Topic: Already delays in BFL shipment plans? - page 34. (Read 49567 times)

This doesn't take into account someone who is buying one unit. What the F happens to them if they get the "shitty" one ?

They get screwed while the rest who got "lucky" benefit.

Jeez folks, BFL sure does take some shet here, some for good reason, some not.

This is NOT a good reason to give them a hard time. If you were so concerned about a single day of operating your unit, you should have made your own device.

BFL has to back up their product. If they even send out 100 defective units which could have been caught by better testing, they have 100 more pissed off folks potentially and 100 more transactions, etc. etc.

You can get a lot of the benefit of a longer burn in by running at elevated temperature.

I actually do consider 24 hours to be a short burn-in test. You've never overclocked your CPU and done a Prime95 test for 48+ hours?

No.  I think it could be proven that short burn-in times have low risk and high rewards.

Now you guys are just applying logic. Where's the fun in that?

Indeed. Run a burn-in on a fixed number batch for 24 hours, check failure number of units after 2 hours, after 6 hours, ectetera. Apply burn-in test duration that proved safe to rely on for next batches.

Their practice of doing a 24 hour burn in has been discussed before. I'm perfectly ok with receiving my unit a day or two later if it means guaranteeing it will work when I get it. Of course there's the chance I might make a few BTC less, but I'd just be greedy if I said the Risk>Reward.

+1

I agree with mm here.  People always say you should burn your stuff in for a couple of days to make sure everything is good, but honestly, after a few hours of burn in, if something is going to have failed, it probably will have failed.  BFL could easily shift to say 6 hour burn in cycle and get 4x the amount of burn ins done per day and probably not notice a huge bump in failure rates of units that get sent out.  There's not a whole lot of moving parts in these units, and other than stuff dying almost right away, I'd bet you won't really start seeing failures till people run these units for months at a time 24x7, and BFL can never simulate that.

Hopefully not the same people in those 24 hours.

I, like many other people, have a lot of vacation saved up. I propose coming to BFL headquarters to help with the assembly for a week. You can pay me in product instead of money.

Well then why not a burn in for 48 hours... or 96 hours... or how about a whole week?

Because the math doesn't work.  And it doesn't work for 24 hours, unless the amount of defective units that can not be identified within a 1 hour burn in is enormous, like 20% of the total manufactured products.

If they test it for 24 hours... everyone still gets theirs at the same time from BFL (just 1 day later), but it is guaranteed to work. It makes way more sense to burn them for 24 hours to make sure they work then to screw someone by testing it for an hour, shipping it, then it doesn't work and they have to wait about a week to ship it back and get shipped a replacement. I would rather wait that extra 24 hours for a product that works vs taking a chance (most likely a really small chance) that it is defective and be down for a week.

I can't see why a 24 hour burn in would be helpful on average.  If the Butterfly Labs community could prove to you that they would take as little as a 1 hour burn in, would you do so?

I think it is important to consider the following...

Suppose you're shipping 300 units (to a single entity, assuming the BFL buyers have a substantive consensus they would like a short burn-in) and you expect 5% to have major problems, surely, many of those problems would be identified within an hour.  But suppose none of the problems would be identified in an hour.  Now, out of the 300 units shipped, 15 of them have serious problems.  Which would the consumer prefer?  Would he rather have 285 units hashing away a day earlier than 300 units a day later?

Let's look at the math...

To make the numbers easy, suppose each unit is 1GH.  Now the customer, in that single day, has 285GH (300, less the 15 RMA's).  At the current difficulty, he would earn 86.75 with the 285 units on day 1.  Suppose it takes a week before his 15 rigs arrive after he sends them back.  So after 1 week, he earns 607.25 BTC, and then for the following week, he earns 639.24 (7days * 91.32BTC per day).  After 2 weeks, even having to wait a substantial period of time for his RMA's, he still makes 1,246.49BTC.

Let's see how it works out if he has to wait a day...

He makes 91.32BTC per day, but only has 13 days to mine to catch up (you held on to all 300 of his units for 24 hours before shipping) so he makes in his first potential two weeks... only 1,187.16.

The community on average, becomes more and more worse off the longer you withhold their products from them, even in for burn-ins.

In addition, suppose someone needs to do an RMA, could they not simply send the device back and have BFL mine for them with a deposit of some sort?  "I bought 1 rig and it is a dud.  I'll send you back the rig along with an immediate deposit of 3BTC.  In the meantime, start up another rig and let it mine on my pool account at your facility until you receive the bad unit and ship out the good unit.  Then return the 3BTC as well."

Is this not a workable solution?

Or do you expect a much greater percentage of malfunctioning products than 5%, even with a 1 hour burn in?

Scrybe pretty much describes the process.  Yes, I would like to have every order shipped by the end of the year, but it may be around mid January before we do that for the people ordering right now.  I plan on putting my people on 24 hour shifts until all back orders are assembled and shipped.  

Right now, the bottleneck will be the 24 hour burnin period, not the assembly... and I'm looking at ways to mitigate that or increase our burn in capacity.  

- Awesome Dick out

This implies that they have hired ppl who are already proficient at operating the SMT machines.

With that said, their statement that they "are currently being installed in our new facility" implied that they had already received the machines, but when questioned, they backpedaled and admitted they did not in fact possess the machines yet.  I wonder if the same is true of this "capable team".

When you preorder something, you get it when you get it...plain and simple. Don't want the stress? Don't pay in advance for a product that hasn't made it into production and shipped yet...

You are surprised by this?

They were pretty clear what was in the first shipping batch long before they ever announced how large the FABRICATION RUN was. These are independent quantities, one related to the very expensive and automated FAB process is done very quickly in high volume, the other related to the number of folks and tools they have for doing final assembly.

The 40 ppl comment was a "IF we do them all at once" scenario, but that would be irresponsible because we would have total noobs with this hardware ruining the boards for 2 weeks instead of getting a few folks trained right and having the boards come out over a couple months. They would also have to fire most of the 40 workers after the 2 week contract, as well as some of their permanent employees if their shipping forecasts are too low.

I would expect that August orders will be filled in late December, Sept and October orders through the end of January. Of course if their are additional supply or reliability issues then it will slip back more, but this is my baseline expectation at this point.

Wasn't it stated that all preorders should be shipped by Jan,but the first 500 or so (no amount has been confirmed though) would be the 1/3,1/3,1/3

I read it somewhere awhile ago,can't seem to find the posts now,grrr 

Posted under Fair Use for the purposes of disussion.

Courtesy of: https://forums.butterflylabs.com/showthread.php/331-manufacturing-process-question



Conclusion? There may be significant delays. [speculation and personal opinion]

I still can't decide if Inaba is awesome or a dick. I think both. Awesome dick. That should be his tag.