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

[Exoskeleton]

Why do you think so? Its very easy for burnin to see the ordernumbers in his shop, but it would be much work to find out in which batch a customer bought, when this batch was ordered and so on. I, in burnin's shoes wouldnt do the work, and i dont see why that should be done. Burnin doesnt have to favor early chip buyers.

How exactly is burnin supposed to build your miner without your chips in hand? If he receives your chips at the same time as he gets mine, and your first in line (first to pay), you go first obviously. But if he gets my chips first and he does not have your chips yet, I go first. You have to wait untill he has your chips in hand to start your order. Its that simple.

As I just stated in a previous post, being first in line only helps if he gets multiple batches on the same day and your batch is one of those. If you ordered in may, and your batch does not show up untill 1 month after the april orders, you won't get your miner made untill your chips show up a month later. Burnin cannot make your boards untill he has your chips in hand.

This is why I was stressing that the real hold up could be the delay between batches. If Avalon ships all the chips on the same day, then order placement with burnin will be the only factor who goes first. Otherwise the order in which he receives batches of chips is the main determining factor in what order miners get made. Within the constraints of the chips he has available, order placement with him will put you in the front of the line. But if he does not have your batch of chips in hand your not going to be in line yet.

This seems fair, simple, and obvious to me, but I don't want to speak for burnin. Perhaps he could clarify this to those still arguing against this model.

Oh wait... he already did...

Clarification coming up:
The batch is relevant, because that's the order in which I will receive the chips.
Even if you were to be the very first person to order on my website, i could only assemble and send you your boards once I received the batch your chip are in.
And that's why the batch is relevant.
Within one batch the succession of the order number is determining the position in the queue.

And if your chips are in seprate batches...

I wait until all chips needed for an order are present.
If you want per batch shipping do separate orders for each batch.

Case closed. 

[driksson]

Case closed.  


[/quote]

I concurr. He wont ship someone elses chips to you, just because you ordered your boards earlier. And as you said, if multiple batches arrive same day, they are delivered by your place in order queue..

[jlsminingcorp]

Shop Timing:
I agree that this is not the optimal strategy for things like this.
But with all in all 500 People ordering stuff things should not break instantly.
(did choose a beefy server)

Hi all, thanks for putting my mind to rest. I think I was going a bit crazy last night with visions of a server apocalypse! Sounds like this is unlikely.

Thanks again for all of your hard work burnin, this is a great project.

But i am a professional

Yes indeed, this has been demonstrated here many times !
 

[nemercry]

@burnin: Hey is it possible to say how high the boards gonna be (including the heatsink) ?
I plan a custom cooling on my Asics and need the data how high two stacks would be or what the size of one will be ?
I planned with a height of approximately 60mm based on the assumption that the fan is 80mm of cooling size + some mm for saftey.( comes from the assumption, that the long side of the board has 100mm and the fan needs to be a little smaller)
I would be very happy if you could provide any further information of the approximately height !

Do you think there is still more OC for the WC cooled devices ?(Maybe reaching 500 with 1.5V ?)
What is the Maximum volts, which can be set?

Thanks for you amazing work.

[Fishbones]

You are correct about the pnp machine (i am up to 405 parts now btw).
...snip

Up to 405 parts?... What parts were added? Do you, or will you have all parts and components ready to assemble when chips arrive?

Thanks for the hard work. Best of luck with your business venture.

[SebastianJu]

And i have some numbers to go with those from yesterday:
Slightly different air cooling setup therefore different temperatures with air cooling. (fan placement)
TL;DL : 450Mhz [9Ghash/s] - STABLE
But at the cost of 94Watts of power.

Air:
431 - 54, 48, 1.30V, 87W, stable
450 - 56, 48, 1.30V, 90W, HW Errors
450 - 57, 52, 1.34V, 94W, slightly increased error rate compared to what i normally call "stable" but close enough

Water:
450 - 54, 32, 1.34V, 94W, slightly less hw errors then with air

Genius... you beat the original developer Avalon...

Can you say something about how easy overclocking will be? Is it a switch or something? I believe Klondikes need soldering for doing that, i hope overclocking and overvolting is easier with bitburners.

And can you tell the name of the fan, or if its noname the airvolume, size and so on?

Then... what about breaking asic-chips? Will 10 chips stop working when one breaks? And does this one has to be replaced in order to let all chips run again?

Thanks!

@Exoskeleton ... I wrote the same as you. When chips arrive earlier than others that paid earlier the chips that are available will be worked up. But the order number will be the first priority because otherwise burnin would have to do work without sense by finding out whose chips batch was ordered earlier at avalon and so.
But i think we see it the same way anyway.

[iikun]

Then... what about breaking asic-chips? Will 10 chips stop working when one breaks? And does this one has to be replaced in order to let all chips run again?

Thanks!

I could be wrong but I believe he said before that if one chip stops working it will only affect one side of the board (10 chips). So if a chip stops working on an XX board you will still have one side working at least.

[driksson]

Then... what about breaking asic-chips? Will 10 chips stop working when one breaks? And does this one has to be replaced in order to let all chips run again?

Thanks!

I could be wrong but I believe he said before that if one chip stops working it will only affect one side of the board (10 chips). So if a chip stops working on an XX board you will still have one side working at least.

Yes, which i believe would not be the case on the Klondike boards. Would it be possible to only have that very one chip stop working only instead (and even indicated, what ships are delivering and not so it can be replaced)? i think that is the question. Which i also would like..

[iikun]

Then... what about breaking asic-chips? Will 10 chips stop working when one breaks? And does this one has to be replaced in order to let all chips run again?

Thanks!

I could be wrong but I believe he said before that if one chip stops working it will only affect one side of the board (10 chips). So if a chip stops working on an XX board you will still have one side working at least.

Yes, which i believe would not be the case on the Klondike boards. Would it be possible to only have that very one chip stop working only instead (and even indicated, what ships are delivering and not so it can be replaced)? i think that is the question. Which i also would like..

I can't speak for Burnin of course, but I'm sure if it was easy for his design to do this he would've done it that way already.

Also, I found my previous answer is only partly correct sorry. According to Burnin's quote below it depends on the failure mode:

I have a similar question:

Lets say one Chip is damaged - Miner still working ?

The chips, as mentioned before, are calculating in a "row." Which means, that if one chip in a ten chip cascade is dead, the whole row of ten goes down and produces hardware errors.

Depends on the failure mode of the chip.
bang! magic smoke escapes - whole module down, rest of the stack continues to function
chip produces wrong hashes but continues to forward messages - some hw errors in cgminer, hashrate drops by one chip
chip logic completely dead - one row dead from the point of defective chip, hashrate drops by 1 to 10 Chips

My design is spare-no-expense, build for resiliency and reliability.

[SebastianJu]

This sounds good. So when a chip fully stops working all others remain working.

[ujka]

This sounds good. So when a chip fully stops working all others remain working.

If one chip fully stops working and does not forward messages, all chips from that defective one to the end of that row will not function.

If that defective chip still forwards messages, only its hashes are dropped (hw errors).

[SebastianJu]

This sounds good. So when a chip fully stops working all others remain working.

If one chip fully stops working and does not forward messages, all chips from that defective one to the end of that row will not function.

If that defective chip still forwards messages, only its hashes are dropped (hw errors).

Oh... rest of the stack... I thought that means the remaining 9. Now the question is how often chips break. I guess overclocking will raise that probability.

[GandalfG]

This sounds good. So when a chip fully stops working all others remain working.

If one chip fully stops working and does not forward messages, all chips from that defective one to the end of that row will not function.

If that defective chip still forwards messages, only its hashes are dropped (hw errors).

Oh... rest of the stack... I thought that means the remaining 9. Now the question is how often chips break. I guess overclocking will raise that probability.

All chips are connected in chain in|out-> in|out-> in|out-> in|out ....   1st chip forward message to next. Each miner have different number chip in chain.
In burnin solution are IIR two chain, each contain 10 chip.
How many chip stop working are depend what exactly are broken inside chip.  Most worst scenario: 1st chip in both chain stop working definitely.

[SebastianJu]

This sounds good. So when a chip fully stops working all others remain working.

If one chip fully stops working and does not forward messages, all chips from that defective one to the end of that row will not function.

If that defective chip still forwards messages, only its hashes are dropped (hw errors).

Oh... rest of the stack... I thought that means the remaining 9. Now the question is how often chips break. I guess overclocking will raise that probability.

All chips are connected in chain in|out-> in|out-> in|out-> in|out ....   1st chip forward message to next. Each miner have different number chip in chain.
In burnin solution are IIR two chain, each contain 10 chip.
How many chip stop working are depend what exactly are broken inside chip.  Most worst scenario: 1st chip in both chain stop working definitely.

What can be done in such case? I mean Klondikes has here an advantage. If one chip breaks, regardless where, all other chips still work. But with bitburners one need a replacement chip and replace it, because otherwise more isnt working.
Can it be replaced easily? I might keep some chips at home then.

[burnin]

This sounds good. So when a chip fully stops working all others remain working.

If one chip fully stops working and does not forward messages, all chips from that defective one to the end of that row will not function.

If that defective chip still forwards messages, only its hashes are dropped (hw errors).

Oh... rest of the stack... I thought that means the remaining 9. Now the question is how often chips break. I guess overclocking will raise that probability.

All chips are connected in chain in|out-> in|out-> in|out-> in|out ....   1st chip forward message to next. Each miner have different number chip in chain.
In burnin solution are IIR two chain, each contain 10 chip.
How many chip stop working are depend what exactly are broken inside chip.  Most worst scenario: 1st chip in both chain stop working definitely.

What can be done in such case? I mean Klondikes has here an advantage. If one chip breaks, regardless where, all other chips still work. But with bitburners one need a replacement chip and replace it, because otherwise more isnt working.
Can it be replaced easily? I might keep some chips at home then.

afaik the klondike is even more susceptible to this because all 16 chips are organized in one chain.
But you seem to be overestimating chip failure rate a lot.
Replacement:
if you have a SMD reworking station AND the skills - easy.
If not: impossible

Can you say something about how easy overclocking will be? Is it a switch or something? I believe Klondikes need soldering for doing that, i hope overclocking and overvolting is easier with bitburners.

And can you tell the name of the fan, or if its noname the airvolume, size and so on?

Then... what about breaking asic-chips? Will 10 chips stop working when one breaks? And does this one has to be replaced in order to let all chips run again?

Setting the Frequency is easily done through the cgminer Avalon frequency config parameter.
Voltage will be done through a config parameter as well.
Fan size: 92mm 55CFM, 32dBa, 3000rpm.

@burnin: Hey is it possible to say how high the boards gonna be (including the heatsink) ?
I plan a custom cooling on my Asics and need the data how high two stacks would be or what the size of one will be ?
I planned with a height of approximately 60mm based on the assumption that the fan is 80mm of cooling size + some mm for saftey.( comes from the assumption, that the long side of the board has 100mm and the fan needs to be a little smaller)
I would be very happy if you could provide any further information of the approximately height !

Do you think there is still more OC for the WC cooled devices ?(Maybe reaching 500 with 1.5V ?)
What is the Maximum volts, which can be set?

One module is about 45mm high with heatsink, 92mm fan.
Much higher OC would be possible with the 10 Chip version because it can supply the whole 60 Amps to 10 chips.
At 450Mhz the 20 chip modules are already close to the power limit of the voltage regulators .
Probably already exceeding the 60 Amps design spec but with good cooling they should be able to push up to 80 Amps.
The upper voltage limit is 1.8 Volts  .

Edit: burnin please plan the PSU so that it accommodates with the overclocking figures and with stacking. thank you.

+1
Yes Please!
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Burnin,

You said in regards to air cooled over-clocking:
"Slightly different air cooling setup therefore different temperatures with air cooling. (fan placement)"
-Would this work with stacked boards or is the fan placement maybe unconventional just to get that umph?

You said:
"slightly increased error rate compared to what i normally call "stable" but close enough."
- What does that mean, maybe the numbers or the log/report from cgminer would clear this up?

-What is your personal opinion or gut feeling on longterm air cooled over-clocking at
 450 - 57, 52, 1.34V, 94W?

Normally stable would be lower then 0.01% errors, here we are talking 0.1%.
My gut feeling would be not to exceed 1.30V in the long term.
The PSU i am offering with the boards is the EVGA SuperNOVA G2 1000.
But i'll see if i can find an equally good 1.5kw model.

[mike2kt]

afaik the klondike is even more susceptible to this because all 16 chips are organized in one chain.

OK, just to clear this up, the K16 also uses two chains.

ALL: We are talking and 8 chip chain vs a 10 chip chain. This design is an Avalon ASIC result, not a BB^XX nor K16 design limitation.


PS: Burnin, any chance of a BoM soon?   

[mike2kt]

Probably already exceeding the 60 Amps design spec but with good cooling they should be able to push up to 80 Amps.

Burnin - is it ok to add heatsinks to the top of the buck regulators or would this be counterproductive with the thermal vias?

Thanks!!! Can't wait until I can order BB^XX boards!!

[tiros]

The PSU i am offering with the boards is the EVGA SuperNOVA G2 1000.[/b]
But i'll see if i can find an equally good 1.5kw model.

@burnin,
If any chance please include various extension/conversion cables on your website for those who already have PSU.
I switched to Cooler Master V1000 PSU and probably will be needing some of the variants listed below.
http://www.coolermaster.com/product/Detail/powersupply/v-series-psu/v1000.html

SATA to PCI-E 6P (need three of these)
http://www.ebay.co.uk/itm/6-pin-PCIe-PCI-E-Grafikkarte-SATA-Stromkabel-Adapter-/220732874310?pt=EN_Computing_Festplattenkabel_Adapter&hash=item3364b45e46

MOLEX 4-pin to PCI-E 6P (now I have these but someone else might need them)
http://www.ebay.co.uk/itm/6-Pin-Power-Adapter-Cable-for-PCI-Express-Video-Card-/190517701999?pt=UK_Computing_CablesConnectors_RL&hash=item2c5bbd6d6f

PCI-E 6P extension cable (probably need 8 of these since I am not sure what will be final distance between two bitburner xx)
http://www.ebay.co.uk/itm/6-Pin-Male-to-Female-PCI-Express-Power-Extension-Cable-20cm-/200722859585?pt=UK_Computing_CablesConnectors_RL&hash=item2ebc03c641
http://www.ebay.co.uk/itm/NZXT-CB-6V-250mm-Single-Sleeved-6-Pin-PCI-E-Ext-Cable-Black-/330786586551?pt=US_Power_Cables_Connectors&hash=item4d046abfb7

Please excuse my ignorance but I am still unsure about other type of connectors we will be able to use beside PCI-E 6P when using PSU 

[dani]

1.8V - nice!
Are the fans regulated by temperature/cgminer? Will they cost extra? (I guess they are included in the stacking kit).
Can you give some datasheets about the voltage regulator? I want to use one bitburner to push the chips to their maximum, with sufficient cooling. I'm talking about 450mhz+ and as much voltage as needed. This should be fun (don't worry, I won't complain if it breaks though  ). Also can you just resolder a bigger voltage regulator by hand and overclock even more this way? Can you deactivate One row of chips? This way you could go for pertty much insane clocks and voltages and if anything breaks you can go back to a safer value and use the other 10 chip row 

Yesterday the first chinese chip-preorder guys received their chips, so we should see some chips outside of china in the next two weeks hopefully. Prepare for massive impact  

[fiktionist]

i am excited about all those overclocking possibilities... but i also think that 450mhz might be pretty close to the end of the line.
on the other hand, i had some cpus on hand which made more than 4ghz, while having a stock frequency of 1.8ghz...

More voltage solves nearly any problem