wild and unsubstantiated speculation about BFL's power woes

[eldentyrell]

Disclaimer: I don't know anything that hasn't been posted here in the forum, and I have been too busy to read more than 5-10% of that, I have no inside info, this is just speculation based on past chip design experience, yadda yadda.

It sounds like BFL is having some difficulties with their chips (a) running at ~25% of rated speed and (b) consuming 4-9x more power than they planned.  These numbers are vague since the info in the postings I've come across has always been patchy and has been presented very poorly (they'll quote an actual hashrate but decline to say what device it came from, or show a photo of a kill-a-watt but decline to say exactly what's plugged into it, etc).  I would welcome a simple and straightforward posting by BFL saying "we have the chips running at X% of advertised speed and consuming Y J/GH".

They've also mentioned that the wafer-probing tests (which I assume ran only a tiny portion of one chip at a time due to the fact that unpackaged chips overheat when run at full speed) produced the power results they expected, but the packaged chips consume way too much power.

I'm going to make a wild guess here and speculate that they ran all their pre-production SPICE simulations using the default 25 degree C temperature.

This is a pretty common mistake.  It would also explain everything I've seen so far.

Circuits always simulate ridiculously well if you run SPICE at 25 degrees.  The problem is that any circuit doing substantial amounts of computation will generate enough heat to raise the local temperature to at least 100 degrees C.  This in turn reduces the power efficiency even further (circuits running hot run slower and burn more power), an effect that feeds on itself.  Even the best packaging and heatsinking still leaves 3-4 degrees C per watt, and often those figures neglect to include the thermal resistivity of the CMOS bulk (another few degrees C per watt).  Multiply all of that by a 10-20W chip and the junction temperature is going to be a lot closer to 100-110 degrees than 25 degrees in any sort of reasonable ambient temperature.

Repeated disclaimer: this is just a wild guess, I have almost zero information, I have been way too busy with other urgent crap to read most of the forum threads, etc, etc.

[Unacceptable]

The chips are ok....it's the board that is the problem....so they say.

They are doing a revision on the PCB & adding "power regulators" or something  

Congrats on the "devil" post #666 

[Frankie Delaney]

It's also entirely possible that they never simulated anything, that they're flying by the seat of their pants, and their first prototype is educating them on why you don't go whole hog and order the parts to mass produce a product before any testing is done.

Let's not forget this is their FIRST PHYSICAL PROTOTYPE. Yes, the results they are getting would be a problem if they were nearing mass production, but they are still in the very early stages with their FIRST PHYSICAL PROTOTYPE.

[mezzomix]

I'm not sure that this is their first prototype. If it is, they had more luck than they deserve.

I assume they made one or more runs that where not working at all. If they spend enough money they might be able to create a cooling that is able to cool down the chips. I assume they will not invest in a cooling solution with a price of several thousand dollars in order to cool down a single. At the moment they are in a trap. They promised too much and have no product they can afford to deliver. Trying another ASIC run and shift the shipment by another several months will hit them very hard with Avalon creating one batch after the other.

[BR0KK]

The chips are ok....it's the board that is the problem....so they say.

They are doing a revision on the PCB & adding "power regulators" or something  

Congrats on the "devil" post #666 

could someone explain to a noob how they could potentialy archive a cut in power this large?
Dissabling chips (hashing engines), Underclocking, Undervolting and more Regulators won't do the trick i think

[Littleshop]

Wild speculation....

The could still use existing boards and chips to make Jallys. 

[BR0KK]

Wild speculation....

The could still use existing boards and chips to make Jallys. 

Powered by unicorn blood generators :/ ....

Is there a way to speculate what one chip will use in power?

[Frankie Delaney]

Wild speculation....

The could still use existing boards and chips to make Jallys. 

The jalapeno is powered by 2 x 2.5 watt USB plugs. It was originally intended to be 4.5GH/s and 4.5 watt, but if the chips are more than 12% out of spec, the jalapeno will not run at all at 4.5GH/s. You're right though, they're saying the problem isn't the chips, it's the regulator, blah blah blah, but the jalapeno boards were considerably less populated than the other boards. They lacked a lot of the power handling stuff, the capacitors and such, that the larger boards have. You would think if it was a board problem, not a chip problem, the jalapeno wouldn't be affected, and they would be flying out the door.

https://i.imgur.com/nqaKxHw.jpg

[2112]

Could you please speculate wildly on why the idle power usage is so high?

Originally advertised values for a Single SC: 40 Watt while hashing (at 40GHash/s). Actual values for this little prototype board (with unknown hashrate): 42 Watt idle (!!), 160 Watt when hashing.

The ratio of loaded/idle is unlike anything CMOS that I have ever seen. It reminds me of ECL or linear analog designs.

How can one get to such predicament? I've heard of latchup in CMOS circuits, but it becomes a short circuit and therefore completely prevents the normal operation until powered down.

[MrTeal]

The chips are ok....it's the board that is the problem....so they say.

They are doing a revision on the PCB & adding "power regulators" or something  

Congrats on the "devil" post #666  

Looking at the video and picture posted, I have a hard time believing the problem just lies outside the chips if the power meter in Luke's pictures is accurate at 180W. Even if it's only 150W DC, there just isn't anything else on the board that could handle dissipating that kind of heat with minimal airflow other than the big heat sink. Even with those little heat sinks on the power fets and under the board, there's no way even half of that 150W is being dissipated in the board. The chips themselves are almost definitely running at several times higher than their estimated efficiency, and it's unlikely they're going to be able to solve that just with board changes.

[Littleshop]

Wild speculation....

The could still use existing boards and chips to make Jallys. 

The jalapeno is powered by 2 x 2.5 watt USB plugs. It was originally intended to be 4.5GH/s and 4.5 watt, but if the chips are more than 12% out of spec, the jalapeno will not run at all at 4.5GH/s. You're right though, they're saying the problem isn't the chips, it's the regulator, blah blah blah, but the jalapeno boards were considerably less populated than the other boards. They lacked a lot of the power handling stuff, the capacitors and such, that the larger boards have. You would think if it was a board problem, not a chip problem, the jalapeno wouldn't be affected, and they would be flying out the door.

As originally designed yes.  But with an external power brick and the original regulators (though more of them) it should work fine.  It would need a heatsink as well, but again, they could still reuse those boards and components related to the original mini single SC.

[Frankie Delaney]

Wild speculation....

The could still use existing boards and chips to make Jallys. 

The jalapeno is powered by 2 x 2.5 watt USB plugs. It was originally intended to be 4.5GH/s and 4.5 watt, but if the chips are more than 12% out of spec, the jalapeno will not run at all at 4.5GH/s. You're right though, they're saying the problem isn't the chips, it's the regulator, blah blah blah, but the jalapeno boards were considerably less populated than the other boards. They lacked a lot of the power handling stuff, the capacitors and such, that the larger boards have. You would think if it was a board problem, not a chip problem, the jalapeno wouldn't be affected, and they would be flying out the door.

As originally designed yes.  But with an external power brick and the original regulators (though more of them) it should work fine.  It would need a heatsink as well, but again, they could still reuse those boards and components related to the original mini single SC.

AT that point you have a 40watt idle board powering a 4.5GH/s device.

[Frizz23]

I'm going to make a wild guess here and speculate that they ran all their pre-production SPICE simulations ...

The fact that only recently "discovered" the real power consumption of their ASICS shows that they never had a working prototype up until now.

What were those fuckers supposed to ship in October 2012 (their original shipping date) ? They had nothing back then. Literally nothing! Stringing customers along for months ...

[Frizz23]

The could still use existing boards and chips to make Jallys.  

They already said they probably have to add an external power supply plus heatsink and fan to the Jalapeno.

That means they can throw away lots of existing Jalapeno casings and PCBs. And burnt lots of our pre order money for this "Lessons Learned".

[BR0KK]

Could it be that they had top change the specs of the chip (from 65 nm to 90nm or higher) while they switched to FCBGA?

[muyuu]

The chips are ok....it's the board that is the problem....so they say.

They are doing a revision on the PCB & adding "power regulators" or something  

Congrats on the "devil" post #666 

If that was the case, v2 would be so incredibly superior to v1, the first buyers would be completely fucked over.

On the bright side, it doesn't seem true at all. They have this habit of over-promising and under-delivering.

[Gator-hex]

No idea how many chips are running in the video...

http://www.youtube.com/watch?v=4C4bgho5JSI

...or if the thing hashing 25GH/s in the photos is a FPGA Mini Rig or an ASIC...

...but assuming it's a 4 chip 30GH/s Little Single @ 160w/25GH/s...

1) 4.5GH/s Jalapeño is toast, needs complete board redesign. USB bricks are out. 160w / 4 = 40w/3.75GH/s

2) 60GH/s SC Single is toast, needs an redesign with enormous cooler, and more voltage regulators for 160w x 2 = 320w/50GH/s

3) 1500GH/s Mini Rig SC likewise is going to have cooling, power issues, and require a bigger case for the bigger coolers.
    (If they built it, it would blow your home mains fuses anyway!) 320w x 25 SC Singles = 8000w/1250GH/s

    1500Gh/s could be reached with 60x Little Singles but this would be 9600w!

[Minor Miner]

I would really like someone to put up a bet on a date when they will have 5 mini rigs operating in paying customers hands in the USA (really want to see if any BFLers take the other side).   Mini Rig casings are all garbage at this point but if they do not figure this out and ship, they will basically be shipping fire starters and bitcoin will draw the wrath of the ignorant media when some idiot burns an apartment (or dorm) down "mining for millions".   I have ZERO faith they will send anything to UL which could have MO shut them down after they start shipping since most munis do not allow untested electrical devices in their municipalities for this very reason.

[Gator-hex]

I would really like someone to put up a bet on a date when they will have 5 mini rigs operating in paying customers hands in the USA (really want to see if any BFLers take the other side).   Mini Rig casings are all garbage at this point but if they do not figure this out and ship, they will basically be shipping fire starters and bitcoin will draw the wrath of the ignorant media when some idiot burns an apartment (or dorm) down "mining for millions".   I have ZERO faith they will send anything to UL which could have MO shut them down after they start shipping since most munis do not allow untested electrical devices in their municipalities for this very reason.

It would be a good bet to create, because nobody is going to be able to run a 1500GH/s (now 1250GH/s) Mini Rig SC in their home @ 8000w

120v x 20A USA You're kitchen might have a 2400w rated socket for your cooker, everywhere else will be 1800w.
230v x 13A UK might get up to 3000w max on the kitchen cooker socket.

[Littleshop]

I would really like someone to put up a bet on a date when they will have 5 mini rigs operating in paying customers hands in the USA (really want to see if any BFLers take the other side).   Mini Rig casings are all garbage at this point but if they do not figure this out and ship, they will basically be shipping fire starters and bitcoin will draw the wrath of the ignorant media when some idiot burns an apartment (or dorm) down "mining for millions".   I have ZERO faith they will send anything to UL which could have MO shut them down after they start shipping since most munis do not allow untested electrical devices in their municipalities for this very reason.

It would be a good bet to create, because nobody is going to be able to run a 1500GH/s (now 1250GH/s) Mini Rig SC in their home @ 8000w

120v x 20A USA You're kitchen might have a 2400w rated socket for your cooker, everywhere else will be 1800w.
230v x 13A UK might get up to 3000w max on the kitchen cooker socket.

Many US homes have a 230v x 30A dryer plug.  I have similar plugs for servers though not the same physical configuration.   7200W derated to 80% gives you 5760.   That is about as high as you want to go and of course getting rid of that kind of heat is challenging.