[Work in progess] Burnins Avalon Chip to mining board service

[Foofighter]

awesome update! 

Can someone confirm this one will do for 15 modules
http://www.coolermaster.com/product/Detail/powersupply/silent-pro-m2/silent-pro-m2-720w.html
Input Voltage    90-264Vac (Auto Range)
Input Current    11 – 5.5A
Input Frequency Range    47 - 63Hz
PFC    Active PFC (>0.9)
Power Good Signal    100-500ms
Hold Up Time    >17ms
Efficiency    85% Typically
MTBF    100,000 Hours
Protection    OVP/UVP/OCP/OPP/OTP/SCP
Output Capacity    720W
Operation Temperature    0~40°C (Nominal Input Voltage)
Regulatory    TUV / CE / UL / FCC / BSMI / GOST / C-tick / KC / CCC
Fan    135mm Hydraulic Dynamic Bearings
Certifications    80 Plus Bronze
Connector    M/B 24 Pin Connector x 1
CPU 4+4 Pin x 1
PCI-E 6+2 Pin x 4
SATA x 9
4 Pin Peripheral x 5
4 Pin Floppy x 1

Thanks.

Theoretically it is sufficient, but its at the limit particularly regarding to the efficiency. I would suggest the next size that is >= 850 W. The 720 W also has no buffer for possible overclocking. But even without overclocking I would suggest a larger one.

should work fine, as it is labeld with 720 w it will also provide 720W but will use propably more then 800w

But for overclocking I would not suggest this product.

[geopolisch]

Burnin is doing his job sooo much better than BFL ever did !!!! THX 4 ALL THOSE FANTASTIC NEWS & PICS

I cannot agree more!!

[m5]

What about something like this http://www.pugetsystems.com/mineral-oil-pc.php for those with many boards?

[mine2mind112]

What about something like this http://www.pugetsystems.com/mineral-oil-pc.php for those with many boards?

I think is funny but the point is that this board should make money not eat them 

[whisper]

I thing about the same...

[krasnolud]

What about something like this http://www.pugetsystems.com/mineral-oil-pc.php for those with many boards?

I think is funny but the point is that this board should make money not eat them 

Better cooling enables more overclocking = more income.

This oil cooling looks interesting.
But for lager rigs a swimming pool of oil would be needed...

[tigerbit]

What about something like this http://www.pugetsystems.com/mineral-oil-pc.php for those with many boards?

I think is funny but the point is that this board should make money not eat them 

Better cooling enables more overclocking = more income.

This oil cooling looks interesting.
But for lager rigs a swimming pool of oil would be needed...

I hear mineral oil brought up in discussion so many times, but these sort of solutions are surely just more trouble than they're worth.  The heat captured still has to go somewhere - if air cooling isn't enough, then waterblocks are a far more sensible and proven solution if you insist the stock cooling is inadequate for your needs.

Yes, but what about overclocking I hear you say.  450MHz dare I say it, is a fantasy figure.  Theoretically possible according to Avalon but based on pretty much all real world overclocking experiences the absolute maximum is typically impractical for either stability, safety, energy consumption, chip longevity etc..  If the majority of Avalon chips were capable of this, for what conceivable reason would they have limited their clock to 285 MHz in their own units.

What's really most important is getting boards in hands.  I wouldn't want to see any attempt to build in exotic overclocking abilities stand in the way of that.  Sure, a modest overclock to mid-300s would be awesome but I reckon after that you're into severely diminishing returns.  There's also the warranty issue - Burnin offers a 2 year warranty.  What about the fools that try 450Mhz on air cooling?

[Schrankwand]

What's really most important is getting boards in hands.  I wouldn't want to see any attempt to build in exotic overclocking abilities stand in the way of that.  Sure, a modest overclock to mid-300s would be awesome but I reckon after that you're into severely diminishing returns.  There's also the warranty issue - Burnin offers a 2 year warranty.  What about the fools that try 450Mhz on air cooling?

Well, overclcoking needs to be thought of first. Any later revision is pointless with the boards.

So, he can just add the power supply and the regulator and then let people do whatever they want. He can then insist that overclocking above 282mhz breaks warranty and is fine off.

[FullFathom5]

I hear mineral oil brought up in discussion so many times, but these sort of solutions are surely just more trouble than they're worth.

Edit: I'll see your tad of unqualified FUD and raise you an iota of qualified optimism:

Read the links that discuss this. Cooling electric/electronic systems by direct oil contact has been in use for decades. Computer users have historically not done so, likely because there hasn't been a need. The likes of Intel have done studies spanning a year of nonstop server operation. Follow up failure analysis indicated hardware was in fine shape. The benefit was significant operational cost reductions, primarily in the form of reduced air conditioning costs. There are others who have had similar results, these studies are interesting reads.

The heat captured still has to go somewhere - if air cooling isn't enough, then waterblocks are a far more sensible and proven solution if you insist the stock cooling is inadequate for your needs.

Yes, the heat capacity of water is the best. Cheap, widely available. It is one alternative. Indeed, thermodynamics tells us that the heat must go somewhere. In this particular case, away from the ASIC chip is all that matters. There are many ways to do this. Water is one, oil another. The combination of the two... or just air. Creativity is the only limiting factor here.

Yes, but what about overclocking I hear you say.  450MHz dare I say it, is a fantasy figure.  Theoretically possible according to Avalon but based on pretty much all real world overclocking experiences the absolute maximum is typically impractical for either stability, safety, energy consumption, chip longevity etc..  

From what I have seen on the forum, the OC attempts were performed under air cooled conditions. I may have missed a thread, if I am mistaken I would be interested in reading it. If all these OC studies were under air cooling, these are not the conditions that Yifu (BitsynCom) identified that 450MHz was achieved at Avalon. Your point is taken, I would like to see more info from Avalon regarding that study (duration, conditions, errors thrown, etc.). The only info offered was that "450" was not achieved under air cooled conditions.

If the majority of Avalon chips were capable of this, for what conceivable reason would they have limited their clock to 285 MHz in their own units.
[...] What about the fools that try 450Mhz on air cooling?

I think you answered your own question there.

I also agree with you: we need to get these boards in hand as soon as the PCB designers have achieved their design aim.

[tiros]

awesome update! 

Can someone confirm this one will do for 15 modules
http://www.coolermaster.com/product/Detail/powersupply/silent-pro-m2/silent-pro-m2-720w.html
Input Voltage    90-264Vac (Auto Range)
Input Current    11 – 5.5A
Input Frequency Range    47 - 63Hz
PFC    Active PFC (>0.9)
Power Good Signal    100-500ms
Hold Up Time    >17ms
Efficiency    85% Typically
MTBF    100,000 Hours
Protection    OVP/UVP/OCP/OPP/OTP/SCP
Output Capacity    720W
Operation Temperature    0~40°C (Nominal Input Voltage)
Regulatory    TUV / CE / UL / FCC / BSMI / GOST / C-tick / KC / CCC
Fan    135mm Hydraulic Dynamic Bearings
Certifications    80 Plus Bronze
Connector    M/B 24 Pin Connector x 1
CPU 4+4 Pin x 1
PCI-E 6+2 Pin x 4
SATA x 9
4 Pin Peripheral x 5
4 Pin Floppy x 1

Thanks.

Theoretically it is sufficient, but its at the limit particularly regarding to the efficiency. I would suggest the next size that is >= 850 W. The 720 W also has no buffer for possible overclocking. But even without overclocking I would suggest a larger one.

should work fine, as it is labeld with 720 w it will also provide 720W but will use propably more then 800w

But for overclocking I would not suggest this product.

At first I don't plan to overclock so it should be fine
Thanks for replies.

[SebastianJu]

What are these 4 slots? The middle at the end is for power? What about the other 2 at the end and the 2 at the left side?

Offtopic: Its funny how much such pcb's remind me of cities. Streets, flat buildings, oil containers, foundries with chimneys...

[Toolhead]

What are these 4 slots? The middle at the end is for power? What about the other 2 at the end and the 2 at the left side?

Offtopic: Its funny how much such pcb's remind me of cities. Streets, flat buildings, oil containers, foundries with chimneys...

I hope that no magic smoke comes up from the "chimneys" 

[FullFathom5]

No that's not the Production heatsink:

Am working on a case/cooling solution, are dimensions of the production heatsink known?

[tigerbit]

What are these 4 slots? The middle at the end is for power? What about the other 2 at the end and the 2 at the left side?

Offtopic: Its funny how much such pcb's remind me of cities. Streets, flat buildings, oil containers, foundries with chimneys...

Nearest us - USB, Power, Power & Power.  3 connection options depending on how you intend to power it.
Left side - CAN Bus header connector, and what looks like a cooling fan power connector.

[finnzwerg]

do the chips really transfer more heat into the PCB than to the upper side?

Yes, they are designed this way. And you shouldnt put a heatsink on top. Yifu said this will have a negative effect instead.

Can anybody verify this?
On computer mainboards. (NB/SB , MOSFETS) or RAM you can cool with out problem from the top.

The differential temperature in one chip should be less than 1°C ... i can not image that this would have a negativ effect.

The only can explain the negativ effect throuhg wrong/to high mounting force. (so that the chip will break)

Regards
Finn 

[Sophokles]

do the chips really transfer more heat into the PCB than to the upper side?

Yes, they are designed this way. And you shouldnt put a heatsink on top. Yifu said this will have a negative effect instead.

Can anybody verify this?
On computer mainboards. (NB/SB , MOSFETS) or RAM you can cool with out problem from the top.

The differential temperature in one chip should be less than 1°C ... i can not image that this would have a negativ effect.

The only can explain the negativ effect throuhg wrong/to high mounting force. (so that the chip will break)

Regards
Finn 

Putting a heat sink on top will restrict the airflow over the pcb. The pcb has internal copper layers ('ground' is usually designed as a plane) which act as heat spreaders as well.

[FullFathom5]

I think Sophokles answered your question.

Links: http://tinyurl.com/m4zy7p3  <-- this was posted elsewhere, maybe in this thread IIRC.

do the chips really transfer more heat into the PCB than to the upper side?

Yes, they are designed this way. And you shouldnt put a heatsink on top. Yifu said this will have a negative effect instead.

Can anybody verify this?
On computer mainboards. (NB/SB , MOSFETS) or RAM you can cool with out problem from the top.

The differential temperature in one chip should be less than 1°C ... i can not image that this would have a negativ effect.

The only can explain the negativ effect throuhg wrong/to high mounting force. (so that the chip will break)

[bezzeb]

do the chips really transfer more heat into the PCB than to the upper side?

Yes, they are designed this way. And you shouldnt put a heatsink on top. Yifu said this will have a negative effect instead.

Can anybody verify this?
On computer mainboards. (NB/SB , MOSFETS) or RAM you can cool with out problem from the top.

The differential temperature in one chip should be less than 1°C ... i can not image that this would have a negativ effect.

The only can explain the negativ effect throuhg wrong/to high mounting force. (so that the chip will break)

Regards
Finn 

Hey guys, to all who are concerned about cooling from the other side of the PCB, please chill (pun intended) and study "thermal vias" - a common method to cool modest wattage surface mount components from the back of the board.  Read up here and feel better:
http://www.electronicproducts.com/Thermal_Management/Heat_Sinks_and_Thermal_Materials/PCB-cooling_techniques_and_strategies_for_IC_packages.aspx

Bottom line:  Don't put heat sinks on top of your avalons - it won't do anything good or bad, but will risk breaking the chips off of the board if the heat sink gets twisted or bumped or over tightened.  SMC's are not the strongest type of chip package, and these babies are not really designed for top cooling.  Chips you should cool with top heat sinks were designed for top heat sinks - and if you study the package form factors closely you will see clever reasons why they should be top or bottom cooled.  (Like for top cooled chips, too many lead contacts on the bottom, or the chip fits in a socket which blocks downward thermal flow, or the wattage is just too high for through PCB cooling, blah blah etc..)

These asics want back side cooling, and we should rejoice because it's a reliable and easy (cheap) way to go.  They have a metal pad on the bottom designed for heat flow into the PCB, with a kind of thick layer of plastic on top which does not really conduct heat that well.

And to all the liquid coolers and mineral oil guys - the wattage of these ASIC's is not super high so they don't make big heat like a hard core top end zillion core super clocked beast chip.  So spend the money on fancy cooling systems if you want, but the only real benefit might be noise reduction.  I don't think the ROI would be in your favor on technical merits, Burnin has demonstrated that he knows what he's doing.  Now I love liquid cooling as much as the next guy and have a pretty sweet water chilled silent rig, but pragmatism is a virtue.  ;-)  Remember these ASICS are killer tech because they do more work for less power which means dramatically less total heat per square mm per hash (generally speaking) than an equivalent old school rig (cpu, gpu, fpga, whatever).

So keep cool when keeping it cool, peace out. 

[burnin]

So keep cool when keeping it cool, peace out.

True, but watercooling could increase the packing density of the Boards.
And with maximum overclocking the stock heat sink will not suffice (more then 70 Watts/board).

No that's not the Production heatsink:

Am working on a case/cooling solution, are dimensions of the production heatsink known?

Production heatsink is 83x100x40mm.
Board dimensions:
https://www.dropbox.com/s/gkuayv84tr63x70/BitBurnerXX_rev3.pdf
I could provide a iges/step files of the assembled board if that helps.

Todays status:
Found a nasty bug that produced the flood of nonce errors.
And I fried one part of that chip that i planned to replace tomorrow. - just another reason to change it.
But it meant that i could only run it with one chain today.
It was pretty stable though, about 2% HW errors.
But with the replacement chip that should hopefully go down to 0%.
And then I can pull the trigger on the website and the PCB Orders.

[dan99]

That is cool and Thanks.