Topic: ANTMINER S3+ Discussion and Support Thread - page 159. (Read 710164 times)

Thank you

Checked the top of the R47 chokes and got a even 0.81v on all of them.

looks like an entire chain is down. try checking the voltage if you have access to a multimeter - its likely the DC/DC regulator for the first chip in the series isnt fuctioning (should see 0.7V across the large grey cube-like inductor) and as a result the entire chain does not function.

check if there are any damaged resistors or missing bits

Antminer S3+ Batch 9, 8 chips showing as non-operational.

Tried removing the thermal compound, cleaning and re-apply non-conductive thermal compound without any luck. Tried switching the cables around, etc..

Anyone else having similar issues? Should I request a RMA (Anyone have any experience with the RMA processes? Do i need to send the entire unit back?)

https://i.imgur.com/cTYcpTx.png

We run Batch 8 and 9 units over clocked to 237.5 and 243.5 in groups of 3 on the EVGA 1300w G2 PSUs. They pull exactly 1205w off the wall. So that's 400w each.

Strato

I don't know.  If you  find what you need to test I'd like to see the results!

Each board has an LM75A in its center.  Between the two rows of ASICs.  The IC has LM75A printed on its top.

One can't flip the rear heatsink to change the curve relative to the fans as the top fin is above what would then be the bottom - prevents bottom of heatsink from reaching the floor.

I think you are onto something here .... thanks for sharing!


You are definitely onto something!

I had to upgrade a few of my S1's with the kit bitmain offered and rather than apply thermal paste to the chips before screwing on the front heatsink, I applied thermal pads instead. the unit is hashing away without issue, and the cost of the heat pads for each unit came to 1 GBP (negligible cost though a bit fiddly to cut to size thus takes a bit more time to apply than paste). My query is, do you think any heat pad would do for the board-heatsink contact in case I can get it in the 150mm x 250mm dimension(s)? It would then be a simple case of curving out the spot where the LM75A resides. Talking of which, how do I tell which is the LM75A? Is there only one on each board?

My training in heatsink building back in the late 1970's informed me that there should be no voids between the device and the heatsink and that the purpose of heatsink compound was to fill any voids as the two surfaces aren't perfect but also that the compound moves heat less well than the metals of the devices.

When I repasted my first S3 hashingboard I found the factory installed heatsink thermal compound on the inner heatsink to be what looked like stripes of machine applied compound the length of the heatsink in two groups of stripes, targeting the upper and lower rows of ASICs.  The individual stripes had air gaps between them and the two groups of strips had a larger stripe of air gap between them:

_________________________________
_________________________________
_________________________________
_________________________________

_________________________________
_________________________________
_________________________________
_________________________________

Following my training, I applied a thinner, evenly applied, layer of thermal paste to the entire heatsink and ground plane of the board and mated the surfaces.  This worked so well that that board maintains the lowest registered temperature.

This  would be ideal if there were no hashrate problems.  Ideal because it keeps the fans running at a minimum speed so as to conserve energy.  If there are hashrate problems one wants cooler ASICs.  The fudge to get cooler ASICS is to get higher fan speeds.  To get a higher fan speed to have cooler ASICs one might apply a thin even coating on the entire inner heatsink but then clean the area in the middle where the LM75A is situated.  The LM75A is what measures the board's temperature.  By applying a full even layer of heatsink compound then cleaning where the heatsink would make contact under the LM75A, the LM75A then is given to have a deceptively higher temperature than otherwise.  Remember, the heatsink is cooler than the board.  When that LM75A is allowed to read higher due to no heatsink compound in that spot, the higher registered temperature dictates a higher fan speed.  The higher fan speed means less chance of lower hashrate.

Yes, from the ethernet connection end chain 1 is on the right and generally runs hotter.

The ASICs are numbered on the board as if a snake or dragon starting with the top ASIC beneath the fan connection.

Hi, I've always one "x", always at the same place, without overclock:
oooooooo ooxooooo
oooooooo oooooooo

I've clean thermal paste twice, very carefully the second time.
I've change PSU with another one. Both are Corsair 750W.


Now I want to check the bad chip more carefully.

Can someone confirm that the chain in WebUI are the same order as this picture:
https://i.imgur.com/bu6BjO0.jpg
I want to confirm that my bad chip is on physical chain #1 before unscrew and cleaning again.

And where is the bad chip on this picture :
https://i.imgur.com/KWnIrsdl.jpg

(sorry if bad english, I'm french)

Thanks.  Will retry.

Discarded shares has been discussed here ad nauseam. Simply ignore it, it means absolutely nothing.

I didn't do anything too special, just put westhash pool info and that's it.  My BTC address showed up in stats immediately.

I do have huge number of "discarded", about 30:1 to "accepted" but hashrate shown on westhash is roughly the same as shown in ant UI.  Try setting difficulty using "d=xxx" password, I use "d=128" for S1 and "d=256" for S3 however S3 usually switches to diff 512.  I also initially had better luck connecting to nicehash, which is strange (I'm in US, however on the east coast).

FYI after putting thermal compound on the inner heatsink, before mating it with the hashing board, if one takes a bit of solvent and cleans in the area of the LM75A such that there is no thermal compound there, the board will read a higher than expected temperature by about a degree and the fan will run faster.

When will Batch 10 go up for sale???  I'm ready to get a couple more.

The specs. you request have been posted here before, I bought my kilowatt meter for like $25.  But I don't want to overclock them so you'll have to search.  I would assume overclocked, no more than 475-500w per antminer using 4 pci-e connectors.

Yes please post what a oveclocked s3 does I am dying to know. I run two of these over clocked each on a corsair pro hx1000 (1000w) gold rate psu. Well because i got the psu's on ebay for around $75-$90 each I know I know the server psus are cheaper but they are so hard to find specs on or info on wiring up the needed cables etc.
Maybe something to try one day as a backup in case

I have 1` s3
and 1 s3+ (another one  on the way)

I guess I need to order a killawatt meter- figured though save myself a bit of cash and find someone else who has the readings. If you got those readings miners everywhere will love you, because i cant find that info anywhere.

  

I have 3 on one psu and the psu is plugged into the meter.  Current average for 3 is 1112w off of 120v outlet.  So Bitmain claims 355w per so I just took an average of all.

360watts at the wall  at stock speed. Have you checked the at the wall wattage @ 231.25, 237.5, 243.75 and 250 That info would be a big help Thanks in advance