hashrate unexpectedly drops! damaged hash units?

[tarui]

i have a 4 module batch 1 and by right it should be hashing at 100-120gh/s (MHSav and MHS5s) I made sure the heatsinks do not contact any capacitors or components




but sometime after my latest round of cooling mods (placing alu heatsinks on the square mosfets?) on the lower 3 modules (1st mode at the top is left alone)

the hashrate has dropped to around 85 to 104gh/s and my mhSav has dropped to 98500gh/s

I have a power meter attached to the power socket for this unit and it confirms the actual power draw has dropped from the usual 970-990Watts to as low as 770-940Watts

on the miner status pag, it still shows miner count 32, asic count 10. if there is a fault within the miners, does it reduce those numbers?

[candoo]

Never touch a running system.

own fault.

[tarui]

Never touch a running system.

own fault.

 

[candoo]

Never touch a running system.

own fault.

 

Did you attach heatsinks on the chips? i can see some remains of glue or thermal compound

[AdamKD]

Those inductors can go up to 125 degrees C ... they don't need heatsinks.  

Also, those inductors aren't mosfets (http://www.vishay.com/docs/34123/lp50fd01.pdf).  There's 3 mosfets (depends on hasboard being used) and 1 inductor.  The tallest of the 4 is the one I'm assuming you connected the heatsink to.  That'd be the inductor.  

There's vias for the mosfets to dissipate heat onto the heatsink on the reverse of the PCB.  Those mosfets are designed to dissipate heat to the board, not the other way around.  Adding heatsinks is not a good idea.

[Bogart]

Sounds like you've lost a module.  Check the match_work_count numbers in the cgminer api log to see which.

Reseat the power and data cables going to each module, and check the power cable pins for signs of poor contact/heat damage.

Speaking of those big inductors, mine is cracked on one of my boards, and that board is "weak" (match_work_count is about 10% below the rest).  Does anyone have a link to a replacement I can order handy?

[tarui]

Those inductors can go up to 125 degrees C ... they don't need heatsinks.  

Also, those inductors aren't mosfets (http://www.vishay.com/docs/34123/lp50fd01.pdf).  There's 3 mosfets (depends on hasboard being used) and 1 inductor.  The tallest of the 4 is the one I'm assuming you connected the heatsink to.  That'd be the inductor.  

There's vias for the mosfets to dissipate heat onto the heatsink on the reverse of the PCB.  Those mosfets are designed to dissipate heat to the board, not the other way around.  Adding heatsinks is not a good idea.

oh i see,  sorry i had no idea what those components are.

how would they be a bad idea i mean even now, the base of the pcb is still in contact with the heatsink.

i'm just adding another heatsink on other side.

[tarui]

Sounds like you've lost a module.  Check the match_work_count numbers in the cgminer api log to see which.

Reseat the power and data cables going to each module, and check the power cable pins for signs of poor contact/heat damage.

Speaking of those big inductors, mine is cracked on one of my boards, and that board is "weak" (match_work_count is about 10% below the rest).  Does anyone have a link to a replacement I can order handy?

managed to fix the problem or rather found the source of it

btw, what does "no matching work" mean? it says (after the problem has been resolved)  31
and some of the match work count (i assume to be the miner boards) show a range of frequencies, how do i know which board is which, and why do they vary so much?

[wpgdeez]

what was the problem?

[StringTheory]

What is the ambient temperature of your unit?  I was only getting around 85 GH/s out of my batch 4 Avalons until we moved them to our colocation center (72 degrees F) and now they are hashing stable at 115 GH/s.

We operate a Bitcoin ASIC Hosting center (www.bitcoinasichosting.com) if you are interested.  We have not experienced any drop in our performance.