Pages:
Author

Topic: IBM Bladecenter 2880W Breakout! - page 2. (Read 2634 times)

member
Activity: 111
Merit: 10
March 30, 2017, 11:24:41 PM
#9
Hokay,

In order to protect these boards and make them mimic the wiring that this board is interfacing; we have added several processes.

The back side of the 2880W connector has been sealed with a high dielectric strength potting compound, made specially for electronics.


This layer is essentially impenetrable.


Quite industrial.

These boards have also been coated with silicone specific to lead free processes to prevent whisker growth and other contamination over time.


We can verify the coating via black light and this layer may be repaired at any time.

Cheers,
-Optim
member
Activity: 111
Merit: 10
March 30, 2017, 05:59:01 PM
#8
A bump for the GPU guys.  Get that ETH!

We have added potting compounds to the exposed side of the board for safety and reliability.

Pics incoming.

-Opt
legendary
Activity: 4172
Merit: 8075
'The right to privacy matters'
December 22, 2016, 12:29:39 AM
#7
I got one last week  I have been really busy but finally started testing it.

I am using  a dps 2980 unit  with modified fans.   This unit  does plat and is my most efficient psu.

We moved all gear from buysolars barn/loft into his basement today.

this allows for  heating his home .   It also means  I had to mod fans to cool the psu.

For now I am running these 2 units.



about 24.6 th
the meter shows 12.2 volts no draw
and 12.0 volts with about 80% to 85% draw

say 2400 watts used of 2980 watts max

remember the 2980 is a true plat  with the fan mods I did to it last year.

more to come  but I will be slow due to holidays

This is first time I ran 2 s9's on a 2980  should do it with ease  (I hope)

member
Activity: 111
Merit: 10
December 20, 2016, 07:03:55 PM
#6
Thank you very much Optim!

I got the package a week ago but did not have enough time to post the photos.

http://imgur.com/a/HWyjr

I totally recommend your board though I do not have a thermal gun atm to measure the temperatures but I touched them by hand and they are way lower than my body temperatures.

I got a great deal from you + you ship internationally and it was cheap for me for such a big package!


Looking forward to deal more with you.

Cool!  I always advise against touching active electrical components, but if you ever get the heat gun out, do let me know how it goes.  I am curious of the performance at different ambient temperatures.

-Optim
full member
Activity: 143
Merit: 100
December 19, 2016, 05:23:43 AM
#5
Thank you very much Optim!

I got the package a week ago but did not have enough time to post the photos.

http://imgur.com/a/HWyjr

I totally recommend your board though I do not have a thermal gun atm to measure the temperatures but I touched them by hand and they are way lower than my body temperatures.

I got a great deal from you + you ship internationally and it was cheap for me for such a big package!


Looking forward to deal more with you.
member
Activity: 111
Merit: 10
November 30, 2016, 01:34:36 AM
#4
Yes.  We will be potting shipped units with epoxy for insulation, but there is 220VAC on the input line.  Always use caution.  That first article photoed above is not sealed yet.  That is fressh off an assembly bench.

-Optim
legendary
Activity: 4172
Merit: 8075
'The right to privacy matters'
November 30, 2016, 01:31:59 AM
#3
Hello fellow miners.

I wanted to update the community on my work to optimize the 2880W breakout efficiency and overall thermal performance.  We have put together a design and am currently adding this breakout board to my manufacturing list.  It is a pain in the behind, but we figured it out.

This board has 22 PCI-e 6-Pin output for 12V power.  There are four, 3 pin fan connectors and a power switch with a secondary jumper for network switching.  

Behold the 2880W Breakout:

Main Breakout Board


Power Input Board



This is what it looks like on the supply.  I have also added holes for zip-ties so that you may secure the boards together in an efficient manner.

On Supply Top:


On Supply Bottom:


I am currently running extended thermal tests, but here are my preliminary results.  My initial testing was done with limited airflow and an ambient of 20.7C.

Ambient:


The testing was done at approximately 1900W (I need to move everything to the machine/tooling side of my shop for 30A breakers) with the test stand my team has put together.

Test Stand:


This test stand loads each PCI-e pin tested to ~150W.  This is used for initial thermal testing as well as individual unit testing to ensure functional and safe products.  During this test, I measured heat using FLIR and used a 4-wire (6 digit) DMM to verify power losses (though still an approximation).

The heat measured as follows at 1900W:

Connector Side (cables coming through)


Back side (closest to 2880 PSU)


We had some air blowing, so the static temperatures may be a bit higher, though this construction matches that of the 4k design and we have never seen more than a 40C rise out of those units at 3kW loading.

To give you a reference point, this is what our load stand looked like under FLIR:



To measure and approximate power losses, I have measured the voltage drop at 1900W, on the furthest connector from the power supply output.  For the first article, we measured 15.215mV and 13.000mV drop on the return and +12V pins respectively.  At 1900W, we are consuming about 158.3A with a total drop on this board of 28.215mV.  This puts the design at around 4.47W loss at 1900W.  That is about 1/3% loss (with margin).

My largest concern with this breakout is the exposure of 220VAC.  Anyone that tries these boards must exercise caution and common sense.  These boards and supply should never be handled without powering down the entire system.

Anyhow, my team is currently operating out of www.price-technology.com if anyone has any questions or would like to try this product.  They cost $70 in singles and I am sure it is worth it.  I am also happy to answer any questions on the forums.

Cheers,
-Optim






Okay are the three silver dots hot?

I left the photo on it.
full member
Activity: 143
Merit: 100
November 30, 2016, 12:49:42 AM
#2
Wohoooo.

Finally these are out.

Love your boards separate design.

Looking forward to receive them!
member
Activity: 111
Merit: 10
November 29, 2016, 11:57:06 PM
#1
Hello fellow miners.

I wanted to update the community on my work to optimize the 2880W breakout efficiency and overall thermal performance.  We have put together a design and are currently adding this breakout board to my manufacturing list.  It is a pain in the behind, but we figured it out.

This board has 22 PCI-e 6-Pin output for 12V power.  There are four, 3 pin fan connectors and a power switch with a secondary jumper for network switching.  

Behold the 2880W Breakout:

Main Breakout Board


Power Input Board



This is what it looks like on the supply.  I have also added holes for zip-ties so that you may secure the boards together in an efficient manner.

On Supply Top:


On Supply Bottom:


I am currently running extended thermal tests, but here are my preliminary results.  My initial testing was done with limited airflow and an ambient of 20.7C.

Ambient:


The testing was done at approximately 1900W (I need to move everything to the machine/tooling side of my shop for 30A breakers) with the test stand my team has put together.

Test Stand:


This test stand loads each PCI-e pin tested to ~150W.  This is used for initial thermal testing as well as individual unit testing to ensure functional and safe products.  During this test, I measured heat using FLIR and used a 4-wire (6 digit) DMM to verify power losses (though still an approximation).

The heat measured as follows at 1900W:

Connector Side (cables coming through)


Back side (closest to 2880 PSU)


We had some air blowing, so the static temperatures may be a bit higher, though this construction matches that of the 4k design and we have never seen more than a 40C rise out of those units at 3kW loading.

To give you a reference point, this is what our load stand looked like under FLIR:



To measure and approximate power losses, I have measured the voltage drop at 1900W, on the furthest connector from the power supply output.  For the first article, we measured 15.215mV and 13.000mV drop on the return and +12V pins respectively.  At 1900W, we are consuming about 158.3A with a total drop on this board of 28.215mV.  This puts the design at around 4.47W loss at 1900W.  That is about 1/3% loss (with margin).

My largest concern with this breakout is the exposure of 220VAC.  Anyone that tries these boards must exercise caution and common sense.  These boards and supply should never be handled without powering down the entire system.

Anyhow, my team is currently operating out of www.price-technology.com if anyone has any questions or would like to try this product.  They cost $70 in singles and I am sure it is worth it.  I am also happy to answer any questions on the forums.

Cheers,
-Optim




Pages:
Jump to: