Server Power Supply Interface Board - for standalone miners and GPU rigs

[sidehack]

If it's a linear regulator shifting 5V down to 3.3V, running 12V into it is going to increase the power dissipation (and therefore heat) by five times. If it is indeed designed to give you 3A at 3.3V, it's dissipating 5W from a 5V source and that would be 26W from a 12V source - likely enough to explode it. I'm kinda surprised they didn't put a switching regulator on there instead, which is common for anything expecting to pull more than 1.5A or so.

My boards give you 5V rated for 2A; the circuit should handle 3A output.

[repairguy]

Hi,

At the expence of sounding like a newbee, i am a bit at a loss.
After browsing trough the comments, i am still trying to figure out where to get the 3.3V line from on the DPS-2000BB.
AFAIK, pci-e slot draws 12v and 3.3v.
I am using the USB RISERS - https://www.hashratestore.com/shop/cables/usb-pcie-riser-powered-molex-60-cm .
Apart from the standard pci-e 6 AND 6/8 PIN SETUP (12V), it also uses 4 pin molex which carries 12v G G 5v.
However, according to this:
Except for a potential maximum 10W from the PCIe +3.3V slot, all remaining power consumption will be on a +12V output.
there should be 3.3v comming from the molex and there isnt.
I can only conclude that instead of 3.3v it uses the 5v from the molex in the same way.

The PCIe X16 slot supplies 75W of which 10W comes from the +3.3V output (3.0A)

The following questions comes to my mind:
1 - How many WATT/AMP can i pull from the DPS-2000BB 5V line ??
2 - If not, can i use a molex from PSU powering the RIGS ?

I do apologize in advance if this has been covered before.

Cheers,
Mac

It uses a voltage regulator to make 3.3v.

ok, do you mean the usb-pcie-riser uses a voltage regulator ?
That would mean i would use the molex as follows: 12v G G 12V instead of 12V G G 5V and the riser does the rest.
Or do i need to adapt one to get the 3.3v from the 12v line ?

Cheers,
Mac

Yes, the usb riser has a regulator on the 16x board that plugs into the bottom of the power supply.  I do not know right off of my head if it converts the 5v to 3.3 or the 12v to 3.3 but logic would tell me it uses the 5v, otherwise it would probably get really hot. You would need to look at the regulator and make sure that it can accept 12v input and then see how hot it gets if you feed it with 12v. The regulator is going to dissipate almost 3x the watts that the card uses at 3.3v if you feed it with 12v, so it may only work for a little while, then burn out the regulator.

looking at my usb riser right now - yes the 5V is used, and is connected to the flat 3-pin chip, presumably to drop down to 3.3V as mentioned.

putting 12V through that is a TERRIBLE idea.

It might sound like terrible idea, however, after looking at the data sheet, i found that max input current is in fact 12V.
It does get quite hot though.
Considering that u only connect the GFX cards to a single PSU, and the the 12V line is not an option, where do i get a 5V or a 3.3V line from if not supplied by the PSU ?

You could use the connections from the power supply that powers the motherboard to power the risers

[repairguy]

If it's a linear regulator shifting 5V down to 3.3V, running 12V into it is going to increase the power dissipation (and therefore heat) by five times. If it is indeed designed to give you 3A at 3.3V, it's dissipating 5W from a 5V source and that would be 26W from a 12V source - likely enough to explode it. I'm kinda surprised they didn't put a switching regulator on there instead, which is common for anything expecting to pull more than 1.5A or so.

My boards give you 5V rated for 2A; the circuit should handle 3A output.

Other manufactures may be different, I have two different types of usb risers.  One has a linear regulator and the other I cannot find a data sheet for, but I would guess it is a linear also.  I also doubt the card pulls the full 10w at 3.3v. I will measure it when I get a minute. 

You also can't assume that the riser board is designed to deliver all 10w.

[MacStone]

If it's a linear regulator shifting 5V down to 3.3V, running 12V into it is going to increase the power dissipation (and therefore heat) by five times. If it is indeed designed to give you 3A at 3.3V, it's dissipating 5W from a 5V source and that would be 26W from a 12V source - likely enough to explode it. I'm kinda surprised they didn't put a switching regulator on there instead, which is common for anything expecting to pull more than 1.5A or so.

My boards give you 5V rated for 2A; the circuit should handle 3A output.

Other manufactures may be different, I have two different types of usb risers.  One has a linear regulator and the other I cannot find a data sheet for, but I would guess it is a linear also.  I also doubt the card pulls the full 10w at 3.3v. I will measure it when I get a minute. 

You also can't assume that the riser board is designed to deliver all 10w.

I thought you "shouldnt mix" PSU´s when powering a GFX.
EX.:
PSU 1 - MB + 1 GFX PCI-e & Riser (by riser i mean the 4 pin molex)
PSU 2 - GFX PCI-e & Riser 2, 3, etc, etc.....

From what i read, it seams that mixing and matching PSU´s will cause undesired efects (Burn baby Burn)

Or am i just getting this wrong ??  

[MacStone]

If it's a linear regulator shifting 5V down to 3.3V, running 12V into it is going to increase the power dissipation (and therefore heat) by five times. If it is indeed designed to give you 3A at 3.3V, it's dissipating 5W from a 5V source and that would be 26W from a 12V source - likely enough to explode it. I'm kinda surprised they didn't put a switching regulator on there instead, which is common for anything expecting to pull more than 1.5A or so.

My boards give you 5V rated for 2A; the circuit should handle 3A output.

Other manufactures may be different, I have two different types of usb risers.  One has a linear regulator and the other I cannot find a data sheet for, but I would guess it is a linear also.  I also doubt the card pulls the full 10w at 3.3v. I will measure it when I get a minute. 

You also can't assume that the riser board is designed to deliver all 10w.

You are correct.
As my Dad always said: Never Assume anything. It makes an ASS out of U and ME.
While i do not assume it will deliver the 10w, i do believe the it should be thereabout.
The model is:  AZ1084S-3.3E1
I did a quick test using the 12V.
1 - I does not work....   
2 - It does get very hot...

[repairguy]

If it's a linear regulator shifting 5V down to 3.3V, running 12V into it is going to increase the power dissipation (and therefore heat) by five times. If it is indeed designed to give you 3A at 3.3V, it's dissipating 5W from a 5V source and that would be 26W from a 12V source - likely enough to explode it. I'm kinda surprised they didn't put a switching regulator on there instead, which is common for anything expecting to pull more than 1.5A or so.

My boards give you 5V rated for 2A; the circuit should handle 3A output.

Other manufactures may be different, I have two different types of usb risers.  One has a linear regulator and the other I cannot find a data sheet for, but I would guess it is a linear also.  I also doubt the card pulls the full 10w at 3.3v. I will measure it when I get a minute. 

You also can't assume that the riser board is designed to deliver all 10w.

I thought you "shouldnt mix" PSU´s when powering a GFX.
EX.:
PSU 1 - MB + 1 GFX PCI-e & Riser (by riser i mean the 4 pin molex)
PSU 2 - GFX PCI-e & Riser 2, 3, etc, etc.....

From what i read, it seams that mixing and matching PSU´s will cause undesired efects (Burn baby Burn)

Or am i just getting this wrong ?? 

I can confirm with 100% certainty you can use a server power supply to provide the power for the 6pin connections and the motherboard power supply to power the slots.  I have built alot of machines this way.  All of the power connections are isolated inside the card.  You should never directly connect 2 power supplies together, unless they have a current share pin and it is connected correctly.

I used a 650 watt supply to run the motherboard and the risers for 5 cards.  When I added one more it wouldn't run.  I switched to an 850watt for 6 cards and works great. I use a 1570w dell supply to run the 6pin connectors

[tolip_wen]

Excellent thread!

Thx all who posted links for parts.

I'm DIYing my IBM and HP PSU's but I'll throw down for the cause when they ready.
(My HW setup is typically a moving target anyways)

A foreward thinking enterpreneur might consider adding a small brain to monitor temps and generate PWM for fans.

I am most familiar with TI MSP430 mixed signal micro's but Atmel(Adruino) has similar offerings.
Adruino prolly has more web support from users but TI has a good collection of examples included with the free IDE. I am TI centric but not biased

EX:
With a USD $9.00 launchpad you could measure 2-3 temps and generate a couple clocks.
bare chips are WAY less, each board could have a chip and user could add USB(launchpad or adruino) if desired.

Copper is free on PCB's, it could remain an unpopulated option.
(assumes SMD parts, holes are expensive on PCB's)
Just my .000001 Satoshis.

Thx again all, for the HW  pr0n!

YMMV

[sidehack]

A foreward thinking enterpreneur might consider adding a small brain to monitor temps and generate PWM for fans.

We're designing the DPS-2000BB boards with a manual PWM fan controller for 4-wire fans, and two 4-pin fan headers per board. An external fan speed input pin will also exist so you can rig up your own automatic controls if you want. People are already complaining about prices (bitcoin people tend to be a greedy bunch) so adding more features (which requires more dev time including firmware and flashing setups) is just gonna hurt more. Also, please don't equate Atmel with Arduino. It's an insult to Atmel.

[MacStone]

A foreward thinking enterpreneur might consider adding a small brain to monitor temps and generate PWM for fans.

We're designing the DPS-2000BB boards with a manual PWM fan controller for 4-wire fans, and two 4-pin fan headers per board. An external fan speed input pin will also exist so you can rig up your own automatic controls if you want. People are already complaining about prices (bitcoin people tend to be a greedy bunch) so adding more features (which requires more dev time including firmware and flashing setups) is just gonna hurt more. Also, please don't equate Atmel with Arduino. It's an insult to Atmel.

I take it that the boards are still in development ?

[sidehack]

Unfortunately, yes. I need to go over the PCB design as soon as I have time, and then we'll make any necessary changes and send off for a small batch to test with. Progress is kinda slow going because the PCB guy is also the web guy and order handling. I'm the electronics designer but also 90% of the manufacturing. There's only so many hours in a day.

[RickJamesBTC]

On all of my setups I have the dps2000bb powering the pcie connectors and the 12v wire for the riser card. The 5v line goes to a hd connector on the psu running the motherboard, and both grounds are tied together at the riser connection. No problems whatsoever.

[repairguy]

On all of my setups I have the dps2000bb powering the pcie connectors and the 12v wire for the riser card. The 5v line goes to a hd connector on the psu running the motherboard, and both grounds are tied together at the riser connection. No problems whatsoever.

Yep, that will work just fine.  Connecting negative terminals between power supplies will not hurt anything, most of the time they are all ready bonded together anyway.  They are bonded to ground in the power supply and all of the ground are connected together in the electrical panel. All of the positive connections are isolated in the card.

[xgtele]

Here is a board I built for my DPS-800GB, It hooks up to a Raspberry Pi so I can monitor the supply's status and power cycle it remotely...

I like your setup. Could you please tell more about your board and PRi software you use to control PSU?

[pmorici]

Here is a board I built for my DPS-800GB, It hooks up to a Raspberry Pi so I can monitor the supply's status and power cycle it remotely...

I like your setup. Could you please tell more about your board and PRi software you use to control PSU?

    Sure; The board, made to work with the HP DPS-800GBA, is my own design it's got logic to switch the supply on and off and convert the signal levels coming out of the supply into voltages the Rpi can handle w/o blowing up.  If you don't want to bother with a Pi you can also just use a jumper to turn it on and off manually.  The 12 position screw terminal is rated to supply up to 15 Amps on each position so you can push the supply to it max w/o worrying about starting a fire.  The supply itself can handle 2 - 2.5 AntMiner S1's.

    The board interfaces with the Rpi via GPIO.  There are 4 wires one to turn the supply on and off, one to tell the Pi the supply is plugged into a wall socket and receiving power, one to tell the Pi the 12 Volt power is on and working and the 4th is a ground wire.  I have a couple shell scripts that are a few lines each that configure the GPIO setting on the Pi and then you can ssh into and check the status and switch the power on / off by the running the corresponding on off command.  Each Pi has enough GPIO pins to control and monitor multiple supplies.  My plan is to make an add-on for MinePeon were you can control the supply right from the web interface.  I'm working on that now.

    Here are some photos of my latest revision, If anyone is interested, I'm building a bunch of these I haven't settled on an exact price yet but I'm thinking $40 each including shipping.  I'll post a separate thread soon with more details but anyone interested in a board for the DPS-800 or a remotely controlable PSU keep an eye out for more info to come.

Edit: There is a separate thread for more info on these now https://bitcointalk.org/index.php?topic=622783.0 or check out my website http://gigampz.com

http://imgur.com/a/FmUl6

[xgtele]

pmorici, I have a pair of Artesyn PSU 7001138-Y000, used with IBM X-Series x3650 Server, P/N 24R2730 and would like to use RPi to control them. I think what I need is control interface slide-in board to connect it to RPi. Could you please advice me where to start if you can? Thanks!

[xgtele]

bringing this up again:
Delta DPS-835AB, Artesyn 7001138-Y000, IBM x3650 3655 835W (12.1V 69A) Power Supply


at 835W, these units turn on with a simple pin shorting (pins 3 and 4 - which are easy to bend into contact and place a small solder bead across) and have large, well-spaced sockets that can either be filled with wire leads and solder, or have the outer plastic snipped away to expose the metal contacts.

I converted a unit with about 20min, some AWG14 leads, and a few different sizes of heat shrink (1 size for the leads and 1 size to seal off the entire terminal to any accidental contact or shorting). I plan to post some pictures shortly, but right now its doing an excellent job of powering a 7-card bitfury system usin 2+ and 2- power lines that experience no warmth. It will also power an antminer shortly, either using 1+ and 1- per blade (pushing it a bit) or 4 wires to one blade and a seperate PSU for the other

http://www.rcgroups.com/forums/showthread.php?t=1766215

Please post your pics

[sidehack]

Just throwing this out there, but my 750W boards (12V 62A) aren't much less power than that, and already have a header interface for logic-level external control. Also an on-off switch and the ability to be wired in parallel for higher currents and a built-in current meter and both 3.3V and 5V outputs to actually power an external controller if necessary.

Also I'm gonna start listing some of those three-board panels for sale, wherein it's pre-wired for three 750W supplies to work together as a single 2200W PSU.

[klondike_bar]

bringing this up again:
Delta DPS-835AB, Artesyn 7001138-Y000, IBM x3650 3655 835W (12.1V 69A) Power Supply


at 835W, these units turn on with a simple pin shorting (pins 3 and 4 - which are easy to bend into contact and place a small solder bead across) and have large, well-spaced sockets that can either be filled with wire leads and solder, or have the outer plastic snipped away to expose the metal contacts.

I converted a unit with about 20min, some AWG14 leads, and a few different sizes of heat shrink (1 size for the leads and 1 size to seal off the entire terminal to any accidental contact or shorting). I plan to post some pictures shortly, but right now its doing an excellent job of powering a 7-card bitfury system usin 2+ and 2- power lines that experience no warmth. It will also power an antminer shortly, either using 1+ and 1- per blade (pushing it a bit) or 4 wires to one blade and a seperate PSU for the other

http://www.rcgroups.com/forums/showthread.php?t=1766215

Please post your pics

I dont have a convenient photo, but i basically just bent the PS-on pin towards the GND pin beside it, then left a very small dot of solder to keep them connected so that when given AC power, the DC automatically triggers on after a 0.5-1 second pause (presumably charging a capacitor?).  I then snipped away the plastic on the GND and 12V sockets (i left the middle one alone since it has pins for both and i didnt want to much that up somehow). \

snipped away the plastic, bent the 6 pins/socket outwards slightly, and jammed in my wires and a TON of solder to make sure everything was sealed tight. You'll need a 100W+ soldering gun to do this, otherwise you'll be sitting there for 20 minutes trying to even get the solder to melt against such thick connectors/wires.  150W soldering gun makes the task a lot easier

[pmorici]

pmorici, I have a pair of Artesyn PSU 7001138-Y000, used with IBM X-Series x3650 Server, P/N 24R2730 and would like to use RPi to control them. I think what I need is control interface slide-in board to connect it to RPi. Could you please advice me where to start if you can? Thanks!

What you need to do to interface with a Pi varies from supply to supply you basically need to design a circuit to switch the supply on and off that works with the 3.3v logic of the Pi and then if you want to monitor the status you need circuitry to shift the outputs of the supply to 3.3v as well.  Sometimes all you need is a standard NPN transistor but there are a lot of cases where that won't work for various reasons.

That supply you have would be kind of expensive to make a breakout board for, I'd estimate in the neighborhood of $60-$70 The connector alone looks to cost around $20

[xgtele]

pmorici, I have a pair of Artesyn PSU 7001138-Y000, used with IBM X-Series x3650 Server, P/N 24R2730 and would like to use RPi to control them. I think what I need is control interface slide-in board to connect it to RPi. Could you please advice me where to start if you can? Thanks!

What you need to do to interface with a Pi varies from supply to supply you basically need to design a circuit to switch the supply on and off that works with the 3.3v logic of the Pi and then if you want to monitor the status you need circuitry to shift the outputs of the supply to 3.3v as well.  Sometimes all you need is a standard NPN transistor but there are a lot of cases where that won't work for various reasons.

That supply you have would be kind of expensive to make a make a breakout board for, I'd estimate in the neighborhood of $60-$70 The connector looks to cost around $20

I intend to use server PSU as a 2nd PSU to power a couple of extra GPUs on the rig. If I use your board + HP DPS-800GBA for this purpose, how to turn both PSUs on at the same time? The first PSU is a standard PC power supply.