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

[_r2h]

I see the DPS-xxxxBB have no fans. How would one cool it, if it is even needed?

I was just going to stack two on top of each other, roughly the height and width of two 120mm fans.

[sidehack]

Probably going to put a pair of three-pin headers on the board, with a PWM speed adjust tied to both fans. As with the Z750 board, this'll have both manual control, and auto control from an external signal pin so you can implement temp sensors if you want.

[volosator]

Found it. This is what HP uses in DL580G5. Exact match.
P/N number on my PSU is 438202-001

digikey

PM me for bounty  

Finally found some information on how to turn on the HP Common Slot Power Supplies:

http://www.rchelination.com/setting-hp-dps-1200fb-power-supply/

Still looking for a connector though.

Here are few more details (Pin37 has 12V standby power and Pin30 is a ground):



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I've confirmed the following connector can work, but it's not an exact match:

http://www.te.com/commerce/DocumentDelivery/DDEController?Action=srchrtrv&DocNm=1761426&DocType=Customer+Drawing&DocLang=English
http://www.digikey.com/product-search/en?pv88=667&k=1761426&mnonly=0&newproducts=0&ColumnSort=0&page=1&quantity=0&ptm=0&fid=0&pageSize=25

Here's a picture of how it matches up. Notice some of the data pins connect on the ground plane and also, there is some over hanging on the left:


I'll give a $100 in BTC to the person that can find me an exact match!

[joeventura]

Insane $8 http://www.ebay.com/sch/i.html?_trksid=p2050601.m570.l1313.TR0.TRC0&_nkw=IBM+39Y7355&_sacat=0&_from=R40

I ordered one.

1440 Watts for $16 (with shipping) pretty small risk, nice payoff!

[sidehack]

I'm probably gonna use this guy since it's a right-angle connector.

[RickJamesBTC]

Don't mount your dps2000bb upside down. Board has to be on the bottom!!  They just turn themselves off though, so that's good to know if a problem occurs. Damn these things work well! I've got each one running 10x r9-270x. I did have to do some rewiring for stability, the 5v line for the risers I sell should come from the computer controlling the cards. That lets them do a proper restart when they need to. I was trying to run them in bunches and it didn't work as well.

Someone asked about fans, I just put a little fan connector header on the side of the plug, and strap four fans to each bank of two psus. Zip ties, real ugly.

[mxz280]

I now run by DPS-1520AB Jupiter, fans are very quiet.

[BotwinBG]

Insane $8 http://www.ebay.com/sch/i.html?_trksid=p2050601.m570.l1313.TR0.TRC0&_nkw=IBM+39Y7355&_sacat=0&_from=R40

I ordered one.

1440 Watts for $16 (with shipping) pretty small risk, nice payoff!

Got 5 of those...at 140USD including shipping from the US to Scandinavia, why the heck not

[mxz280]

I now run by DPS-1520AB Jupiter, fans are very quiet.

[mxz280]

dps 2500bba two。

[mxz280]

 

[sidehack]

Update - most of the parts to build a few hundred of the Z750P boards came in today. We assembled a few for initial testing; all the signal pins work as expected, as well as the 3.3V and 5V power outputs. I used a couple Cubes as test loads, one of which had some customized cards that pulled more power than usual.

Here's the finished product, hooked up to two Cubes. The pigtail leads on the right are for the current sense signal.



These two Cubes, one running at high clock and the other at even higher clock, collectively pulled about 101% power from a single supply. The current sense signal ranges 0-1.8V for 0-60A (the supplies are rated for 62.4A) so 1.887V equates 62.9A or 101% power. The tolerance on this measurement is about 5%. It ran at this level standalone for at least 15 minutes with no issues while I set up the next test.



Here we demonstrate the load-balancing capability of these supply boards. The 12V busses are tied together, and current-share pins are linked on the header. One supply is sourcing 42.3A, the other 37A; the percent difference from mean current (39.7A) on each supply is about 6.5% in this configuration. Without the current share pins tied, one supply gave 46.3A and the other 33A, a difference from the mean of 17%.



For this test we tied one supply's External On pin to the other supply's 5V output. The 5V is powered off the 12V rail, so it comes on with the main 12V power. Flipping on the main power on this supply automatically powered on the second supply as soon as its 5V was brought up, which was so immediate that I didn't notice a delay and the two supplies came on as one. The EON signal works reliably off 3.3V signals as well. I haven't tested below this voltage but the signal is rated for 3.3-12V without issue. The 5V supply should source 2A comfortably.

These two boards have been running, averaging a 64% load, with zero problems for over four hours now.

We're not taking orders yet, need to build up stock and get bobsag3 taken care of (since he paid down on an order, to get this project rolling he gets first dibs) but possibly sometime next week we'll have some stock to start selling.

[_r2h]

Outstanding work!

[freddyfarnsworth]

Inside of those powersupplys, is just amazing value, those huge caps will hold up on a 70% load for a second or so, maybe a tad longer, just like my PC with a good HX1050 in it when power blinks around here, cheaper PS's reboot not the HX.

These are massive 400000MTBF Delta quality Powersupplies, without the $2000.00 price they were when made.
Protected lab and bank data from fail, most were/are meant to be run redundant. Have built in failover ability.

the 2000watt ones with the fans in the cases are Really Loud I mean screamin, have some 15k scsi arrays you cannot talk over the fan noise

God I love this stuff, KUDOS for pathing the way for miners.

[RickJamesBTC]

I may need to switch to power supplies with internal fans. I have four fans strapped to each pair of power supplies, and they still occasionally turn off from overheating I presume. It's annoying and I've had to spread the loads out a little. I don't want to spend more on fans than I'm spending on power supplies... What are some good ones with fans built in? I need at least 2000W per, 240v.

[sidehack]

The fans in these guys are pretty loud when left on full. These boards have a knob for manual speed adjustment, and the option for driving the fan from an external signal so you can rig up a load-based or thermostatic speed control. I haven't tried it, but in theory you could hook the external fan pin directly to the current sense pin, since high-speed for the fan would correspond with the 60A mark. Unfortunately the fan speed is not linear with voltage; it'll be at fairly low speed until you hit about 50A, which would still only be dissipating about 60W inside the supply so it might be okay. Wouldn't be very hard to rig up a simple op-amp circuit to scale the current measurement into an approximately linear fan speed signal.

I've also got parts on hand to start prototyping DPS-800 and DPS-2000 boards, so designs for those will be forthcoming. They should have compatible headers so hardware that works for one model's interface board should work with another, but that doesn't mean I recommend mixing and matching supplies in a single bank.
Right now the main effort will be in manufacture of the first batch of these boards. We're also planning on developing a control board which can be linked to several (not sure, 6 to 12 probably) boards and handle them all together, with individual supply health indicators and power readings, as well as composite power output data. The backplane guys will probably like that.

Three of these Z750P load-linked will supply about 2200W and they're rated for 240V input. Just saying. Two of the DPS-800 would supply 2000W off 240V, and have internal fans.

[RickJamesBTC]

Not worried about loud, and I don't want to have to stack even more power supplies. I'm just finding that fans zip tied to the cases isn't providing sufficient cooling. I guess I'll need to get some much more powerful fans and maybe make some ducting to ensure the airflow all works as it needs.

[freddyfarnsworth]

Not worried about loud, and I don't want to have to stack even more power supplies. I'm just finding that fans zip tied to the cases isn't providing sufficient cooling. I guess I'll need to get some much more powerful fans and maybe make some ducting to ensure the airflow all works as it needs.

If you cant hear em they are wrong ones. need 12v 2 or 3 amp 3800-4000+ rpm fans push 140cfm or so.
120x120x25 or x38 but they are thick, would only need one or two of those - as original design pulled outside air from case thru the PS to exit.
They had xeon passive sinks on the CPUS so it was pre warmed air, with ALOT of flow.

Here some links to get you goin - I would use even heavier for this application on the 2000watt jobbers


AFB1212SH
http://www.ebay.com/itm/Gateway-960-8006323-Delta-AFB1212SH-Back-Case-Chassis-Fan-30-Day-Warranty-/170822936596?pt=US_Computer_Case_Fans&hash=item27c5d72814

http://www.ebay.com/itm/Gateway-Delta-AFB1212SH-Back-Case-Chassis-Fan-/281046185168?pt=US_Server_Fans_Cooling_Systems&hash=item416fa874d0

[PMB]

Update - most of the parts to build a few hundred of the Z750P boards came in today. We assembled a few for initial testing; all the signal pins work as expected, as well as the 3.3V and 5V power outputs. I used a couple Cubes as test loads, one of which had some customized cards that pulled more power than usual.

Here's the finished product, hooked up to two Cubes. The pigtail leads on the right are for the current sense signal.

Hi,

the PCB looks really nice! Keep up the good work! Do already have prices for these?

regards,

P.

[r00t$]

Nice work sidehack. I'm anxious to order a couple!