Discussion Will it smoke? 2x RPS-200, 1= picoPSU 1= GPU

[jabcoebi]

Planning on using two MW RPS-200-12-C for a water cooling build, not the most compact at about 9~10L with 280 radiator support. The power supplies will sit under the radiator fans and be cooled that way.

Layout:
1x RPS-200 +12V -> picoPSU -> board + CPU
2x ground -12V -> picoPSU + GPU
1x RPS-200 +12V -> GPU

Is it a good idea to have the ground linked together or is better to keep them separate? And will the motherboard supply the GPU with power through the PCI-e socket?
My concern is for any conflicts between the two 12V supplies, technically a two rail PSU but with separate ground.

Is this achievable or I should just look for a more powerful 12V supply?

 

[GuilleAcoustic]

What GPU are you using ? Are you using a riser ? A powered riser ? If you need a PCIe power connector for your GPU there are solutions like the HDPlex DC-DC, instead of using multiple PSU (unless you already own them).

 

[smitty2k1]

Just get the RPS-400-12-C?

 

[jabcoebi]

What GPU are you using ? Are you using a riser ? A powered riser ? If you need a PCIe power connector for your GPU there are solutions like the HDPlex DC-DC, instead of using multiple PSU (unless you already own them).

I'm using an undervolted 5700 (non XT) for the GPU, R5 3600 for CPU, no overclock and nothing high power, last time I checked the whole system was drawing 260W under stress test.

Thinking of not using a riser and going with the M1 layout style with some modifications and main reason for going with watercooling is low noise.

HDPlex is difficult to get in the UK.

Currently, the loudest part in my system is the PSU and would like to switch to a semi-passive solution where the PSU is cooled by the radiator fans, two 140mm fans in total for the whole system.

 

[GuilleAcoustic]

There are fully passive PSU, from Seasonic or Silverstone to name only 2 . HDPlex stuff are sent from US or Germany if you buy directly from them.

 

[jabcoebi]

Just get the RPS-400-12-C?

Thanks for the sanity check! Upon further research I found the RPS-500-12-C which costs less than 2xRPS-200 and does not take twice the volume, even shorter than an FlexATX PSU. Rated at 320W with convection cooling it should have no issues with a 140mm fan spinning at 500RPM blowing air over it.

 

[jabcoebi]

There are fully passive PSU, from Seasonic or Silverstone to name only 2 . HDPlex stuff are sent from US or Germany if you buy directly from them.

Yeah, there are many options like the Nightjar SFX-L from SilverStone, but it will take too much space and it will be wasted since this is not truly a fanless build. An RPS-500 + PicoPSU solution will take less space and I will not have to deal with all the cables. I prefer soldering a few wires here and there over using modular connectors, and not having to deal with that 24pin is a big plus for me.

I know that an off the shelf solution is better in every way, and less of a headache for me, but where's the fun in that. ?

 

[GuilleAcoustic]

and not having to deal with that 24pin is a big plus for me

I read you on that, 24pins can burn in hell. Server world has been moving away from it for quite some time now.

I know that an off the shelf solution is better in every way

Not always

but where's the fun in that. ?

Nowhere ! Make sure to write a build log to share your experience with us

 

[smitty2k1]

Thanks for the sanity check! Upon further research I found the RPS-500-12-C which costs less than 2xRPS-200 and does not take twice the volume, even shorter than an FlexATX PSU. Rated at 320W with convection cooling it should have no issues with a 140mm fan spinning at 500RPM blowing air over it.

Great! Make sure to post pictures! This going in a custom case?

 

[jabcoebi]

Great! Make sure to post pictures! This going in a custom case?

Yes, custom case, not 100% decided on the layout, but a custom PSU will be necessary.

Vertical builds like the H1 have become popular lately, but I have already did a 280CLC 12L tower and other vertical builds.

With a custom PSU like this I have the option between a 10L H1 layout or a 6L A4 layout with better cooling. I’m leaning more towards the 6L build since I already had cases from 50L down to 8L, but never built something so small.

My idea is to have sandwich style layout but with the CPU and GPU fans being exposed and not sit under a grill to prevent any unnecessary noise.

concept case

It will definitely be a challenge.

 

[smitty2k1]

Yes, custom case, not 100% decided on the layout, but a custom PSU will be necessary.

Vertical builds like the H1 have become popular lately, but I have already did a 280CLC 12L tower and other vertical builds.

With a custom PSU like this I have the option between a 10L H1 layout or a 6L A4 layout with better cooling. I’m leaning more towards the 6L build since I already had cases from 50L down to 8L, but never built something so small.

My idea is to have sandwich style layout but with the CPU and GPU fans being exposed and not sit under a grill to prevent any unnecessary noise.

concept case

It will definitely be a challenge.

Great layout concept! I'd think a pair of 92mm x 15mm fans on top would be plenty. No need for anything thicker or additional fans

 

[Valantar]

Remember that the RPS-500-12 has a relatively high ripple spec (IIRC higher than what the ATX spec allows for), so while it will no doubt work, you can't expect your current undervolt to be stable.

 

[smitty2k1]

Remember that the RPS-500-12 has a relatively high ripple spec (IIRC higher than what the ATX spec allows for), so while it will no doubt work, you can't expect your current undervolt to be stable.

RPS 400 has better ripple IIRC

 

[richi-123]

negative (black) wires have to be connected together for "interference", ground wires (green with yellow) go over metal parts of cabinet or psu

so you can use a low power picopsu, it does not need so much because the sources give maximum 50W in the 24 atx pins

gpu and cpu have their own 12v Molex that in this situation would come from the RPS-500 source

these two only need to have the negative cable in common, just in case not one because the current that passes through 12v and the negative is the same and if you have several 12v cables and a single negative that would have to withstand the current of several 12v

translated from google ??

 

[jabcoebi]

Remember that the RPS-500-12 has a relatively high ripple spec (IIRC higher than what the ATX spec allows for), so while it will no doubt work, you can't expect your current undervolt to be stable.

Wow, totally forgot about output stability, thanks!
According to the datasheet:
RPS-500-12 = 200mV (peak to peak)
RPS-400-12 = 120mV (peak to peak)
RPS-200-12 = 100mV (peak to peak)

Meanwhile, something like the SilverStone Nightjar NJ450-SXL gets 26mV ripple at 12V 90% load and 45°C ambient temp, according to anandtech.com.
Going from ~20/30mV ripple to +100mV will definitely affect system stability.
Quite a difference, this changes things. Many thanks Valantar!

 

[Valantar]

Wow, totally forgot about output stability, thanks!
According to the datasheet:
RPS-500-12 = 200mV (peak to peak)
RPS-400-12 = 120mV (peak to peak)
RPS-200-12 = 100mV (peak to peak)

Meanwhile, something like the SilverStone Nightjar NJ450-SXL gets 26mV ripple at 12V 90% load and 45°C ambient temp, according to anandtech.com.
Going from ~20/30mV ripple to +100mV will definitely affect system stability.
Quite a difference, this changes things. Many thanks Valantar!

Even the technically put of ATX spec 500 is likely to be entirely stable at stock settings (at least for motherboards and GPUs with decent filtering in their VRMs, but any undervolt or overclock will inevitably be less stable with that kind of ripple. So yeah, it's something to look out for for sure. But also, remember that specs and actual performance don't necessarily match - at least for the lower power MW units their actual ripple is a good deal below the spec. MW publishes detailed test reports for their units on their site. Of course one should expect some variance, but they would have a serious QC issue if there were 2-3x ripple variance between units of the same model, so for the lower power units I tend to read the spec as a worst case scenario type of thing. The 400 still has a good deal more ripple than a top-end PC power supply, but whether it matters at all is another thing entirely.
EPP-500-12 report - 187mVp-p (spec 200)
EPP-400-12 report - 65mVp-p (spec 120)
EPP-300-12 report - 28mVp-p (spec 120)

 

[jabcoebi]

Even the technically put of ATX spec 500 is likely to be entirely stable at stock settings (at least for motherboards and GPUs with decent filtering in their VRMs, but any undervolt or overclock will inevitably be less stable with that kind of ripple. So yeah, it's something to look out for for sure. But also, remember that specs and actual performance don't necessarily match - at least for the lower power MW units their actual ripple is a good deal below the spec. MW publishes detailed test reports for their units on their site. Of course one should expect some variance, but they would have a serious QC issue if there were 2-3x ripple variance between units of the same model, so for the lower power units I tend to read the spec as a worst case scenario type of thing. The 400 still has a good deal more ripple than a top-end PC power supply, but whether it matters at all is another thing entirely.
EPP-500-12 report - 187mVp-p (spec 200)
EPP-400-12 report - 65mVp-p (spec 120)
EPP-300-12 report - 28mVp-p (spec 120)

Even the technically put of ATX spec 500 is likely to be entirely stable at stock settings (at least for motherboards and GPUs with decent filtering in their VRMs, but any undervolt or overclock will inevitably be less stable with that kind of ripple. So yeah, it's something to look out for for sure. But also, remember that specs and actual performance don't necessarily match - at least for the lower power MW units their actual ripple is a good deal below the spec. MW publishes detailed test reports for their units on their site. Of course one should expect some variance, but they would have a serious QC issue if there were 2-3x ripple variance between units of the same model, so for the lower power units I tend to read the spec as a worst case scenario type of thing. The 400 still has a good deal more ripple than a top-end PC power supply, but whether it matters at all is another thing entirely.
EPP-500-12 report - 187mVp-p (spec 200)
EPP-400-12 report - 65mVp-p (spec 120)
EPP-300-12 report - 28mVp-p (spec 120)

Thanks for the info! Any idea if an DC-DC convertor like the picoPSU XT-160 will amplify the ripple further? I’m assuming the 12V will not be greatly affected, but not sure how the ICs on the picoPSU will crosstalk back with the MW. Currently have an -96mV undervolt on an RX5700 with SGX-450 and it’s stable, but the city power is also relatively clean. Back home at the country side I have an R9 Fury with an -87mV undervolt and it’s not 100% stable. The power quality is very poor and fluctuates like crazy, I can also hear clicks and pops through some of my speaker amplifiers. Maybe a capacitor bank can help with filtering but it won’t fix the voltage fluctuations.

 

[Thehack]

Thanks for the info! Any idea if an DC-DC convertor like the picoPSU XT-160 will amplify the ripple further? I’m assuming the 12V will not be greatly affected, but not sure how the ICs on the picoPSU will crosstalk back with the MW. Currently have an -96mV undervolt on an RX5700 with SGX-450 and it’s stable, but the city power is also relatively clean. Back home at the country side I have an R9 Fury with an -87mV undervolt and it’s not 100% stable. The power quality is very poor and fluctuates like crazy, I can also hear clicks and pops through some of my speaker amplifiers. Maybe a capacitor bank can help with filtering but it won’t fix the voltage fluctuations.

You can use a pass-through design.

Search for the J-Hack M2426. It converts RPS-400/500 into regular ATX PSU. No need for a load switch.

If you have outlet power issue, you need a power conditioner. Many UPS also power conditions. But it may also just be just a function for Furys. They're known to have high transient power requirements.