Looks like the caps are 10mm diameter with a 5mm lead spacing.  Height is around 20mm, but there is plenty of room for taller ones if needed.

I'm getting these for replacements:

http://www.digikey.com/product-detail/en/UHW1C152MPD/493-6909-ND/3664326

I suppose if they wanted to be jerks they could try to do that.

Really though I didn't hack it at all until after the failure had occurred, and had they given me a timely response/replacement I would not have needed to hack it.

While they're at it they could send me a fixed backplane with good caps, but I doubt that will happen anytime soon given how swamped they appear to be, so I'll again have to take the matter into my own hands.  At least caps are relatively cheap compared to a power supply, though before I can order replacements I will need to determine the lead spacing (5mm vs 7.5mm), which means removing the backplane, which means more downtime...

It's alive.  I finally had success hacking my dead X3 PSU so that I could attach an ATX power supply to it, and it's up and mining again, bulging caps and all.

I'm using an older Antec 1200W unit to power it.  At first I saw it drawing 1460W from the wall, and after recovering from my surprise that it kept running, i quickly backed the settings off from 1200MHz to 1000MHz, and it settled around 1180W at the wall.

I had at first attached four 6-pin PCIe sockets each coming off a pair of 12-ga wires, but the connectors got too hot to touch after just a few minutes, so I added two more sockets.  Now they just get warm to the touch.

https://i.imgur.com/0e1DVVG.jpg
https://i.imgur.com/AI6Acj8.jpg
https://i.imgur.com/uadu548.jpg

Yes I know (now) I got the colors/polarity backwards.  After a magic smoke incident I fixed it outside the unit, not wanting to resolder all those wires on the PCB.  I really need to work on being less retarded when it comes to reading pinouts.

Soldering all that was a giant pain in the ass.  I used 3 irons at once, and used a propane torch to help heat them more hotter.

Still no response from BA about a replacement PSU.  I sure hope they don't ask me to send back the dead/hacked one first.

I've received no human response yet from BA to my support ticket requesting a PSU replacement for my failed unit.  It's been about 45 hours now.

Since the unit is offline anyway I thought I'd take it apart and see if I could find another way to get power into it.  No luck on that front yet, but I did find something else.

I saw that there are 10 decent-sized electrolytic caps on the backplane, and I scanned the tops of them out of habit for signs of bulging, and guess what...out of the 10 caps, 5 had bulging tops, 2 of them to the point where the electrolyte was leaking out.  That's after just 13 days of use.

They're labeled as Sanyo brand caps, rated for 16V 1500uf, 105c.

I guess I should think about ordering up some replacements.  I wonder what kind of shape those caps are in in the still-operating miner...

I tried swapping cables/switch ports, but that did not help matters.

Unfortunately I have a new problem now.  One of the power supplies seems to have died.  When powered-up the red and yellow LEDs come on briefly and then go out, and then repeat about every half second.

I tried swapping PSUs, and the problem does follow the PSU this time.

I also tried powering it using 120V instead of 240V as normal.  This lengthens the cycle time to about 2 seconds, but does not cause it to power on as normal.

I also tried powering it up outside of the miner with no load, but the behavior is the same.

The sticker on the PSU says it's "Sunshine Electronics model AP188".

I guess I'll try contacting BA and hope they will replace it under warranty.  It ran for 13 days before dying.

So yeah, these guys never fully repaid the DRK owed to me.  Out of ~1160 DRK owed, they repaid ~260 DRK with the last payment on 29-May, and now they won't even answer my emails.

My X3 seems to have a problem.  Every 24-48 hours one of the two units stops mining.  Always the same unit.

The PSU and NIC LEDs look normal, the fans are running, and the blue "heartbeat" LED visible through the left front fan is blinking as normal.

So far as abnormal LEDs, there are yellow LEDs visible through the rear fans (I assume these LEDs indicate that the hashing modules are idle), and the blue activity LEDs on the front do not blink as the normally do when mining.

The unit is unresponsive on the network.  It does not respond to ARP requests, and of course is equally unresponsive to ICMP pings and HTTP(S) requests (since those require ARP to work).

Yesterday I swapped the PSUs between the two units.  The problem did not follow the PSU.  The same miner still has the same problem.

 

(Both units are running software V1.8.)

That worked.  I guess I'd still had the https:// prefix in there.  Thanks.  :tips hat:

Thanks.  I tried that, but navigating /LCD/ just gives a 404 error, and /lcd/ gives 403 Forbidden.  (Yes, I tried logging in to the regular web UI first.)

I'm running software version 1.7.

DHL delivered a big heavy box from BA today, with "0.5kg" printed as the weight on the shipping label.

Pics here: http://imgur.com/a/xGtWI

Some observations:

It seems that one X3 is actually two 3U rack devices that make ~1TH/s each.  I had hoped for 3TH/s total as was alluded to earlier in the thread regarding "compensation", but I only received the two units.

They have six buttons on the front and a blank area that I guess was meant to be occupied by a large LCD.

When I plugged them in and checked the DHCP server on my router, they registered themselves as "android-a_16_digit_hex_number".

The units came with software version 1.0 installed, and when I did "check for updates", they updated themselves to 1.4, then 1.5, and finally 1.7.

The units detect each other on the LAN, and allow me to manage both units from a single web UI instance, which is convenient.

One of my units came clocked at 1100MHz, and the other at 1000MHz, which I can't seem to change.  I discovered a small "Hz" icon in the sidebar on the left of the web UI, which offers me the choice of 800, 1000, 1100, or 1200, but the buttons seem to have no effect.

They appear to be running OpenSSH server, but I'm not able to login via SSH.  The user:blackarrow login that works for the web UI does not work for ssh, nor does root:blackarrow.

Mine is supposedly on its way.  Order #10x for a single X3.

https://i.imgur.com/7xCvoUB.png

1:3.3 compensation would be more appropriate.

Difficulty on Feb 24th when our miners were supposed to ship: 3,129,573,175
Difficulty today: 13,462,580,115

(Probably more by the time they actually ship.)

This was on one of my CoinCraft Desk modules, which draws around 200W.

Replacing the connectors was a pain in the butt.  It took me about 4 hours from start to finish.

It can't.  At least not for very long.  It's a seriously retarded design.

The connectors don't always make perfect contact either.  I have had to replace several of these 6-pin connectors on my AntMiner S1s after they fried themselves, and they only move around 190W each.

Well don't just sit there.  If you've got some "big rigs" on order, get the necessary infrastructure in place now, before delivery.

I'm sure you can buy a few 1500W space heaters for maybe $20 each to simulate the load in order to test the setup.

Order #102 here (one X3), placed on September 26th.  Mine is a trade-in order where BA is supposed to request my old Lancelots when it's nearing time to ship.  I have not yet received such a request.

Difficulty is now 375% of what it was on February 24th when they were supposed to ship. 

What calendar are you using?

So far as I know, historically no mining ASIC maker has ever hit their planned energy efficiency targets.  Avalons first chips consumed about 150% of the planned amount, where BFLs consumed 450%.  I don't have numbers for Hashfast and Cointerra but I hear they had the same problem.

I would advise Alpha to prepare for similar results, and to plan accordingly when designing the PCBs.  Use extra copper to make fatter traces able to handle more amps, design in pad provisions to add more voltage regulators and power connectors if they turn out to be needed, etc.  The same goes for the cooling system.

There seems to be some confusion about what "tapeout" means.  Tapeout refers to the creation of the mask set which will later be used to fabricate the chips.  As I understand it, it is not a long process and is done inside of a day, though it is one of the most expensive steps in the production process.

Once the mask set has been made, the masks are then used to fabricate the chips themselves one layer at a time onto the silicon wafers, using a photolithography process.  This is the part that takes 4 to 8 weeks.