[Announcement] Avalon ASIC Development Status [Batch #1]

[BitSyncom]

the purpose of this thread is to provide transparency and keep our existing and potential customers up-to-date whilst we work out some unforeseen website integration issues.

Hello everyone, my name is Yifu. Despise my fairly-lengthy involvement with Bitcoin, most of you do not know me. Rest assured however, I am no stranger to many in the Bitcoin community that is currently building/have built exciting business and/or systems. I must confess it has been my intention to stay away from the forums from the start as I enjoy being in the state of "doing" rather than in the acts of "blowing hot air". Alas, I do what I must.

I decided to look into ASICs awhile back because a simple and powerful reason: the lack of competition in Bitcoin mining dedicated hardware. Utilizing my well-traveled history, being multilingual and various skill sets is how I got involved with Avalon. Anyways, enough about me. Let's get to the good stuff.


before I begin you may wish to catch up on previous newsletters and a reddit AMA I conducted a few weeks ago.

http://store.avalon-asic.com | http://support.avalon-asic.com

Intro. Avalon launched with a challenge to our competitors: a over 50% dollar per GHash gain. Our competitors since then has also increased their specifications in order to stay competitive. Deviated from their original plan the competition managed to even the playing field, yet we simply proceeded with our original design goals and have reached 64Gh/s, and now 66Gh/s! Once again taking the lead in $/Ghash. While the competition promised early delivery dates but kept pushing them back. Avalon instead gave an very conservative estimate but now is scheduled to ship even earlier at Jan. 20th 2013.

Specification.

Code:

Chip Specification
Technology Summary:
    TSMC 0.11- micron G process
        5 Metal
Core Voltage: 1.2 V
I/O Voltage: 3.3 V
Core Frequency: 256+ MHz
Number of Pads: 48
    8 Data
    40+1 Power
Package Type: QFN48 -0.5 Pitch
Packaged Chip Size: 7 mm x 7 mm

Chip Interface
Data Pins (8 in total):
Clock                     i
Serial Data In  [2]       i
Serial Data Out [2]       o
Serial Data Bypass [2]    o
Reserved    [1]    -

Avalon since inception has been designed to be an stand-alone unit, embed Linux openWRT equipped with WIFI capabilities and an Ethernet Port, while the competition is designed to operated over USB, depended on the need of a host computer. Continuing our trend, we have narrowed down our power estimates from 600W to 400W and this number is guaranteed to decrease further. Did I mention our power supply is a standard ATX?

The true genius of Avalon lies within its modular design. need a better PSU? No problem! Install your own. In fact, it is within our schedule to release the controller ( openWRT ) image.iso before the end of the year to give eager developers a head start to spin your own custom mining setups. Embrace open source!

Trade-Ins. Avalon believes in continuing the trust that have build up with our old customers. Reinstating our policy of $300 per Icarus, $400 per Lancelot, limited one FPGA trade-in per ASIC unit. The trade-in program will continue to be available and will apply to 1st-gen Avalons when the time comes.

Stress Testing. Avalon will not be testing these ASICs on the Bitcoin network period, whether it being the main-net, test-net, or even test-net-in-a-box. Why is that you may ask?

We will be testing these ASICs to do what exactly they are designed for, that is computing double-SHA256 hashes. this algorithm is not bitcoin exclusive at all. It is just so happens to be the block hashing algorithm, thus it would be silly to even set up a mining pool or use the bitcoin network in and form or way. We will be testing using our own custom software written to stress test our chips.

Shipping We at Avalon understands the some what unreliableness of EMS in certain countries, to address this issue, all existing and future orders will have the choice to be shipped out via DHL at no additional charge. This will ensure timely delivery of our products.

Inventory. Avalon learned from the mistake of our competitors and since conception has decided to release small order batches of 300 ( a number solely based on our estimated capacity) and only when a batch is close to finished, we will begin the sales of the next batch. When a new batch is announced, there will be another thread. In short, any units ordered so far, until further notice is part of batch #1, which is currently scheduled to ship at Jan 14th. In addition, as people fail to follow up with their orders, there are still some units left in the store, so if you missed your chance originally, now it's the time.

Facility Photos. Due to popular demand by the community, We have uploaded our facility photos. The full res album is at http://imgur.com/a/KPBTl


Retail Pricing.
$1,299? Yep, to stay with the spirit of the competition. $1,299 for 66Gh/s.
question, comments and grievances welcome.

[BitSyncom]

reserved.

Updating as promised,

Let's be frank. We ran into a potential delay of 4-7 days.

What does this mean?

We when taped out the chips to TSMC we had a "projected out date" of Jan 3, 2013 which is on schedule with our late dec. / early jan. demonstration dates. Now however after this week's website update our "projected out date" has changed to Jan 7th, 2013 and "committed date" has been revealed to be Jan 10th, 2013. TSMC did not give us a reason for this, but as a small customer you are really at the mercy of bigger customers running higher priority lots. e.g. "super hot lot"  The "committed date" is the promised date of delivery, it is very unusual for TSMC to break their "committed date", so we expect our chips to ship on or before Jan 10th, 2013.

Another potential problem is Customs clearing, the packaging company Fujitsu is located in Shanghai's Export Processing Zone (EPZ), while the packaging itself will not see delays, it may take an additional 2-3 days to clear Customs. To make up for this, if it occurs, we will be flying down to Shanghai to the demonstration in the EPZ instead of waiting for the shipping to the factory for assembly.

With all this happening, it has burned through the 1 week additional leeway time we originally left out. I must say it will be difficulty for us to ship on Jan 14th, we however expect to ship around Jan 18 ~ 20.

Since some people were wondering how many chips are on a wafer earlier in the thread after we posted the MT forms I'll tell you, each wafer contains 4055 chips.

Code:

TSMC
TMEM91
================================================
Chip Size :   X = 3.9760 ,Y = 4.0560 mm
Reticle Size :   X/cell =  3 ,Y/cell =  3
Offset Value :   X = -3.7668 ,Y = -2.2990 mm
Alignment Mark :   (118.80,83.20),(-118.80,-83.20)
Alignment Mark Tolerant Distance :      1.6 mm
Notch Reserved Distance :   7.75 mm
Start Distance :   7.75 mm
Ring Edge :   3.0 mm
Photo Die Number:    4055

and this is the chip layout.

[eldentyrell]

Packaged Chip Size: 7 mm x 7 mm

Thanks for the transparency, but the unpackaged die size is what you ought to be posting.  Without that it's impossible to make performance comparisons (or at least those comparisons would have to assume it fills the whole package cavity, which portrays Avalon in the worst possible light).  Please provide the unpackaged die size.

Also, I think you may have neglected to say how many of these chips are in the 60GH/s device… or perhaps I missed that.  Anyways, without knowing how many chips are in the device, none of the information above is really useful.  Did you post this somewhere that I didn't see?

Core Frequency: 256+ MHz

Interesting; an FO4 on TSMC 110nm is around 63ps, so at your 3.9ns cycle time (256mhz cycle rate) gives a datapath that is 62 FO4's deep.  Of course that doesn't mean there are 62 levels of logic!  Wire delays and clock skew (for chips that have clocks) come out of this budget too.

[ngzhang]

reserved.

[PuertoLibre]

 

Finally...

Sounds great that you guys are leveling up the GH/s race. So far I think only Tom from bASIC might compete on this race. I guess the three letter company will ship with the specs they have set to date (60Gh/s).

How many workhorse chips do you have in this rig?

------------------------

Also it was a very nice touch for you guys to use open source code AND a standard power supply. Sounds like you guys thrive on open-ended innovation.

[Dexter770221]

Power consumption is quiet high If belives to BFL 6 times higher than their product.
There will be possibility to buy only chips? QFN package is simple to solder, and PCB design shouldn't be a problem. I have my own vision how rig should look like
I think that you may have many clients like me...

[Tinua]

What do you think about this?
Especially as regards the European market, with electricity costs between $ 0.10-0.30/kwh

Alright, one more scenario, more extreme:

Let's assume $12/BTC, $0.25/kwh power cost, and 1000 TH network hashrate after reward halving.

Hypothetical device that does 60 GH, costs $1300, and uses 60W:
$100K buys about 4600 GH of hardware which will use about 4.6kW of power and earn about $5980 per month, minus about $830 for power, for 5.15% monthly ROI.

Hypothetical device that does 54 GH, costs $1070, and uses 405W:
$100K buys about 5100 GH of hardware which will use about 38kW of power and earn about $6630 per month, minus about $6885 for power, for -0.26% monthly ROI. Uh oh.

Hypothetical device that does 54 GH, costs $1070, and uses 120W:
$100K buys about 5100 GH of hardware which will use about 11kW of power and earn about $6630 per month, minus about $2040 for power, for 4.59% monthly ROI.

But someone with $100K to invest, ought to find a better place to setup than where power costs $0.25/kwh.

And if ~5% monthly ROI were attractive to professional miners, why has mining been historically much more profitable than that, except when the exchange rate fell toward $2? I think 10-20% monthly ROI will continue for the next year or so at least. Though a price war among mining devices could really screw ROI up.

https://bitcointalk.org/index.php?topic=119744.msg1294453#msg1294453

[BitSyncom]

What do you think about this?
Especially as regards the European market, with electricity costs between $ 0.10-0.30/kwh

strictly speaking on power being the only variable I wouldn't even suggest to "mine-for-profit" basis with power rates higher than $0.10 ( and you are unable to get write offs or reductions. )

The real factor in the transition period of ASIC units will ultimately be delivery time, when difficulty is lower and the ROI is much higher.

Let's take this conservative example. using http://bitcoinx.com/profit/index.php

6 times the current difficulty
$.25 power cost.
400W
60GH/s

your break even time is 81 days, thats little over 2 month, of course assuming you can only maintain this ROI rate for 12 days or so, you would have effectively made 15% of your invest back.

with the current trend of everybody pushing their shipping date back I feel everybody has over estimated how quickly the difficulty will rise.

Let's assume Tom sold 1000 units, that's give or take 50TH/s,
Avalon only selling 300 units first batch, ~20TH.
BFL sells a crazy estimate of 2000 units, 120TH.

that barely adds up to 200THash, which is about 10 times the current network rate, and if difficulty goes up by a factor of ten using the same formula above. your ROI is 156 days.

honestly, it isn't that bad.

disclaimer: Avalon nor Yifu will be responsible for information provided here based an educated guess of future Bitcoin mining difficulty.

[benco]

that barely adds up to 200THash, which is about 10 times the current network rate, and if difficulty goes up by a factor of ten using the same formula above. your ROI is 156 days.

honestly, it isn't that bad.

this would be true only in the case that the difficulty will remain unchanged for 5 months, and you know that it will not

real ROI will be, at best, somewhere at 8 months.

[thorvald]

that barely adds up to 200THash, which is about 10 times the current network rate, and if difficulty goes up by a factor of ten using the same formula above. your ROI is 156 days.

honestly, it isn't that bad.

this would be true only in the case that the difficulty will remain unchanged for 5 months, and you know that it will not

real ROI will be, at best, somewhere at 8 months.

+1

[ice_chill]

While the price and performance is great, 400watts vs 60watts is far from competitive. Also a laptop will consume at most 30watts when mining hardware is attached, so it's strange that your all in one unit needs so much more.

I see you mention the power consumption will drop, but it needs to drop a lot.

[crazyates]

They've admitted to using a larger die size than their competitors. What else did you expect? The only consolation is being able to swap out to a more efficient PSU, and the fact that it doesn't need a host computer to run. Still, that GHs/1kUSD is nice for those of us with access to free electric.

[Bogart]

They've admitted to using a larger die size than their competitors. What else did you expect? The only consolation is being able to swap out to a more efficient PSU, and the fact that it doesn't need a host computer to run. Still, that GHs/1kUSD is nice for those of us with access to free electric.

A larger size than bASIC's 90nm, but smaller than ASICMINER's 130nm.  I don't believe any other ASIC makers have disclosed their process node size.

[cablepair]

Awesome Job Guys.

+1 and lots of respect to your team - Good luck with everything, I know your going to bring a really special and innovative product to the table.

If there is anything myself or anyone on my team can do for you please don't hesitate to reach out. We are are competitors but we still are friends

Good luck!

Tom
tom@btcfpga.com
(315) 514-0269

[ice_chill]

Awesome Job Guys.

+1 and lots of respect to your team - Good luck with everything, I know your going to bring a really special and innovative product to the table.

If there is anything myself or anyone on my team can do for you please don't hesitate to reach out. We are are competitors but we still are friends

Good luck!

Tom
tom@btcfpga.com
(315) 514-0269

Easy to say that when you have a superior product, will see how well wishing you are to BFL

[flynn]

Very interesting.

Amazingly small chips. Do you know yet how many of them will be on one board ?

[kaerf]

Avalon since inception has been designed to be an stand-alone unit, embed Linux openWRT equipped with WIFI capabilities and an Ethernet Port, while the competition is designed to operated over USB, depended on the need of a host computer. Continuing our trend, we have narrowed down our power estimates from 600W to 400W and this number is guaranteed to decrease further. Did I mention our power supply is a standard ATX?

Will you eventually be selling a non-standalone version or will the unit come with the ability to disable the openWRT hardware? What are the power estimates for the non-mining (linux host) hardware?

With such high power usage, it seems you'd need to at least double the hash rate to be competitive from a pricing standpoint.

strictly speaking on power being the only variable I wouldn't even suggest to "mine-for-profit" basis with power rates higher than $0.10 ( and you are unable to get write offs or reductions. )

When designing your hardware, please don't assume that only people who can get power for <= $0.10 would like to "mine for profit". Power usage is a significant concern if your competitors are offering 60GH/s @ 60W. To be competitive (assuming similar hash rates and cost per device), you really need to have something less than ~200W (personally, I'd really only see it being competitive at <150W for similar hash rates).

that said, i'm happy there is competition in the marketplace and commend the avalon team on their efforts.

[ice_chill]

Yes to above post, because when the difficulty increases far enough, it's the running costs that will determine the lifespan of the product.

[crazyates]

Retail Pricing.
$1,299? Yep, to stay with the spirit of the competition. $1,299 for 66Gh/s.

Good news! Do you have any idea when the non-preorder, retail ordering will start?

[jjshabadoo]

Hey Ice_chill,

Hard for Tom to be friendly with BFL when Josh trolls his thread mercilessly.