Topic: The Chili – 30+GH/s BFL based Bitcoin Miner Assembly - page 12. (Read 138078 times)

Restarts are part of the gamble. Only time will tell.

I have a theory that the Chili was designed and prototype tested in an "American Air conditioned summer workshop" 
Now that they are out in the wild (of the rest of the world) There are some "exceptions to the standard" being found.
The next firmware release will probably bring things back on track.   

If we are still hashing them in the coming summer then they will have had a good testing cycle.
The unfortunate reality is that they will be way out of date by then ( in 6 months they may be the "graphics cards" of the asics)
Such is the speed of development of mining hardware. 

Is the firmware open source yet? I would like to mess w/ voltages ...

The trick is to get them about 60c and stable.
If you get some gentle cooling on the back of the board, then the temp will rise again to around 68-70c and you will get another 3Gh

Thats what ive been doing. O well new PSU comes in today. Will test more when I get in from work.

If the above sequence works on your board, "then you need the slow rise firmware".

The "standard" installed firmware is trying to start too fast.

The 1v1 limiting firmware will also still try to start at the "standard" speed. (So this is really only for boards that hash too fast 39-40Gh then crash).

MrTeal has said earlier in this thread, that a new firmware will be released soon, incorporating the slow rise and a manual voltage setup, (watch this thread)

The only symptom that sounds different for me, they never show SICK in bfg -- they just disappear after a few minutes.

I only have 2 cards. They both do the same exact thing as this. They start normal and ChiliFlash shows everything within normal parameters. Start mining and they ramp up to about 22-26ghs for anywhere from 30-60 sec. If I don't do the warming, they start throwing more errors, hash rate falls until I get a flood of the temp errors and then they'll disappear from the bfg console.

If I kill bfg and go back to ChiliFlash, the device still opens, but it shows only 2 engines still enabled. Power cycling the Chilis is required to bring them back.

Not to confuse the thread, as so many have the 1.1v fw working for them, but for some reason neither of my boards worked better with it -- they both actually sermed worse, as o barely had time to pick up the hairdryer before they'd die.... I decided to reflash the "standard" Chili14e fw before giving up on them and to my surprise, they were stable enough to be able to use the warming technique. Once I got this worked out, based on all the previous cooling/mounting/shielding research of others on this thread, now they run nice and steady about 32-34ghs -- up for well over 48 hrs now.

I have 2 old Arctic USB fans rigged on a hub pointing down onto the pcbs for long term cooling -- hairdryer on cool to cool it down during the startup dance...

As Ronin4bits said:-

Get the hairdryer ready, then start it hashing. As soon as it's hashing, start blasting the top of the fets with it. Give it a minute -- (temp should rise to 68-70c)the hash rate should stabilize for about 30 sec, then it'll start to fall -- (turn off hairdryer) (temp will drop to 60c) switch over to cool air and start cooling the back of the pcb -- when you get good cooling on the back, hash rate will start climbing along with the reported asic temp -- when airflow keeps this around 68-70 degrees, it should run.  (you may need a secondary fan to take over the cooling of the back of the board)

+1 the hairdryer, af_newbie.

Get the hairdryer ready, then start it hashing. As soon as it's hashing, start blasting the top of the fets with it. Give it a minute -- the hash rate should stabilize for about 30 sec, then it'll start to fall -- switch over to cool air and start cooling the back of the pcb -- when you get good cooling on the back, hash rate will start climbing along with the reported asic temp -- when airflow keeps this around 68-70 degrees, it should run.

I've been average 32-33ghs for over 48 hours now after getting this one going like this.

One twist - I'm running the "regular" Chili14e fw - the 1.1v limited version would crash this board after about 30 sec of hashing, no matter what... 

Good luck.

Your other board sounds like it needs the "hairdryer" mod to get it started.

For the first one, I would double check that there doesn't appear to be any board flex that would cause one chip to lift off the heatsink. Alternately, open the unit in putty or another terminal program as soon as it starts and while it's doing its self-test (the top two LEDs flashing for ~20s) send it "ZlX". That will report the temperatures of all the dies. Note if one appears to be a lot higher than the others.

For the second board, it does sound like it might have the same issues that many boards from that group buy are experiencing. We're working on firmware as a workaround for some of those issues, and hope to have it out soon.

The one is a little strange, but nothing too crazy. The board starts at a 0.85V and a lower frequency setting during initial turn-on and self-test to provide a factor of safety if there is a problem with heatsink contact. If there is one chip that has poor or no contact we want to watch that before it causes damage.
You won't see those numbers increase until you actually start to accept work from the mining software.

The direct PCIe will be better than a dual 4 pin to PCIe, so use the one coming directly out of your power supply.

Just out of curiosity. I have been running the single chili off a single pcie 6 pin coming from the PSU would that be a problem? I can try and find a dual 4 pin molex to 6 pin pcie adapter if needed. Anyone think this might be an issue? Still up and running since my last post. But i am sure by the time i wake up tomorrow it will be down. It has not stayed up more then 12-16 hours yet. I would ideally like to run both chilis off 1 500 wat supply if I can. Doesnt matter if its on the pcies themselves or if I need to get adapters. Please let me know. Thx!

Check the voltages on the PSU while it's running. It's a group regulated design, so if you're pulling enough current on the 12V line without a load on the 5V line it's possible it's shutting off due to overvoltage on the 5V line.

Weird. I'm running a single and a chili on a Corsair CX500 no problems. Make sure the side vent isn't covered, but that unit should easily handle the load.

C

It's a corsair cx500.  I just cam home on my lunch break and was able to get it to turn on. I'm sure it will stay on for a couple hours but can go down at any thime. Then I will be unable to get the psu to turn on again for a long time. Thx again.

I used normal Radio shack thermal grease on my chips, Runs at 38/35gh with a sedion water block, which isn't bad. I really should take it off and try artic silver type stuff but I'm lame.

As for replacing the FTDI, might be a cold joint, replacing the FTDI chip on one of my jallies was a pain in the *rear*. I used leaded solder balls on the reflow, put 1-2 .45mm balls on each pad and used my Aoyue with 400c, very very low flow rate to keep from blowing things off nearby. And a pre-heater under the board set to 350F for about 10 minutes before starting.

I've found it to be best to think of the air coming out of the Aoyue as forming a "bubble" of heat off the end as opposed to a stream. Low air flows allow the hot air to transfer into the joint as opposed to going around the joint.

C

I don't use the pads. I have 8 miners all using thermal compound on the ASIC chips. I use thermal adhesive on the heatsinks under the FETS.

I was using a hot-air station at 400.  The current behavior is after is after I re-went around each side for 30 seconds and also reinstalled a nearby small brown smd cap that got blown off by the airstream (not sure if it happened then or the first time around when I originally replaced the FTDI).  My hands are way to imprecise (and my eyes too) for hand-soldering the teeny-tiny legs!  I could add a bit of flux and try again, but I'll check the usb area first.

BTW, I definitely appreciate all the responses as I work on this board - you've got to be the most responsive and helpful developer/vendor I've run across here - kudos and thanks!

Yes, that's really interesting. I'd agree it might be a loose connection or a soldering issue if it crashes when you bump the table. Possibly a cold solder joint on the FTDI chip? The grounds especially are difficult to solder effectively as they wick heat away very quickly especially if you're using a fine tip. I would go over the leads again, and if possible heat the whole board up before you do that.