1) So the update said pretty much what I was expecting. The software is almost done, we have changed the power profile and had to change the board design to accommodate which is almost done, we are shipping out the dev units, and doing all the little 100's of details that make for a properly engineered and tested product. No surprise here. In fact, encouraging as it gives insight into exactly where everything stands.
I think we are 2 months or less at this point.
Notice that this is not surprising exactly because they have been saying these kind of things since october.
Also, please read the latest BFL propaganda again.
They want to stress that the chips are absolutely fine, except that they use up more power than was designed. But that's still ok, according to BFL.
But it's not. If the ICs use up a lot more power than designed for than there is something bad with the chips as well.
You have to look for those details because if there is one thing that i've learned about the way BFL operates then its that this little "the chips are fine" will turn into "We'll have to redo most of the ICs" in a couple of weeks.
The problems with the board were also fixable 'with a little tuning' and they would have delivered their product a couple of weeks ago.
Now the story has progressed over a couple of communications from 'Just some minor power problems with the board, will be done in a few days' to 'we need new boards and have no real shipping date at the moment'.
I guess that's just the difference in perception between us. Having been involved in a myriad of products from the initial design all they way to finished product... I can tell you from personal experience that having chips come out drawing more power or generating more heat than expected isn't the end of the production cycle. Sure it can change the game. But you don't just junk the entire design and do something else.
Once upon a time intel was developing Silverthorne - it was designed to an x86 processor for cellphones. They were trying to compete with motorola and mitsubishi. The tech worked, but needed more power and generated way more heat than expected. This tech didn't just dry up and go away... The changed the sales model, rebranded it under centrino as "Atom" and pushed it for handheld devices like tables and netbooks. This lead directly to the explosion of tablets as a consumer product. I don't believe there's a single cell phone that has an atom in it... but there are micro computers (in use as terminals and budget pcs) and hundreds types of tablets and 'netbooks' (micro laptops) that use them.
Nice story! Never knew atom was designed as a phone cpu.
But even then, intel is meaningless in the phone/tablet market. They never were able to catch up.
I seriously hope for everyone involved that BFL won't screw up their market like intel has been doing the past decade.
BFL is on the right track - redesign the board to supply the power needed while keeping temps stable, change the sales specs and ship the coffee warmer to consume the old tech board that would have been singles. Personally I think the very best move they could make would be devolping something where the top of it was a massive heatsink. and making it modular, to be plugged into the rack assembly they would sell. I'm thinking 4u rack that takes 3 mini-single modules. maybe just 4 chips per device.
I think the biggest hurdle for them is giving up the tiny size they had from FPGA days, get over it, go a bit bigger so you can cool the thing effectively... etc. Give me a metal packaged chip (or 4) that on a board that has AM3 holes for mounting a cpu heatsink and fan on top of it... that would make me very happy. Design the entire thing with a stacked heatsink that has a pair of 120mm fans on each side... design them so they could be assembled into a wall via interlocks on the sides, top and bottom of the case... one side air intake and other side of the wall being the exhaust... have the thing blow through.
Modular would be cool.
In fact, i don't understand why they try to make such good
looking products while the shelf life is pretty short.
I'd rather buy something more, eeh, russian (
) in design that can come to market faster.
Thermal design is a bit hard to do if your ic's use more power than anticipated. It screws up everything about what you thought will happen inside the box.
BFL had designed their board and cases before even having a prototype.
Now that the chips are out of spec they need to redesign the rest of the system to accomodate.
I'm not sure, but i'm getting the idea that their yields are pretty broad, quality wise. Which would suggest that they either have to divide the batches in quality stacks or they have to make sure every assembled box is cool enough to handle the worst cases.
I think it depends on the spread of the yield.
But there is something fishy going on whatever it is. I have the feeling they are not telling the whole truth about the situation.
Two weeks ago they were saying that they will probably ship the product last week. Now they say they are waiting for the revised boards to be shipped to them which means they are a week late from their last prognosis and still dont have the boards.
My guess is they're finding new problems but time will tell.