Topic: The Chili – 30+GH/s BFL based Bitcoin Miner Assembly - page 23. (Read 137904 times)

You rock. I've replaced a few of those on the BFL devices; bitch in heat.

How did you do it on the Chili? And what is with those chips?

C

Yes second hand.

Made a few changes, hopefully it will stabilize

Out of curiosity, how did you verify that the temperature sensor wasn't working correctly? It is specced to work linearly down to -40C, but I haven't actually tested that.

replacing the FTDI chip fixed it!!!!!!!!
 

Just one LED is on, and it just stays on? Was this one you got second hand from someone?
PM the serial number and any other details on the symptoms if you can.

Hi there, I have several chili's, One of the new ones I just received is having issues. When its first plugged in the LED's 5,6,7,8 come on and stay on and never go off.

Any suggestions? Thanks in Advance!~

Just received a Chili today and within minutes of it starting to hash it starts reporting that it is receiving an empty string from the temp sensor and the last Diag LED goes on, its in a 40 degree F room with three fans on it and heat sinks.

Any ideas?

Can I get an RMA?

I wonder if that's it: The normal BFL chips don't take so much of a hit on an error because they all run all over the place. The Chili runs itself in lock-step, so when it has an error it's a total fail on all chips and not just the one chattering.

Which brings up a point: Is the error due to a heat issue? It's amazingly constant, maybe the microcode runs things fast enough to have that many errors or something.

But yes, 38 is pretty darn good. Might just enjoy that.

6% of 40Gh is 2.4Gh  so    40/40/38  is quite respectable.

I found the reason. U25 temperature sensor doesn't work properly under low temperature. I covered it with a small tape and now all chilis are okay.

Still hashing this morning at 39+Gh, peaking at 40gh, but the third statistic in BFGMiner is still about 2gh less (the one that is shares submitted to pool and accepted). Still not quite sure why, the rejected shares is <1% and errors are at a solid 5.5%.

Have heat sinks coming for the FETs, yes they are hot even with a fan on them. But man this $49 water cooling thing from newEgg is incredible. I have a dead jalapeno board coming in without chips so I can use that to see how/if I can fit water cooling on my J units. Well worth it.

C

My chilis in garage are not working if cold. Yesterday they stopped at 3 am and today same. I rebooted everything and LEDs are blinking well, but my raspberry pi doesn't recognize the devices. They may be gonna recovered Morning as the yesterday case. How are your chilis? Are they okay in -7 Celsius degree?

I can tell you that even if you think the heatsink is well seated, it still might need to be redone. Every single one of the chilis that were running around ~20GH needed the heatsink redone along with the pad underneath.
They all used the Evo212 Heatsinks. You just have to completely take it off, redo it and try it again. Almost always fixed the issue. The one chili that it didn't fix completely still ran about 5-7GH better ~27GH, I just assume it got a bad batch of chips on that one.

Is it possible the heat sink itself is broken (bad heat pipes or something)? I've found that if I run my water block at too fast a flow rate the temps on the board go way down along with the hash rate.

Try a water block cooler like that $50 one on NewEgg. 36gh hashing as reported by Eligus, 39gh reported locally (why the difference I do not know...)

Any advice will be appreciated
One of my chilis is running a little slow
http://imageshack.com/a/img202/6720/n29r.png
I know this has been discussed before But fresh ideas are welcome
as you can see by the picture its hashing in the 20's. The heat sink is an evo212 and it's well seated
http://imageshack.us/scaled/large/24/jpaw.jpg
the power supply puts out or at least is supposed to put out 750W
and its feeding a single and the two chillis,
the three units are connected to an unpowered usb hub
The speed won't change if I plug them directly to a usb port on the computer
and the temp. on the questioned chilli won't go over 60 degrees.
............................................................................... ...
Well after some research I found out that a lot of engines are off.
Does any one know how to force the the chips to enable more engines?
http://imageshack.com/a/img17/7939/b6oy.png

That's the one. He hasn't, but I have. I applied it to the heatsink, not the ASICs, and a very thin layer that has contacted the tops of the ICs but never the sides or board.

The chili flash utility gives me the following:

DEVICE: Chili SC
MANUFACTURER: MrTeal and ChipGeek
FIRMWARE: 1.2.14e
CHIP PARALLELIZATION: NO
QUEUE DEPTH:40
PROCESSOR 3: 1 engines @ 172 MHz -- MAP: 0080
PROCESSOR 4: 11 engines @ 194 MHz -- MAP: FFE0
THEORETICAL MAX: 2.30 GH/s
ENGINES: 12
FREQUENCY: 192 MHz
CRITICAL TEMPERATURE: 0
TOTAL THERMAL CYCLES: 0
XLINK MODE: MASTER
XLINK PRESENT: NO
OK

At this point I'm giving up. Normally, I'd return it, but it's hardly Keefe's fault this happened. I think simply the stress of removing and replacing the heatsink may have done it in. it won't mine and I don't know how to fix it.

If anyone would like to buy the mostly dead board off me for experimentation, PM me. Be aware, the ARM chip works, but everything else is unverified/unknown. All I changed was putting stick on memory cooler chips on the power mosfets, and ultrasound clean it when I discovered debris under the edge of an ASIC (which may, but probably may not, be responsible)... Failing a buyer, I'll mail it to MrTeal so he might learn from the failure if it has any lessons that apply to other boards.

Also, Keefe has been more than fair, and this in no way represents any reflection on him. He sold me a known glitchy board, and stated so up front. It worked here for 18 hours before dying while I wasn't present. I appreciate the supportive PM that suggested I get my money back, but it's not Keefe's fault.

This was the one with the short on the 3.3V rail, right?
I've never seen anything like that. I had one that came in that had one ASIC that didn't work; it had a crack along one side of an ASIC and another ASIC had a crushed corner. That was pretty obvious physical damage though.

It's too bad that the ultrasonic bath didn't help any further. Do you know if Keefe used something conductive/capacitive like Arctic Silver?

Ok thanks I see it happy Xmas

See the second post of the thread. You need to have the FTDI drivers installed to use the flash utility.

The factory ones with Windows. Fastest way to get them is to go to device settings, find the chili, uninstall drivers, and remove drivers from device. Then unplug replug and it should install the right ftdi ones.

Then use BFGMiner.

So how can I get the stale shares count down here? I'm at 39gh shares, but only 35 recognized as valid by the pool (eligus). Cut down my difficulty? Get a better pool?

C