|
Frizz23 (OP)
|
 |
January 16, 2013, 07:00:15 PM
Last edit: January 16, 2013, 10:40:45 PM by Frizz23 |
|
And the winner is: BFL_Josh of course!
Short: We had our ASICs ready waaaaaaaay back in 2012 [yeah, right. LOL], but we decided not to ship and redesign our product so that we can give you guys muuuuuuuuch better devices. We just need 30 days more (as usual). Honest Abe. Long: (attention, lots of blah blah blah) https://forums.butterflylabs.com/announcements/692-bfl-asic-status.htmlHello everyone. It's been about a month since our last update and it's high time we had another, since we have lots of information to share this time around.
To start off with, our chip production is going well. I am traveling to our packaging facility in California on Wednesday the 16th of January to do a walk through with the lead developer on our ASIC team as well as speak with some of the engineers to make sure everyone is on the same page. We have confirmed with the fab directly (not through any intermediaries) that our chips will be done no later than the 31st of January and ready for delivery. I will be at the foundry prior to that date to speak with some of the engineers and accept delivery of a portion of the chips to hand deliver to our packaging facility. This means our delivery date is expected the week of the 10th of February. We know it's not the end of October and we want to deeply apologize for the delays. However, I think that once you get to the end of this post many people will understand why we've been delayed and hopefully how it will ultimately benefit all of our customers a great deal. So please, read on!
So, I know the question first and foremost on everyone's mind (besides when we are shipping) is "Why are we delayed so long?"
Several things have happened to push us out from our target date... most things have been minor, but they have added up and I have detailed many of them in previous updates and on various threads throughout the forums. However, the biggest delay we've incurred has been twofold:
We absolutely, positively want to ship a chip that works perfectly. We don't want to send a chip that has problem or limitations. To that end, we have made some optimizations to the final layers of silicon that we feel will benefit everyone going forward due to point two:
The biggest single reason for the delay is due to a package change.
When we announced the delay last month, we decided to go back and switch from our QFN package to a flip chip BGA (FCBGA) package.
This necessitated a lot of changes with regards to the PCB and packaging (obviously) and caused us to really scramble to make things happen in a reasonable time frame. We did this for a number of reasons, but primarily it was due to heat concerns. Our chips would have functioned as currently spec'd within our thermal tolerances, but we would have very little thermal headroom to really crank things up. With an eye towards the future, we decided to bite the bullet and release a product that is ready for the long haul as opposed to releasing something now that would require exotic cooling to be pushed past 80 GH/s or so.
When redesigning things, we have made some various internal changes that may cause our power usage to potentially rise slightly beyond our original spec, and we are still tweaking things from other angles to bring power usage back down. Originally, we were at about .8w per GH (Yes, we built 20% into our announcements) and we are estimating that it's possible that a worst case scenario would lead to a bit under 1.2w per GH. If that ends up being the case, then yes we will have missed our power targets and yes we will gladly pay out the bounty. However, at less than 72 watts for 60 GH/s, it's still more than 5x better than any other product. However, we still believe that our .8w / GH target is what we will ship, but we like to prepare for the worst case scenario.
I will post more information once I have a better handle on our power usage.
I am sure many people may wonder why a flip chip BGA package is better than a QFN package (and are likely wondering what exactly is the difference, but that is a discussion for another thread). In a nutshell, our QFN package has plastic on top of it. Plastic is a terrible conductor of heat, so the heat stays trapped in the chip and shed out through vias on the bottom of the board. In extreme situations, where there's lots of heat and the vias can't handle the load you end up with heat migration to places you really don't want it to be, such as the ground plane, etc... and this affects surrounding components. With a flip chip BGA package, the chip is essentially flipped over and filled with silicon on top. Silicon is an excellent thermal conductor, so the heat has no problems being shunted away from the rest of the system. With our top mounted HSF, this makes a huge impact. Originally we had potentially planned on doing a bottom mounted HSF and using the vias. It turned out that the vias could not handle the thermal load and heat would migrate all along our thermal and ground planes and had the potential to overheat a number of other components. At 60 GH/s, heat was still within thermal tolerances, but only barely. If you were in an unusually hot environment, it could potentially cross the thermal threshold and the unit would start to fail and/or throttle. Switching to this new package essentially completely eliminates this, as almost all of the heat can be wicked away through the top of the chip, leaving the board and surrounding components completely cool.
If you are wondering what a flip chip package looks like, many (most) laptop or otherwise permanently mounted CPUs are Flip Chip BGA packaging (FCBGA). In retrospect, of course this is the route we should have gone, but we believed at the time that the QFN package was adequate for our needs. Hindsight is great for making predictions and if we knew then what we knew now... but that aside, we are almost to the end of the right track now. The QFN package is not suitable for dense, high speed bitcoin mining applications. Much like the toggle rate issue in the FPGA era, issues that are not normally a problem for traditional ASIC applications become monster issues when the bitcoin beast sinks it's teeth into them, surprising and biting engineers in the butt. The only way to find some of these things out is to get bitten by them, since this is new territory for everyone.
In other news, we are implementing a queuing system in the MCU of our units, to allow multiple jobs to be queued up. This should prevent any communication latency bottlenecks. It was an often requested feature, so we are happy to say that we will be headed in that direction. With the announcement of our new Android based mining application, I want to confirm that we are in contact with developers of the other major mining applications, BFGMiner, CGminer and Bitminter. We will be sending development units to them for integration and testing as soon as we can and we are confident all three mining applications, in addition to our own Easyminer and Android miner will be ready for our units when we ship.
I have often fielded the question as to if Chinese New Year will affect our shipping. No, Chinese New Year it will not affect us. We have all of our parts needed to build the units in stock, with the exception of the chips. We have taken delivery of all of our Chinese made components at this time and they are filling up our warehouse and the assembly plant warehouse. Our fab is not located in China and our chips will be done before CNY in any event. Our packaging plant is located in California and obviously not affected by CNY, our assembly house is located in the US and thus also not affected.
With regards to upgrading your shipping, our new system should be online this week. Once we are comfortable with it, you should be able to email or phone us to upgrade your shipping to DHL, EMS, FedEx or UPS as your convenience. Email is best of course, and probably the fastest way to start the process. We have been working with Bitpay to get the ability to send non-time dependent invoices to customers and we are now able to do that, so if you want to pay in BTC, we can send you an invoice which can be paid any time as opposed with within 15 minutes. This will make upgrading your shipping or products much easier on everyone.
Our new manufacturing facility is almost complete. We will be posting a video walk through hopefully by the end of the week so that you can see the facility and all the parts we have, ready and waiting for the chips to finish baking (Mmmm delicious chips). We will accept visitors, but please call before you decide to drop by so that we know when to expect you and can make time to show you around. We are (understandably) very busy right now and I do not want to turn anyone away because everyone is already engaged with other tasks. I will be traveling a lot, as will several other members of our team, so make sure we are there if you want to talk to us about anything.
I have been in contact with both Yochdog and Kano and have a tentative date scheduled for their visit. I will be confirming dates with both of them later this week and arranging the travel plans as soon as we have everything nailed down.
I am often asked the question "If I order now, when do you expect my order to ship." This is a hard question to answer, since we don't know exactly how long it takes to assemble a unit. However, the best estimate we have for this is that if you order today, 13 January 2013, you would likely receive your order around the middle of March or possibly sooner if you're ordering a Jalapeno or a Single. Minirigs take much more time to build than the other product lines, so they will take longer to deliver. However, without actually going through a bulk assembly process, we aren't sure what the major bottle necks will be. We can do estimates and we have strived for simplicity in everything related to the SC line. To that end, it requires 4 screws per Jalapeno and 6 screws per Single. Assembly of these units should take less than 5 minutes (probably around 3 minutes) per unit, so working through our pre-order backlog should be very quick. It will take more time to box and label them than it will to assemble them. The March ship date is a worst case scenario and it's entirely possible you would get your unit much sooner than that.
|
