The Chili – 30+GH/s BFL based Bitcoin Miner Assembly

[ChipGeek]

Any updates on batch 2 production status?
0-25%
25-50%
50-75%
75%-finished

75%-finished 

[1713582416]

1713582416

[philips]

My two Chili against a Little-Single.
BAL 0  is the LS.


BAL  0:  max 48C 3.27V | 31.24G/31.33Gh/s | A:  235213 R: 1250 HW:     896 WU: 435.0/m
BAL  1:  max 60C 3.29V | 34.38G/34.33Gh/s | A:  230372 R:   624 HW: 32387 WU: 419.2/m
BAL  2:  max 63C 3.29V | 40.38G/40.34Gh/s | A:  277958 R:   697 HW: 25154 WU: 516.9/m

[Spieder]

Mr. Teal & ChipGeek,

Thank you both!

[BrimStone]

Thank you both once again for making this happen.  I know it was a ton of work but the results are fantastic!

If you want to see them in action, here is a photo you will either like or cringe when you see.

-BrimStone

[MrTeal]

Thank you both once again for making this happen.  I know it was a ton of work but the results are fantastic!

If you want to see them in action, here is a photo you will either like or cringe when you see.

-BrimStone

Thanks Brimstone. It was definitely a challenge but I'm really glad we were able to get your issues resolved.



On another note, after we closed batch 2 orders we had some excess boards without ASICs, so we teamed up with CanaryInTheMine to get them all built up with the intent to sell them off.
The those units are now for sale in the auction subforum.
https://bitcointalk.org/index.php?topic=334841.0

[Bogart]

What would it take for there to be a batch 3?

[MrTeal]

What would it take for there to be a batch 3?

I would guess at least 100 boards or so.

[agibby5]

What would it take for there to be a batch 3?

I would guess at least 100 boards or so.

Can the prices for end users be lowered enough to make it more enticing?  I know there are some required costs like manufacturing, etc...

[i3luefire]

  thanks for the boards. we have had 4 out of the 13 assembled and running. 23gh, 36gh, 34gh and 39gh so far. when we get the rest of the heatsinks on and the rack set up I will put up a pic or two. we are running them with thermal paste only using the twin turbo ii coolers with lots of small heat sinks on the back side and one 90mm fan on the back side heat sinks.

[MrTeal]

thanks for the boards. we have had 4 out of the 13 assembled and running. 23gh, 36gh, 34gh and 39gh so far. when we get the rest of the heatsinks on and the rack set up I will put up a pic or two. we are running them with thermal paste only using the twin turbo ii coolers with lots of small heat sinks on the back side and one 90mm fan on the back side heat sinks.

Maybe wait until you're all set up (since making setting up a new one is more efficient that fixing an old one), but you might want to reseat the heatsink on the one that's running at 23GH/s. It's very likely that it's throttling due to poor temps on one of the chips.

[MrTeal]

Can the prices for end users be lowered enough to make it more enticing?  I know there are some required costs like manufacturing, etc...

Depends on volume, but probably not. A lot of it will depend on what kind of price you can get for the ASICs themselves, as at list price they're more than everything else combined. For our part of it, an order of 100 boards is enough that it takes you out of the very expense realm you pay for runs of 10-20 or so for prototypes and the ludicrously expensive cost of doing a few, but it's not nearly enough to get down to even the costs we saw in our first batch (which was 3x larger than our second batch).

[i3luefire]

is 70c a good temp for these boards?

[MrTeal]

is 70c a good temp for these boards?

It's pretty normal.

Remember, we're actually reading the die temperature as opposed to a board temperature like BFL is doing. The temperatures are going to be higher than you would see with see on a comparable BFL device and way higher than you'd see on something like an Avalon that thermocouple glued to the heatsink.

[asjfdlksfd]

Hello Mr. Teal,

you told in the past there was a problem with the board by mounting the asics.
What was the problem exactly? I understood the most of the tin-soldering-balls of the asic has to been flooded the holes completly or over the strip-lines?
I've only the photo in my mind which you post here. It looks a little bit there or the mask was not printed to the asic chip area to stop flooding the tin-solder over the whole strip line. In this case there is only a little bit missing of tin-solder.

I want to prevent me from this problem so maybe compensate the situation by a little bit soldering paste filled to the holes ?

Another question, as I understood in this discussions was not only the cooling of the chips is essential, also the cooling of the power chips is very essential. So it looks to me it must not be a big EVE 212 cooler needed only. Maybe a smaller is good, too but I should check that the power chips get there own heatsinks.
I plan to glue a heatsing on top and a big heatsink plate on the back for the power chip area.

I've now ordered two of the most more cheap Scythe Setsugen Rev. 2 Grafikkartenkühler VGA
https://www.google.de/images?output=search&sclient=psy-ab&q=Scythe+Setsugen+Rev.+2+Grafikkartenk%C3%BChler+VGA&gbv=1&sei=fjSHUuX_EMLItQam-oHACg&hl=de&sa=X&oi=image_result_group&ei=gDSHUr5_jcGzBsGUgOgK&ved=0CDIQsAQ
which could be buy in germany from < 22 € and plan to mount them in a way the air flow will be go through the top heatsinks of the power chips.
The heatsink itself is know to handle ~ 200 W which should be enough to cool the asics.
On the back I want to mount a heatsink as bfl did this with the little singles.

I hope it will work ;-)

But in fact, can anybod give me the mechanical dimension from the board and the holes so I can prepare/check my 19 Inch server rack case for mounting some of them? I would like to bring the case in a air conditioned server room which is only possible if I mount them in a server case as I allready did this with my knc miners ;-)
I want to calculate how much asics I can mount in one case. I think 5 should been possible but I need
clarify that.

Cheers...

[oldbushie]

I'm hoping we see firmware updates soon. Mine are still hashing away nicely but it would be nice to see a small bump up in hash speed with any efficiency fixes. Plus I still have that one Chili that randomly hits 100% on bootup unless I powercycle and another that simply takes forever to boot up. I haven't needed to reboot anything for a week or two though.

I think right now all my Chilis are at 60C or thereabouts, maybe 70C at most.

[Mudbankkeith]

I'm hoping we see firmware updates soon. Mine are still hashing away nicely but it would be nice to see a small bump up in hash speed with any efficiency fixes. Plus I still have that one Chili that randomly hits 100% on bootup unless I powercycle and another that simply takes forever to boot up. I haven't needed to reboot anything for a week or two though.

I think right now all my Chilis are at 60C or thereabouts, maybe 70C at most.

If your chili's are running at 60c then try this:-

Place a fan above the mosfets, the tempreature readout will go up to 70c and your hash rate will increase from 32Gh to at least 35Gh maybe even 38Gh.

The chili processors likes to run at 70c but to achieve this the power modules need cooling.

[MrTeal]

Hello Mr. Teal,

you told in the past there was a problem with the board by mounting the asics.
What was the problem exactly? I understood the most of the tin-soldering-balls of the asic has to been flooded the holes completly or over the strip-lines?
I've only the photo in my mind which you post here. It looks a little bit there or the mask was not printed to the asic chip area to stop flooding the tin-solder over the whole strip line. In this case there is only a little bit missing of tin-solder.

I want to prevent me from this problem so maybe compensate the situation by a little bit soldering paste filled to the holes ?

Another question, as I understood in this discussions was not only the cooling of the chips is essential, also the cooling of the power chips is very essential. So it looks to me it must not be a big EVE 212 cooler needed only. Maybe a smaller is good, too but I should check that the power chips get there own heatsinks.
I plan to glue a heatsing on top and a big heatsink plate on the back for the power chip area.

I've now ordered two of the most more cheap Scythe Setsugen Rev. 2 Grafikkartenkühler VGA
https://www.google.de/images?output=search&sclient=psy-ab&q=Scythe+Setsugen+Rev.+2+Grafikkartenk%C3%BChler+VGA&gbv=1&sei=fjSHUuX_EMLItQam-oHACg&hl=de&sa=X&oi=image_result_group&ei=gDSHUr5_jcGzBsGUgOgK&ved=0CDIQsAQ
which could be buy in germany from < 22 € and plan to mount them in a way the air flow will be go through the top heatsinks of the power chips.
The heatsink itself is know to handle ~ 200 W which should be enough to cool the asics.
On the back I want to mount a heatsink as bfl did this with the little singles.

I hope it will work ;-)

But in fact, can anybod give me the mechanical dimension from the board and the holes so I can prepare/check my 19 Inch server rack case for mounting some of them? I would like to bring the case in a air conditioned server room which is only possible if I mount them in a server case as I allready did this with my knc miners ;-)
I want to calculate how much asics I can mount in one case. I think 5 should been possible but I need
clarify that.

Cheers...

Well, the pads for the ASICs are that they are what are called solder mask defined pads. The pads are holes in the solder mask, and when they go through the lead-free solder bath that creates the tinned pads. When they did this process the first time, the board house had a couple issues with the flux not getting into the holes in the mask, so the pads didn't tin properly. The staff there fixed it by manually tinning them, but that left the issues with unlevelness.
We sent that back and they fixed the problem, so the boards you get shouldn't have any issues. Even on the ones we had with some mask issues we had no problems with bridging, but the ones you will get had no missing mask.

That cooler looks like it will work, though I can't tell for sure without knowing the dimension of the contact plate. For safety it should be about 35mm square.

[i3luefire]

• 13 miners
• 400gh
• 3kw
• $13k

[ryanb]

just received my chilli i connected it to my raspberrypi and i am gettting failed to send queue message and retry in 1 second

it is detected and temp is 30 degrees

any ideas?

Thanks

[MrTeal]

just received my chilli i connected it to my raspberrypi and i am gettting failed to send queue message and retry in 1 second

it is detected and temp is 30 degrees

any ideas?

Thanks

I'll let ChipGeek answer this one. I have a Pi in a drawer somewhere, but I've never taken it out. My understanding is that they should just fire up (we used them in our test setup), but I'm not sure on that one.