Wow - industrial DC fans are generally cheaper than PC fans, and should cost nowhere near that much. Try ebay: http://www.ebay.com.au/itm/NMB-Dell-Optiplex-GX270-Case-Fan-DC-Brushless-Model-3615KL-04W-B86-12V-2-10A-/181444763166?pt=LH_DefaultDomain_0&hash=item2a3ef35e1e

I found the biggest difference makers were heatsinks on the vregs (on the back, opposite the asics) and fans suited to purpose. Speaking of which - why would you go with a 90CFM fan on the back? I don't think I ran a single fan under 130CFM+ front and back.

I ditched all my S1s in the spring, but I'll see if I can dig up some stats . I will say that the quality coming out of their factory is sh!t, some miners would be ace and others complete dogs. The cynic in me suspects that they would pull the worst miners from their farm stock and sell those out for each batch, pulling only from new stock if need be.

Newer firmwares (e.g. Kano's binaries) introduce features like temperature control/temperature shutdown. These features could come in handy as we head into the summer..

Binaries are here: https://github.com/kanoi/cgminer-binaries/tree/master/AntS1

The image could have been corrupted, or at least the section where the webui files are stored. To fix this you can download a new disk image here: https://bitmaintech.com/files/download/initramfs.bin.SD-20140410.tar which you can write to a microSD using win32diskimager.exe (or similar program).

Once you're up and running you can apply a patch, Kano's version is (imo) the best version out there: https://github.com/kanoi/cgminer-binaries/tree/master/AntS2

I had 15 S1s between 5 AX1200i - zero problems. Sold them all off in the spring as thermal management was going to be an issue.

Beaglebones chew up SD cards when running from flash? That's interesting...

Aren't patriot about the cheapest cards you can buy?

The keying is to prevent issues with existing equipment which used the same connector series but with different pinouts and/or voltage levels. There are no differences aside from that (assuming you're comparing similar grades of parts).

Yeah, there's a bunch of factors that they'd have to consider. The "standard" molex minifit jr pin is currently rated at 9A, (IIRC the first series were rated at 4A/P, the baseline has been increasing steadily since they were introduced years ago). That said, there's no telling which pins would be used in the connector, they could instead use the top of the line molex pins (13A/pin) or cheap knockoffs (??/pin). Then there's the quality of the connecting cables used, which can't really be planned for aside from setting the bar very low.
 

The only reason I'd prefer those crappy little controllers is they don't chew up SD cards like RPIs can/do. In a perfect world I'd much prefer something with embedded flash storage, either a custom design ($$) or maybe a beaglebone black.

Yeah, that or it was "too difficult" to source an appropriate high-cfm fan with a long 4 pin lead to reach the junction board from both fan locations.. Or to just lengthen the cable or use an extender...  

Dragonminers have virtually no support (at least from the developer/engineering side). The source code for their modifications to cgminer isn't even released, so don't expect anyone from the community to be able to step up and fix/improve it either.

Molex has several series of minifit jr connectors. Their HCS series is spec'd in at 13A per pin: http://www.molex.com/molex/products/family?key=minifit_plus_connector_system&channel=products&chanName=family&pageTitle=Introduction&parentKey=minifit_products

And I should buy it from you rather than directly from Avalon why?

I don't think they should be DNPs, as they're the output caps for the buck regulator (the same part (TPS5335) you used on the Bitfury 25GH boards).

If anyone's interested here's the relevant schematic section.



Edgar: from what I've seen the heatsink compound used is of passable quality - the actual application during assembly varied wildly (too much, too little). The HS screws were also installed erratically, resulting in voids under the heatsink where no or little contact was made. I also had several where cutting/tapping fluid (which hadn't been cleaned out properly) oozed out from the heatsink screw holes and mixed with the compound, resulting in a runny mess.

Don't get me started on the HS design itself..

Avalon. People are pissed at them (some justifiably) but they are shipping out miners from stock. https://ehash.com/

Hi,
I'm not terribly familiar with programming FPGAs - and I'm sure I'm not alone on that. Are there some simple instructions describing equipment required, and how to program?

Do you foresee any problem programming the FPGA on 3rd party 'clone' avalon miners (e.g. lketc's avalon3 miner)? They seem to have much different electronics layout and it is quite possible that they tweak the source to match their design (it's impossible to tell for sure, as they do not provide source code).

You're doing it wrong.

I've had no problems with the supplied openwrt images for either the rpi or WR703, with one caveat: FW upgrade through the UI is not functional in the rpi build, so must be done manually. I'll take the minor inconvenience over dealing with the WR703's limitations thanks.

NMB-MAT 4715KL-04W-B40/50: http://www.nmbtc.com/pdf/dcfans/4715kl.pdf.

A few of my ants were shipped from bitmain with these fans (which had obviously seen significant 'burn-in test" time..) before they switched over to cheaper/more available alternatives.

Personally, I find that these things are a perfect balance between noise and cooling capability. In general fan noise doesn't both me; it's the resonance and/or squeal that some exhibit - I've never had a 4715KL which has this particular problem. Loud yes, but tolerable.

Unfortunately it can be difficult to source PWM capable versions of these fans - eBay usually has them for sale by scavengers who loot various retired IT appliances (e.g. PSUs, servers, UPS) that make use of this fan. Not for cheap either - as they seem to be in high demand.

Digikey stocks the standard 3 wire version, and again not for cheap

Given that you're overall hashrate isn't terrible, maybe it's just a flaky board? The trick would be to figure out which one.

A couple of ideas:
- check cgminer debug log (or 'kernel log' as it's referred to in the OpenWRT UI) and see if a particular chain is reporting more errors than others. If you find a particular chain is having issues investigate why - e.g. is it getting too hot? You can also see this if you log into the shell and watch the cgminer output (screen -r). It may look to you like gibberish to start with.
- Pull all cards aside from one, and then begin systematically adding each card back and verifying that the overall system is performing as expected (tedious)

FYI: The temperature reported for each card is the output of a single IC on each board. I forget the exact part, but it's the same as is used on the S1. If the board is not reporting temp it's possible that this IC is damaged, missing, or has bad solder joint(s).

On first login pw will be root / blank (blank = no password). It will be set with a pw afterwards.

FW development seems to be handled exclusively by Canaan Creative, and he linked to their software repository in the OP. There is a wiki entry which you may want to read that outlines some of the basics: https://en.bitcoin.it/wiki/Avalon2