Pages:
Author

Topic: The Chili – 30+GH/s BFL based Bitcoin Miner Assembly - page 30. (Read 137986 times)

member
Activity: 67
Merit: 10
Would MrTeal or ChipGeek be willing to share a cad/pdf/vector file of the board. I wouldn't want all the traces or chip info or anything else propietary I just want the outline, hole locations, size of the area that needs cooling and where NOT to put pressure on the board (even thru foam). (For instant's it looks like there is an array of small something's near the center of the board on back. Should we avoid them or can they be barried in foam and a backing plate )

Just trying to cad out my holders and back plates to laser out or print.

Thanks
https://bitcointalksearch.org/topic/m.3666970
The area of the ASICs that needs to be cooled is at least 35mm x 35mm (IE, your heatsink bottom should be that large to contact all the ASICs.
The small somethings under the board are just pads where extra decoupling caps could have went if needed. There's no parts there, so just don't short them out with metal and you're good.
For where to add extra cooling, underneath the mosfets (the small black rectangle towards the edge from the grey inductors) are the most important.

Perfect! Thank you :-)
legendary
Activity: 1274
Merit: 1004
Would MrTeal or ChipGeek be willing to share a cad/pdf/vector file of the board. I wouldn't want all the traces or chip info or anything else propietary I just want the outline, hole locations, size of the area that needs cooling and where NOT to put pressure on the board (even thru foam). (For instant's it looks like there is an array of small something's near the center of the board on back. Should we avoid them or can they be barried in foam and a backing plate )

Just trying to cad out my holders and back plates to laser out or print.

Thanks
https://bitcointalksearch.org/topic/m.3666970
The area of the ASICs that needs to be cooled is at least 35mm x 35mm (IE, your heatsink bottom should be that large to contact all the ASICs.
The small somethings under the board are just pads where extra decoupling caps could have went if needed. There's no parts there, so just don't short them out with metal and you're good.
For where to add extra cooling, underneath the mosfets (the small black rectangle towards the edge from the grey inductors) are the most important.
member
Activity: 67
Merit: 10
Would MrTeal or ChipGeek be willing to share a cad/pdf/vector file of the board. I wouldn't want all the traces or chip info or anything else propietary I just want the outline, hole locations, size of the area that needs cooling and where NOT to put pressure on the board (even thru foam). (For instant's it looks like there is an array of small something's near the center of the board on back. Should we avoid them or can they be barried in foam and a backing plate )

Just trying to cad out my holders and back plates to laser out or print.

Thanks
legendary
Activity: 1274
Merit: 1004
Well it was a busy time and it's not over yet, there's lots of boxes that need to be delivered to the post office / FedEx.
Batch 3 is in the bag though, and all outstanding orders/auction wins should go out today.

We also still have some excess units after our auction, and we're going to be putting them up for sale right now. If you're interested, fire me off an email at [email protected]
hero member
Activity: 868
Merit: 1000
How many Raspberry pi's do you guys think would be needed to manage 10 cards.  I was going to run them all off a spare i3 based machine I have around here. But setting up a pi or two looks fun!

I'm not sure what the controller load is actually like.

Thanks

2 is nice
It will give you room to split the load and cover for 'WHEN SH*T HAPPENS'

1 is pushing the load limits to max.(10 boards will push the Pi to 110% load)

Edit:-  my final setup will be 16 chili's across 2 Pi's  with a 3rd Pi as a backup.
member
Activity: 67
Merit: 10
How many Raspberry pi's do you guys think would be needed to manage 10 cards.  I was going to run them all off a spare i3 based machine I have around here. But setting up a pi or two looks fun!

I'm not sure what the controller load is actually like.

Thanks
hero member
Activity: 681
Merit: 500
What is the heat (watt load) for one of these boards?  I am considering bidding in the auction a group of them and reusing some watercooling materials I have.. Just curious what I should scope for :-).

Some questions (apologies if they are answered in chilli thread, didnt have patience to read the entire 44 pages).

1) How much max Watts per chilli ?
2) Do they come preflashed? Or i need to prepare a windoze computer to run the flash utility?
3) Is Chilli supported by mainline cgminer? Or need to use a custom fork? If fork then can i see the source?
1) When they're running full out at 38-40GHs, I'm getting around 220W at the wall. Running at 30-32GH/s, 130W at the wall although the firmware to manually set the voltage and frequency is still pending.
2) They do come preflashed. You only need to update the firmware if there's a new feature or bugfix you need.
3) It's supported by mainline cgminer. I'm using 3.8.5, it has an update that actually lets you see the ASIC voltage rail instead of just reading 3.3V like all BFL based devices previously did.
member
Activity: 67
Merit: 10
What is the heat (watt load) for one of these boards?  I am considering bidding in the auction a group of them and reusing some watercooling materials I have.. Just curious what I should scope for :-).

legendary
Activity: 1274
Merit: 1004
BTW - For those running cgminer, I recommend moving to 3.8.5
It now displays the ASIC power voltage instead of the 3.3V rail, so when you first turn it on you can see it start up at 0.85V, and once it completes its self test watch it rise up as the autoclocking software ramps up voltage and frequency.
legendary
Activity: 1274
Merit: 1004
Hi everyone, time for an update.

Sorry for the huge delay in offering boards for sale. We were wanting to use an auction format for selling the fully populated boards as it really simplifies paperwork between ourselves and Canary (who provided the extra chips), but I couldn't comfortably start an auction without knowing when the boards will be done just in case there were delays.
Batch 3 is coming along though, despite challenges again. SMT is underway and should be finished this morning. QA, programming and testing will happen over the next couple days, and we should (barring any unforeseen setbacks) be done with this before the weekend.

The auction thread is here. It's a quick one (ending tomorrow at midnight GMT) since we don't want to have these sitting on the shelf when they could be mining away for you.
https://bitcointalksearch.org/topic/30ghs-chili-miners-fast-auction-50-available-367170

Please note the auction is only in BTC, if you'd like to talk about a larger order and want to pay some other way, contact me through email at [email protected] and we can discuss it.
For everyone that has already purchased boards, I'll get tracking numbers sent to you once they ship.
member
Activity: 67
Merit: 10
A double Chili please...




Sexy... just plain sexy!
legendary
Activity: 2450
Merit: 1002
think I figured out reason why this one chili runs so hot... I didnt realize in bfgminer you can look at more device specific measurements but doing "manage devices" and just scrolling through them.
Shows more voltages & another temp measurement.
So, the chili that runs hot, runs at a higher voltage than any of my other chilis... must just be very low quality chips on that one =P
newbie
Activity: 29
Merit: 0
Hashrate in cgminer is well maintained for a long time, but hashrate in pool (btcguild) is keep decreasing and stabilized after <10% drop. Is this related to ZNX command? I saw that this happens also in jalapeno. Chili might be same. How can I overcome the problem? https://bitcointalksearch.org/topic/m.2336171

My BTCGuild stats show my Chili hashrate 8 - 10% below what CGminer reports as well.

Does anyone see this with any of the other pools?

I recommend that you click the reset button in the btcguild dashboard. Reset recovers hashrate and shares in my case, which is very weird.
legendary
Activity: 2450
Merit: 1002
hero member
Activity: 700
Merit: 500
Thank you.

What I loved about these boards from MrTeal/ChipGeek is that they gave me back the fun of building things. like in the GPU days.
A ready to go ASIC box is fine, but is also boring.
sr. member
Activity: 267
Merit: 250
hero member
Activity: 700
Merit: 500
A double Chili please...


full member
Activity: 241
Merit: 101
HI, I have a chili where the main power light comes on but no LED's start. It was hashing away at half speed for a while but now nothing. Ideas?? Thanks!
Did the LEDs used to light up? Can you give a bit more information about the application, like what cooling you were using? Normally if it's hashing slower than 30 it's a sign of poor contact with the heatsink; that shouldn't normally cause a hard fault as it will throttle if it gets too hot.

Did you buy the unit used?

Hi MrTeal, I am using a Noctua NH-U9B heatsink/fan with arctic silver paste. When it was working yesterday at 1/2 speed it averaged 48c ... I did buy (2) of these on ebay a few weeks ago. One runs fine around 33ghz at 65c, the other has always ran slow and since the beginning it has always taken a while for the LED's to start up. Instead of the LED's flashing super fast when boot up like the good one this one the LED's stay on about 2 or 3 seconds throughout the whole bootup routine like when 5,6,7,8 first come on they stay on for 3 seconds, then at the end of the check when 7&8 are doing there thing they do the 2 second per light as well. Thanks for your input!

If I could get this to turn on 1 more time at 15ghz I would be happy and never unplug again.
legendary
Activity: 2450
Merit: 1002
For those that might have cooling issues another data point: I have 1 unit that runs at 65 and then it reports 120 and the miner software turns it off... and then right back on and its back at 65. This causes the unit to slowly step back up to its running speed of 36ish GH/S. I'll tinker around with the heatsinks and see if that fixes it up. I didn't think the temp could fluctuate that much and that quickly.

Yeah, one of my chili's is stlil acting up. I reseated the HSF and the original mount looked fine, the AS5 was evenly spread out on all chips. Its weird tho, when all of the chilis are warmer, this one runs cooler. Like this one will register between 62-63 when others are around 55-65C ... but when others drop to 45-55C (due to decreasing ambient temps) this one shoots up to 68-70C and its hashrate slightly increases.... I wonder if its being clocked up but still hitting some type of limit and being throttled .. vs when its running cooler - it runs even slower. Or Im wondering if one of the temp sensors is just fucked. I dont know.
I have one running at 55C and running at 39+ GH .. =)...that one is extremely happy...most of the others average between 36-38GH and this hot one sits right around 35-36GH
Possibly that one is VRM limited. The VRM temperature isn't listed but is limited to 70C. If the board is keeping voltage down to stay within the 70C envelope, it could cause that one to increase in ASIC temperature and hashrate as the ambient goes down if it then has more headroom to increase the voltage.

Well I have the same airflow setup running across all the VRM's on my chili's and they VRM's are pretty damn cool to the touch compared to no airflow ... the VRMs feel the same between all my chili's ... including the hot one.
legendary
Activity: 1274
Merit: 1004
For those that might have cooling issues another data point: I have 1 unit that runs at 65 and then it reports 120 and the miner software turns it off... and then right back on and its back at 65. This causes the unit to slowly step back up to its running speed of 36ish GH/S. I'll tinker around with the heatsinks and see if that fixes it up. I didn't think the temp could fluctuate that much and that quickly.

Yeah, one of my chili's is stlil acting up. I reseated the HSF and the original mount looked fine, the AS5 was evenly spread out on all chips. Its weird tho, when all of the chilis are warmer, this one runs cooler. Like this one will register between 62-63 when others are around 55-65C ... but when others drop to 45-55C (due to decreasing ambient temps) this one shoots up to 68-70C and its hashrate slightly increases.... I wonder if its being clocked up but still hitting some type of limit and being throttled .. vs when its running cooler - it runs even slower. Or Im wondering if one of the temp sensors is just fucked. I dont know.
I have one running at 55C and running at 39+ GH .. =)...that one is extremely happy...most of the others average between 36-38GH and this hot one sits right around 35-36GH
Possibly that one is VRM limited. The VRM temperature isn't listed but is limited to 70C. If the board is keeping voltage down to stay within the 70C envelope, it could cause that one to increase in ASIC temperature and hashrate as the ambient goes down if it then has more headroom to increase the voltage.
Pages:
Jump to: