Topic: 2000W Power Breakout Board (DPS2000BB to PC-Ie x12) - page 7. (Read 13413 times)

I avoided the voltage adjustment based on the issues I've read about with the A6's.  I find adjusting the supply while measuring the unit voltage yields the best performance.  Though my experience is limited.

I am assuming the 2500 model will support this, but I have yet to test.  Looking back at NotFuzzyWarm's link, the header and pinouts appear interchangeable.  It looks like there is a standard wrapped around that specific header.  Someone correct me if this appears incorrect.  I am assuming you are looking to use this 2500 model for a dual S7 configuration (~2200W) and the margin on this board should accommodate that.

If these singles recoup my expenses and do as well as I am hoping, I see no reason not to develop a dual board for power transfer.  If it comes to that, I will let you know. 

PM me for details if you want to try the single on a 2500.  I am somewhat interested in modifying my setup to test this case.

-Optim

When you get time put up the web site , I Polly will need three , two in about three weeks or less then one more a week later.. won't need escrow as long your in the in US.

So. Do I get to adjust volts with a simple switch?

Will they work with the 2500 model psu.

Would you consider making a dual board?

I may want a few singles if they work with the 2500 model psu.

All in all very cool - and I love that there's some other good hardware guys coming into the space, it takes a village.  

About half of my 2000BB's are running the 4k board, and while I haven't had any failure yet, I certainly have had failures in the past... but as he pointed out, there's a practical advantage to pairs of those 2000BB's - but what I did with my single boards was just reverse the orientation of the miners so the plug was on the inside instead of out, and then you can have cables going out in either direction, and still use the same general configuration of back-to-back 2000BB's.  The only downside of this is that the fans aren't blocked on the outside, like they are with the 4k board, but if you're using high SP fans I haven't found that it matters much.  On one of my units I experimented with using foam insulation strips around the perimeter of the fan mount, and it made virtually zero difference, not worth the time it took to do it.

NotFuzzyWarm brought up the 2500BB, would these breakouts be compatible?  If so, that might be VERY awesome, as you could easily run 4 A6's on 2x 2500BB, getting almost full utilization out of the PSU.  I imagine the extra thermal headroom your design has would make a BIG difference in this case.

So I'll take a couple of those to check out - but 5 isn't a great number because I'm going to use them in pairs, but how about either 4 or 6?  PM me and we can work out the details if that's acceptable to you.

@NotFuzzyWarm and @Prelude

Thank you for the post.  It was close.  I'm sure the DPS2000BB curve is similar to this.  There is likely a crest around 50-60% that tapers off slightly as the supply approaches 100%.  I might check one supply over here to get a ball park between 50% load and 100% load.  Will update this thread later in the week.

-Optim

Well poop! Yer, right, everywhere it's about the 2kw supply - except for the performance curves..... Post edited to reflect that.

Those are efficiency numbers for DPS2500BB,  NOT DPS2000BB. Big difference, and the numbers can't be applied to the 2000w version.

Optim, here ya go: Discussion about the 2kw supply - but curves for the IBM DPS2500BB https://litecointalk.org/index.php?topic=5854.0

I come from an area of engineering where there is an emphasis on reliability.  This has led to a bit of paranoia in my personal ventures.  In your mining topology, if a single supply were to fail, three miners would drop out of the system.  With the single board solution, I am rocking a 1:1 supply failure to miner failure ratio.  It makes me a bit more comfortable; to each his own as they say.

I also find that the efficiency of these supplies taper off with max load (This appears to be a phase shifted bridge supply).  I have yet to see an actual efficiency curve for the DPS2000BB (I have yet to measure as well...), but I imagine the dual supply configuration might actually hurt the W/hash slightly.  This is purely speculation, I need to dig for a while to know for sure.

-Optim

Yes.  Those are just resistors.  I ran 14 of them in parallel to get ~72mOhm load.  With the supply across them, they will dissipate the 2kW.  That is why I threw the fan in the mix.  Those resistors will get very, very hot during this test.  During this test, this subjects the power adapter board to the full load (~166A) to measure worst case temperature and dissipation.

I measured the voltage across the load with a DMM.

The temperature was measured with FLIR and a National Instruments thermocouple.
http://www.ni.com/product-documentation/10632/en/

I've thrown a current probe onto the power input to verify efficiency and calculated the current based on the measured load resistance of 72.2mOhm.  This isn't the most accurate calculation, but it should get me right around what I need.

I do enjoy a science fair and I rarely get to fully test things at work with the corporate overhead. 

I have a spare 50 PCBs sitting on my desk and I was hoping to unload.  When that stock runs out, I would have to order more and the two week extra bit of time would get into the mix.  I do not have all of the parts for assembly, but those may be acquired within a day or two and be built in time (the approximate 1-2 weeks before shipment).  PM me for details if you are interested.

I must say that I am impressed with the attention to detail, although I would say look at the double 2000 board he makes and you will see the way it becomes the New standard.  A lot easier to hook up fans to 2 units and also a lot more versatile for hooking up miners.  A single 2000 can run 1 s7 while the double can run 3 and stay cool 

Interesting - I use almost exclusively Jabberwocks boards and haven't had any issues with them, but I'm generally running well under the limit of the PSU's.  They definitely do seem to run warm, so when I made my fan adapter for the dual DPS2000BB's, I actually had it hang over the board to blow some air on it - easy enough to do and worth the extra effort.

Can you explain a bit about your test rig?  I'm assuming those are some industrial resistors just basically converting power into heat, correct?  I like that you actually did pure testing with it, and didn't just hook up a S7 and call it good.  

Also, I'm a bit confused about what exactly you're offering - do you have these boards available now, or do they ship in 4 weeks?

I am a newbie.  I am mostly playing around with this out of interest.   

I may be able to play around with that load stand on Wednesday-ish.  I will let you know if I have some more exact numbers.  My rule of thumb when there is no airflow is that the hotter it is, the less efficient (this does change with surface area).  I also like the reliability aspect.  Cooler things last longer.

-Optim

That make sense, if you ever get some watt waste numbers at different temperature, that could be cool too. I'll stay subbed to the thread.

Cheers, you're new but you seem to know your shit. Welcome (again?) to the forum and thanks for the lesson.

Well, the issue scales with load.  The power lost in these systems is equivalent to I^2*R.  Where I is the current transferred and R is the resistance of the board.  The resistivity of copper also scales with temperature.  The hotter these boards get, the less efficient they are.

That being said, my concern is always I.  Since I is squared in that power loss relationship, R needs to be as small as possible (Since we can't change I without scaling the system).

-Optim

Thanks for humoring me. In depth technicalities is a bit beyond me but i like amassing the general/conceptual knowledge. 100F seem pretty close to body temperature i believe? Thats like 37C? That seem cool to me.

So you mean its a power waste issue? But it would be how much? 2-3watts waste?

 

@sidehack

As do I.  More importantly, I can appreciate the efficiency that level of 'overkill' brings to the system.  That is assuming the designer strayed away from reliefing (which I realize many of these deisgns have).  That appears to generate many other issues for these systems.

-Optim

I'm not sure on the details (Novak did the PCBs for PSU boards) but I think it's 3oz outside 4oz inside. Could be 2oz outside 5oz inside. Many vias to help transfer internal heat to the surface. Also, I like overkill.

@VirosaGITS

I spent the beginning of my venture playing with the J4bberwock's.  They were difficult to amass in quantity and given my ambient conditions those boards rose an additional 30+degrees Celsius.

This is what I was seeing at full load for the J4bberwock:


Versus what I am currently seeing on my unit:


Given the lead times and my quick test, I made a lot specifically for myself to eek out a bit of efficiency.  

@sidehack

I needed the 6-pinners.  14oz seems a tad much.  Is that 3.5oz per layer and full coverage on all layers???  I would expect a 4-layer board around that weight to be overkill (to a point where the connectors might be tested).

-Optim

I'm really hoping my boards were the "out of stock" rather than the heat problem ones. We build 'em with 14oz of copper between the 4 layers in each rail, which tested fine up to 196A draw in about 95F ambient. But no PCIe 6-pins.