbers on 24pin ATX connector.

  1. When power button on the casing is pressed, the motherboard sends signal to 1st PSU by shorting its 16th (green) PWR_ON pin with ground.
  2. I assume the 1st PSU is completly ready if its 8th (grey) PWR_OK pin is pulled up to +5V. (There is the main difference from the Add2Psu approach in which they only check the voltage on molex connector.) I use the 8th (grey) PWR_OK pin from 1st PSU to short 16th (green) PWR_ON pin on 2nd PSU, but they're galvanically separated.
  3. Then I wait until the 2nd PSU is ready, which is denoted by +5V on its 16th (grey) PWR_OK pin. The feedback signal returning to motherboard by its 8th (grey) PWR_OK pin is a logical conjunction of PWR_OK signals from both supplies, but electrically flows only from the 1st PSU, so the second PSU is still galvanically isolated from the first as well as the motherboard.

Could you comment on the correctness of this approach above and a few concerns about it:

  1. In my approach, I slow down the feedback signal that returns to motherboard twice as in the case of single PSU. Does such a delay can affect the waking up process, e.g. the motherboard and CPU would get power first and the graphic cards (via additional PEG connectors) must wait a bit. Could the motherboard detect this delay by measure time between sending PWR_ON and returning PWR_OK and rise an error?

  2. Should I bother with a situation when graphic cards gets power from 2 differet supplies and for instance can have two different grounds. I would feed it via PEGs from 2nd PSU, but get always get a minimal amount of power from PCI-Express connector and I cannot avoid that.

  3. How should look like the shutting down process? Should I literally revere it by triggering the shutting down process on 2nd PSU and waiting for an indicator that it's completed and after that turning down the second or simply shut down the first and the second would turn off anyhow?

  4. Are there any aspetcs that I may not be aware of?

Thank you in advance for your reply.

  • 1
    Is this for study purposes (of electricity and how things works) or you actually want a 2 PSU PC (and for what purpose)? Because if you do want a 2 PSU PC - you also want/need 2 separate main outlet. 2 PSU won't help you if the main to both PSU are dead. – Darius Jun 7 '18 at 1:36
  • @Darius I've asked this question in sheer practical purpose. I'm about to build on my own workstation equiped with 2 CPU & 4 Quadros GV100 and I'd try to set their power target to 120% (using EVGA Precision XOC), thus GPUs TDP will rise from 250W to 300W. – Adam Przedniczek Jun 7 '18 at 16:42
4

There may be issues because power supplies are regulated. If two were paralleled, and one line had, for example, 4.99 V and the other had 5.00 V, both well within tolerance, one PS would be trying to force current backward, while the other tries to limit the voltage. At the best, the full load would be carried by one PS and the other would do nothing. A worse scenario is where the driven supply has a "crowbar" to short the bus on over-voltage.

While it would be possible to use two PS by inserting two series resistors and moving the sensing point to the PCB, it seems more effort than it's worth.

  • 3
    Suppose one PSU had the green and black connected, and is used to power some devices. The other PSU is connected to the motherboard. So you don't have the motherboard or any device having two power supplies. – barlop Jun 7 '18 at 3:30
  • 2
    @barlop That is still potentially risky, because the ground is shared across all devices in the system, but may not have the same voltage between the PSU's. In most cases, you shouldn't have an issue, but I have seen PSU's where the voltage difference between the ground line out of the PSU and the ground connection to the same wall socket was almost 2V. – Austin Hemmelgarn Jun 7 '18 at 18:45
3

Both PSU’s must be earth bonded for EMC reasons therefore cannot be galvanically isolated. The ground noise must be minimal and voltages within tolerance, so wire leads must be similar.

Any delay to trigger power OK should not pose any issues. This triggers Power On Reset then jumps to BIOS. This AUX12V must be present for the BIOS selftest start time when it starts for GPU scanning but that delay after DC OK is unknown.

Any excessive ripple may deteriorate performance on ground or V+.

0

I currently have 2 different PSU's connected..

I have the green power-on hardlinked so one powersupply is on once I switch the power on, Its been removed from the powerbox case, and the bare parts are installed into, where the CD rom or DVD drive would have been..

I have the the ground connected to each other.. VITAL!! this will cause some weired thing if its not. (Bois resets, sparks and stuff)

so once you start the rest of the PC, the GPU already have power, so its normal powerup. I have a 980ti so its really powerhungy.. just turn the power off by the wall once your done.. I have a HPZ400 running a W3550 Xeon and the powersupply cant be replaced, since its not a standard ATX powersupply

If you want to make this more streamline its really easy..

buy a bigger better powersupply :)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.