Pages:
Author

Topic: 2000W Power Breakout Board (DPS2000BB to PC-Ie x12) - page 6. (Read 13413 times)

hero member
Activity: 687
Merit: 511
Just a quick update, I ordered some from him and just got them today - I haven't had time to really inspect them to any great degree but they look to be good quality boards, but so do the Jabberwock ones to my untrained eye and I haven't side-by-sided them yet.  I hooked two of them up to one of my 2x DPS setups (had to flip the PSU's to get them to work how I wanted) - and they're working fine so far.

Hopefully I'll get some time this weekend and I'll post some pictures and do some comparisons between them and the Jabberwock boards I have.
member
Activity: 111
Merit: 10
I'm interested in picking a few of these up. I have some GPU rigs that I would like to be able to reboot remotely, and I like the fact that I can hook these up to relays and reboot them remotely.

If you break them up into smaller even lots I'll take 2...

PM me, we can work out the details.  I am rerunning some tests to post some more results tonight.

-Optim
full member
Activity: 212
Merit: 100
I'm interested in picking a few of these up. I have some GPU rigs that I would like to be able to reboot remotely, and I like the fact that I can hook these up to relays and reboot them remotely.

If you break them up into smaller even lots I'll take 2...
member
Activity: 111
Merit: 10



Op Are you in the U.S.?

Yes I am.  I also like that rig.  The ducting is classic!  Cheesy

-Optim
legendary
Activity: 3822
Merit: 2703
Evil beware: We have waffles!
During the summer the area those are in gets up to low 90'sF and all is happy there.

Like your setup.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'

I will have to look into the 2980s.  If they are that much better, it may be well worth the investment.

I don't think the 24 hour test is necessary.  I would say 1/2 hour to 1 hour (whenever steady state temperature has been achieved).

I need to acquire said hardware for that test.  Not likely, given my current build.  I am finding that more copper isn't necessarily better.  Again, the bottleneck (on the breakout) is at the board to connector interface.  I would have to inspect/measure the sidehack board to make a clear case to switch, but in my case I didn't see a 6-pin PCI-e connector breakout like this from sidehack.  I will dig and see what I can find.

-Optim
Or, to gain a few 10's of watts in eff, ya can always lose the PSU fans entirely like this http://i.i.com/cf5IRyB.jpg
Kills a couple birds with 1 stone Wink I'll look tomorrow to see if the pcb holes for the connector pins match their profile  to reduce the path thru solder or are just round.

yeah that would most likely work okay.

this works on the screamers in the 2980's or 2880's

https://bitcointalksearch.org/topic/update-large-review-of-finksyj4bbrwock-server-psus-compared-to-atxpsu-photos-up-1284563

At first I liked the dual board from j4bberwock  ,

but once I did these fans in the 2980 and got true platinum  power ratings.  I go with 2980's over everything



you only need 1 of the two parts the none filter part.  these are a perfect match same company



back of the 2980's this mod on the 2980's also works on the 2880's saves 40 watts per psu or 1 kwatt a day
cost is 22 bucks per psu.  note the grills are needed my 22 price included grills    


legendary
Activity: 3822
Merit: 2703
Evil beware: We have waffles!

I will have to look into the 2980s.  If they are that much better, it may be well worth the investment.

I don't think the 24 hour test is necessary.  I would say 1/2 hour to 1 hour (whenever steady state temperature has been achieved).

I need to acquire said hardware for that test.  Not likely, given my current build.  I am finding that more copper isn't necessarily better.  Again, the bottleneck (on the breakout) is at the board to connector interface.  I would have to inspect/measure the sidehack board to make a clear case to switch, but in my case I didn't see a 6-pin PCI-e connector breakout like this from sidehack.  I will dig and see what I can find.

-Optim
Or, to gain a few 10's of watts in eff, ya can always lose the PSU fans entirely like this http://i.imgur.com/cf5IRyB.jpg running 3 of my s5's off of a DPS 2000. PSU is on the suction side of course.
Kills a couple birds with 1 stone Wink I'll look tomorrow to see if the pcb holes for the connector pins match their profile  to reduce the path thru solder or are just round.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'



Op Are you in the U.S.?
member
Activity: 111
Merit: 10
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .

That's not the board in question, not to mention 4080w AC is only ~ 3630w DC which isn't full load so not relevant. It would be nice to have that board tested as well, though.


After all the testing I did with the 2980 the 2880 the dual 2000/4000

I go with the 2980s with my fan mod.
I would be really interested in a board for them as they are more efficient psus then the 2000.
Also the fans blow air directly onto the back of the breakout boards so they never overheat.


What would be nice to see is direct 24 hour tests.
With ops board vs other boards.
A s-7 pulling 1300 watts on this board would pull 1339 watts on j4bberwock board if there is a three percent difference on the boards running the 2000 psu.

And if sidehack has more copper then. He should be under the ops.

It is pretty easy to do the testing.  You need 1 s-7 the three 2000 watts break out boards the psu and a meter




I will have to look into the 2980s.  If they are that much better, it may be well worth the investment.

I don't think the 24 hour test is necessary.  I would say 1/2 hour to 1 hour (whenever steady state temperature has been achieved).

I need to acquire said hardware for that test.  Not likely, given my current build.  I am finding that more copper isn't necessarily better.  Again, the bottleneck (on the breakout) is at the board to connector interface.  I would have to inspect/measure the sidehack board to make a clear case to switch, but in my case I didn't see a 6-pin PCI-e connector breakout like this from sidehack.  I will dig and see what I can find.

-Optim
member
Activity: 111
Merit: 10
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.

-Optim
Do correct me if I'm wrong but if you are talking about remote sensing at the load these supplies can do that using the same pins called our for voltage adjustment, pins A1 and A2 so v drop across the PCIe leads can be compensated for by tacking a light gauge twisted pair to those pins and to one of the PCIe sockets where it joins the hashboard. Better yet - a separate pair coming from the board's input power planes to the remote sense in...

The pic does label it as 'sense' and to me at least that means remote load sense vs v.adj.

Just as with the HP DP1200 Commonn Slot Proliant supplies I have found it IMPOSSIBLE to get hold of a schematic for the power backplane these PSU's are used with. That would take all guesswork out of these. Prints for the various server boards or at the server boards backplane pinouts - no problem to find . grumble brumble.

You are correct, though the result is similar.  I have been adjusting with the trim pot internally, though I am planning to implement your Vsense cabling soon.  That should give decent dynamic load performance; however, I have been on the fence about committing to the Vsense version.  These switchers generally run a bit more efficiently (At the miner side, core voltage) when the input voltage is a bit higher.  I have been adjusting the supplies using the aforementioned method under max load.  In this case, I get a high +12V rail when the loads are less than 100%.  I am hoping, this makes the miner slightly more efficiency.

Yes.  No one likes the internal circuitry known.  IP and all.  Maybe when one of these supplies fail, I will do some reverse engineering.  Heh.  Lips sealed

-Optim

I can send you a failed DPS-1200-FB1 if you'd like to muck around with it. I've got about 9 of them. Maybe you can every point me in the right direction in regards to resurrecting them!  Smiley

I can almost guarantee it will be cheaper/easier to just replace those.  If I have some spare time, I might consider this, though the 1200's were generally less power than I was looking for.

-Optim
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .

That's not the board in question, not to mention 4080w AC is only ~ 3630w DC which isn't full load so not relevant. It would be nice to have that board tested as well, though.


After all the testing I did with the 2980 the 2880 the dual 2000/4000

I go with the 2980s with my fan mod.
I would be really interested in a board for them as they are more efficient psus then the 2000.
Also the fans blow air directly onto the back of the breakout boards so they never overheat.


What would be nice to see is direct 24 hour tests.
With ops board vs other boards.
A s-7 pulling 1300 watts on this board would pull 1339 watts on j4bberwock board if there is a three percent difference on the boards running the 2000 psu.

And if sidehack has more copper then. He should be under the ops.

It is pretty easy to do the testing.  You need 1 s-7 the three 2000 watts break out boards the psu and a meter

legendary
Activity: 1596
Merit: 1000
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.

-Optim
Do correct me if I'm wrong but if you are talking about remote sensing at the load these supplies can do that using the same pins called our for voltage adjustment, pins A1 and A2 so v drop across the PCIe leads can be compensated for by tacking a light gauge twisted pair to those pins and to one of the PCIe sockets where it joins the hashboard. Better yet - a separate pair coming from the board's input power planes to the remote sense in...

The pic does label it as 'sense' and to me at least that means remote load sense vs v.adj.

Just as with the HP DP1200 Commonn Slot Proliant supplies I have found it IMPOSSIBLE to get hold of a schematic for the power backplane these PSU's are used with. That would take all guesswork out of these. Prints for the various server boards or at the server boards backplane pinouts - no problem to find . grumble brumble.

You are correct, though the result is similar.  I have been adjusting with the trim pot internally, though I am planning to implement your Vsense cabling soon.  That should give decent dynamic load performance; however, I have been on the fence about committing to the Vsense version.  These switchers generally run a bit more efficiently (At the miner side, core voltage) when the input voltage is a bit higher.  I have been adjusting the supplies using the aforementioned method under max load.  In this case, I get a high +12V rail when the loads are less than 100%.  I am hoping, this makes the miner slightly more efficiency.

Yes.  No one likes the internal circuitry known.  IP and all.  Maybe when one of these supplies fail, I will do some reverse engineering.  Heh.  Lips sealed

-Optim

I can send you a failed DPS-1200-FB1 if you'd like to muck around with it. I've got about 9 of them. Maybe you can every point me in the right direction in regards to resurrecting them!  Smiley
legendary
Activity: 1596
Merit: 1000
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .

That's not the board in question, not to mention 4080w AC is only ~ 3630w DC which isn't full load so not relevant. It would be nice to have that board tested as well, though.
member
Activity: 111
Merit: 10
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .

Definitely C.  Commercial grade PCBs can usually run up to about 120-130C without delaminating.  

I can rerun the test to verify, but that number does not stand out as outlandish to me.  Especially since that temperature is local to the reliefs around the connectors.  That might explain why some people have had burnouts at the connector.  TO give a bit more background, I have been measuring with an ambient of ~30C.  A 70C rise with reliefs is to be expected at that current level.

I will modify my test and see if I can get a hold of a 2500BB to acquire some thermal data.  

-Optim

Well the jacks are warm running avalon6s. I use 15 awg cables .
Do you mean the back of the boards where they attach to the psu?
Or do you mean where the cables attach to the jacks?

I have never felt they were hot enough to burn my hand. But since I know the Avalon 6 pulls a lot of juice I always went with the higher grade 15 awg cables.
I do agree they warm up.

The hot spot I have been most concerned about is right under the PSU connector.  The reliefs in the copper appear to bottleneck the current with a relatively high resistive path.  If you look at the image I previously posted, most of the heat is there.  I imagine the same problem will be true (underneath the PCI-e connectors) when a small amount of PCI-e connectors are used relative to current draw.

The cables should not warm up.  This will be local to the breakout board itself.  You may not notice this until you use FLIR around the equipment.

-Optim
member
Activity: 111
Merit: 10
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.

-Optim
Do correct me if I'm wrong but if you are talking about remote sensing at the load these supplies can do that using the same pins called our for voltage adjustment, pins A1 and A2 so v drop across the PCIe leads can be compensated for by tacking a light gauge twisted pair to those pins and to one of the PCIe sockets where it joins the hashboard. Better yet - a separate pair coming from the board's input power planes to the remote sense in...

The pic does label it as 'sense' and to me at least that means remote load sense vs v.adj.

Just as with the HP DP1200 Commonn Slot Proliant supplies I have found it IMPOSSIBLE to get hold of a schematic for the power backplane these PSU's are used with. That would take all guesswork out of these. Prints for the various server boards or at the server boards backplane pinouts - no problem to find . grumble brumble.

You are correct, though the result is similar.  I have been adjusting with the trim pot internally, though I am planning to implement your Vsense cabling soon.  That should give decent dynamic load performance; however, I have been on the fence about committing to the Vsense version.  These switchers generally run a bit more efficiently (At the miner side, core voltage) when the input voltage is a bit higher.  I have been adjusting the supplies using the aforementioned method under max load.  In this case, I get a high +12V rail when the loads are less than 100%.  I am hoping, this makes the miner slightly more efficiency.

Yes.  No one likes the internal circuitry known.  IP and all.  Maybe when one of these supplies fail, I will do some reverse engineering.  Heh.  Lips sealed

-Optim
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .

Definitely C.  Commercial grade PCBs can usually run up to about 120-130C without delaminating. 

I can rerun the test to verify, but that number does not stand out as outlandish to me.  Especially since that temperature is local to the reliefs around the connectors.  That might explain why some people have had burnouts at the connector.  TO give a bit more background, I have been measuring with an ambient of ~30C.  A 70C rise with reliefs is to be expected at that current level.

I will modify my test and see if I can get a hold of a 2500BB to acquire some thermal data. 

-Optim

Well the jacks are warm running avalon6s. I use 15 awg cables .
Do you mean the back of the boards where they attach to the psu?
Or do you mean where the cables attach to the jacks?

I have never felt they were hot enough to burn my hand. But since I know the Avalon 6 pulls a lot of juice I always went with the higher grade 15 awg cables.
I do agree they warm up.
legendary
Activity: 3822
Merit: 2703
Evil beware: We have waffles!
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.

-Optim
Do correct me if I'm wrong but if you are talking about remote sensing at the load these supplies can do that using the same pins called our for voltage adjustment, pins A1 and A2 so v drop across the PCIe leads can be compensated for by tacking a light gauge twisted pair to those pins and to one of the PCIe sockets where it joins the hashboard. Better yet - a separate pair coming from the board's input power planes to the remote sense in...

The pic does label it as 'sense' and to me at least that means remote load sense vs v.adj.

Just as with the HP DP1200 Commonn Slot Proliant supplies I have found it IMPOSSIBLE to get hold of a schematic for the power backplane these PSU's are used with. That would take all guesswork out of these. Prints for the various server boards or at the server boards backplane pinouts - no problem to find . grumble brumble.
member
Activity: 111
Merit: 10
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .

Definitely C.  Commercial grade PCBs can usually run up to about 120-130C without delaminating. 

I can rerun the test to verify, but that number does not stand out as outlandish to me.  Especially since that temperature is local to the reliefs around the connectors.  That might explain why some people have had burnouts at the connector.  TO give a bit more background, I have been measuring with an ambient of ~30C.  A 70C rise with reliefs is to be expected at that current level.

I will modify my test and see if I can get a hold of a 2500BB to acquire some thermal data. 

-Optim
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.

I ran j4bberwock dual 2000/4000 at 17 amps or 4080 watts for 24 to 48 hours multiple times.

There is no way the breakboard I had  was 104c or 220f

Maybe 105 f to 115 f

Not sure where the 104c  temps are picked up from .
legendary
Activity: 1596
Merit: 1000
The only 2000BB breakout boards I have are made by sidehack, and those work perfectly with the 2500BBs that I have. I'm 100% certain anything that works with the 2000BB will work on it's younger brother, the 2500BB. The pinouts are the same. The only question is, can the breakout boards handle the power? I know sidehack's overbuilt (I also love overbuilt stuff  Grin) boards can handle it no problem, and it looks like these would too. I'd be hesitant with the j4bber board if they're hitting 104c at full load on the 2000BBs. In fact, I probably wouldn't even use them up to 2000w if they're really hitting those temps.
Pages:
Jump to: