new RPi model B planned soon

If the Model A is planned for March 2013, then RPF's banner headline will have declared the Pi as the $25 computer for a whole year by then.  Isn't there a law against such blatant long-term false advertising?

Morgaine Dinova wrote:

 

If the Model A is planned for March 2013, then RPF's banner headline will have declared the Pi as the $25 computer for a whole year by then.  Isn't there a law against such blatant long-term false advertising?

coder27 is being gamesome.  The link is to a 20 Feb 2012 comment where liz promises that "you'll definitely be able to buy Model As in March".  True, she doesn't specify the year.

It's speculative right up to the point someone orders them - until then, they are  chips sitting on a shelf in a warehouse. And look, fabless companies don't like warehouses, they like to make stuff and ship it stright to the user - it's much cheaper. That's how they work. Now I reckon they can sell a shitload, but even Apple under order stuff just in case. And remember, there is a 30wk (still don't beleive that one) lead time - that's over 6 months, a lot can happen in 6 months.

Still, as I said, this is nothing to do with Broadcoms capacity as you have been arguing, since they don't actually have ANY capacity at all. It's all contracted out, and the people they contract out to have LOTS of capacity, and will make whatever you ask. So, rather than changing the subject, absorb that information, and explain how it's not the fault of RS and Farnell's ordering too few.

Billy Thornton wrote:

 

It's speculative right up to the point someone orders them - until then, they are  chips sitting on a shelf in a warehouse.

It's not speculation when there is no uncertainty or risk involved, and you have not offered any viable scenario containing uncertainty or risk.  And the chips don't need to sit on any shelves, they can be offered in advance very trivially.  Indeed this is totally normal, no different to offering for preorder.

Still, as I said, this is nothing to do with Broadcoms capacity as you have been arguing, since they don't actually have ANY capacity at all. It's all contracted out, and the people they contract out to have LOTS of capacity, and will make whatever you ask.

You can't have it both ways.  If Broadcom's capacity is all contracted out, and the people they contract out to have LOTS of capacity, and will make whatever Broadcom asks, then Broadcom's inability to produce SoCs at a higher rate and with less lead time is even less comprehensible.  Third parties would be falling over themselves to manufacture more BCM2835 for Broadcom given the demand for Pi worldwide, because "LOTS of capacity" that is twiddling its thumbs is very costly.

That doesn't seem to be what is happening here.  Admittedly we're speculating, but some things look more likely than others, and a surplus of capacity on which to produce BCM2835 seems extraordinarily unlikely.

Morgaine Dinova wrote:

 

You can't have it both ways.  If Broadcom's capacity is all contracted out, and the people they contract out to have LOTS of capacity, and will make whatever Broadcom asks, then Broadcom's inability to produce SoCs at a higher rate and with less lead time is even less comprehensible.  Third parties would be falling over themselves to manufacture more BCM2835 for Broadcom given the demand for Pi worldwide, because "LOTS of capacity" that is twiddling its thumbs is very costly.

 

That doesn't seem to be what is happening here.  Admittedly we're speculating, but some things look more likely than others, and a surplus of capacity on which to produce BCM2835 seems extraordinarily unlikely.

And that is where the ball returns to my court.  Fabs don't want to handle excess storage (of finished product) either.  So they won't keep a high capacity either (I was quite surprised to hear that B-com did not directly fab the chip themselves though).  You're logic is backwards though, the fact they don't have their own fab facility will actually increases turn around time as they become a middleman (so to speak) in the process.  They are bound by the fab's schedule which means they really can't adjust an order at the last minute.  Which again brings us full circle to the fact that it is the responsibility of the customer to appropriately plan for the demand of a product in accordance with the suppliers schedule. It should also be noted, again, that the pi is not some big ticket item so no, fabs will not be falling over themselves to manufacture the chip.  Besides, I'm willing to bet that B-com has some very specific NDAs and contracts that prevent them from doing just that in exchange for good deals.

And you really got to get the idea that not making the BCM2835 = wasted time or a waste of capacity.  Fab plants will build no more and no less than what has been ordered.  Any time they're not making the BCM2835, they're making other chips.  Fab facilities wait for no one!  Which makes more sense to you?  Fill a big order for a customer to meet deadlines, or building some extra low cost chips in the eventuality that a customer might want some more of it before the next run?  That a waste of time material and storage space, plus the fab facility has no guarantees that the chip will still be in demand in the future.  As much as I like the pi, it could flop at any minute.  People could get tired of the USB issues, or decide they'd rather pay double and get a little bit faster android device for their media center needs.  It's insanely bad business for a fab facility to make an excess of a product that the customer hasn't ordered rather than moving things onto filling the next product that has a customer waiting right now to purchase it.

Morgaine Dinova wrote:

 

 

However, you prefer to place the blame on RS and somehow twist it to look as if an order placed by RS could magically create new fabrication capacity at Broadcom.  It can't.  At most it can affect scheduling of what is fabricated on capacity-limited equipment so that RS's devices get manufactured before those of another customer.  The restricted fabrication capacity remains restricted though, whether it is producing devices for RS or for someone else.  An order won't bring a new fabrication plant online, that costs billions.  It is accurate to state (assuming of course that our speculation is correct) that Broadcom's lack of spare capacity is the limiting factor.

Aye yi yi, you just really don't get the way this works do you?  Turns out the explanation I typed up explaining how fab plants worked was far too complicated and long so it got trashed. I'll go from this approach.  RS had problems Farnell didn't.  If they had both had problems and the roku to boot than I'd be with you 100%!  If it worked the way you're suggesting it does than everyone would have issues equally.  Or at best it would rotate as each distributor got their share of limited capacity chip.  First RS would be short on supplies, then Farnell (then Roku).  Since this has obviously has not been the case, and since Billy has enlightened us that B-com is fabless and capacity would not be an issue for a company whose whole business model is getting the chips that have been ordered in the time frame they have been contractually limited to, then all signs indicate that the issue at hand was Farnell ordered enough to meet demand, Roku ordered enough to meet demand, and RS thought the Pi wouldn't be as successful as it has been and ordered a smaller quantity than Farnell.

Billy Thornton wrote:

 

... And remember, there is a 30wk (still don't [believe] that one) lead time - that's over 6 months, a lot can happen in 6 months...

The source on the 30 week lead time is liz herself.  Last September (2012) I heard Rob Bishop say 23 weeks, so it's gotten worse.  As you say, >6 months is a long time and fads can be short-lived.  For example, Richard Lester made A Hard Day's Night (1964) on a very short schedule since nobody knew how long the Beatles would be popular   So I can see why neither distributor would want to be stuck with unsellable BCM2835 inventory.

Interesting that the fab times are getting worse.  Maybe it's Apple booking up prime fab time so they can peevishly distance themselves further from Samsung.  Or the fabs are swamped making Allwinner A10s and A13s.

Used to work in the industry - times vary according to what's popular - probably new models of phone coming out soon - Christmas tends to cause a rush of new product hitting the market, so the fabs will be backed up which would result in longer lead times. Probably not Apple, although their production is moving away from Samsung, so might be going to TSMC or similar. The sort of volume there would dwarf the raspberry pi, so it's not surprising that's backed up. If I were a manufacturer I'd be building the high profit chips, not low profit stuff like the 2835 (or the Allwinners, but they would be in Chinese fabs anyway). Which is exactly what is happening.

Billy Thornton wrote:

 

... Probably not Apple, although their production is moving away from Samsung, so might be going to TSMC or similar. The sort of volume there would dwarf the raspberry pi, so it's not surprising that's backed up. If I were a manufacturer I'd be building the high profit chips, not low profit stuff like the 2835 (or the Allwinners, but they would be in Chinese fabs anyway). Which is exactly what is happening.

According to this Allwinner press release, the A10 is manufactured by TSMC, one of the fabs used by Broadcom.  So lots of competition for a limited amount of fab.

John Beetem wrote:

 

 

For example, Richard Lester made A Hard Day's Night (1964) on a very short schedule since nobody knew how long the Beatles would be popular

Ahh, the Fab Four

mynameisJim wrote:

 

the fact they don't have their own fab facility will actually increases turn around time as they become a middleman (so to speak) in the process.  They are bound by the fab's schedule which means they really can't adjust an order at the last minute.

That's not consistent with Billy's "the people they contract out to have LOTS of capacity, and will make whatever you ask".  Which is it, lots of capacity allowing Broadcom to enjoy a buyer's market, or very restricted capacity making fabrication a seller's market and giving Broadcom little to no options?

You're both attempting to substantiate the same conclusion from incompatible premises, so one of you is clearly wrong.  Even if each were true separately, they can't both be true simultaneously.

In addition to one of you being necessarily wrong, you're both making out that Broadcom's business planners are incompetent at capitalizing on a massive demand for one of their chips, and that suggests strongly that you're both wrong.  I may not like Broadcom's secrecy practices, but I see no sign that they are incompetent at getting devices made cheaply and in vast numbers, under either of your premises.  If they could make money from this insatiable demand, they would, even if the profit is less than when producing other devices.  It's still profitable for them, and nobody has suggested otherwise (yet).

By Occam's Razor, the fact that they aren't making money from it at the rate for which there is demand suggests that they can't do so, because competent business people (which I believe they are) would find a way if it were at all possible.  They've had since the start of March to fully realize the scale of demand and react to it in a way that would capitalize on the world's hunger for an old chip on which their investment has already been recouped, so it's all straight profit.  I do not think they would look this gift horse in the mouth.

So, you'll have to find a much stronger reason for the company's apparent inability to knock out large numbers of these SoCs than you have so far to be convincing.  They're not some 2-bit outfit like you make them out to be.  To suggest that they  ignored one half of this zero-risk profit stream because they got no order from RS is quite comical.

Morgaine.

I have bought three Raspberry PI's.

The longest I had to wait was 3 weeks for one purchased from Allied Electronics. Original delivery was given as 6 to 8 weeks.

While waiting for Allied, Element 14 advertised a 7 day wait so I bought one from them and it was delivered in 3 to 4 days.

When the 512MB version was announced I bought another one from Element 14 that was also delivered in 3 to 4 days.

When actual availability of the Raspberry Pi was first announced I signed up with RS to be notified when I could order one

from them the next day (couldn't get anything done on the first day because the web site was totally slashdoted). To this

day I am still waiting for that notification. In every web post I've read were some one has been waiting months for delivery

and the name of the company they ordered from was mentioned it was RS.

Most fabs usually have many customers. Some of the customers, like Broadcom have more than one type of device being

made. Each customer has a contract for a certain number of all the devices it wants that the fab is expected to produce. If

a fab is running at or near full capacity as you correctly say they try really hard to do they can't just magically turn on "more

capacity" when one of their customers has a large unexpected increase in demand for one of the devices. Unless other

customer(s) can reduce their needs, more time to adjust doesn't help either. Going to another fab to produce the device

requires at least a 6 to 8 month time frame assuming of course you can find one with unused capacity that supports a

compatable fabrication process and the tools you used to design the device.

As for Broadcoms secrecy practices, they are not much different from all of the other ARM SOC sellers. There is enough

information for most of the peripherals to write drivers for them. Graphics accelerators, multimedia accelerators, and DSPs

tend to be the most highly garded because in many cases the IP for them is actually from another company and they are

under a NDA to not release it. In some cases they may just be trying to not get sued for real or imagined (as in troll) patent

violations. Finally, and what I consider the worst reason is for some perceived competitive advantage. As demand for ARM

SOCs increase this veil of secrecy is slowly being lifted. Just a couple of days ago I read about a rumor that the PowerVR

graphics accelerator was going to get oficial support for an open source driver. I have also read that there is official support

from for an open source driver for the Mali graphics accelerator although I have not been able to find it with Google.

I have bought three Raspberry PI's.

The longest I had to wait was 3 weeks for one purchased from Allied Electronics. Original delivery was given as 6 to 8 weeks.

While waiting for Allied, Element 14 advertised a 7 day wait so I bought one from them and it was delivered in 3 to 4 days.

When the 512MB version was announced I bought another one from Element 14 that was also delivered in 3 to 4 days.

When actual availability of the Raspberry Pi was first announced I signed up with RS to be notified when I could order one

from them the next day (couldn't get anything done on the first day because the web site was totally slashdoted). To this

day I am still waiting for that notification. In every web post I've read were some one has been waiting months for delivery

and the name of the company they ordered from was mentioned it was RS.

Most fabs usually have many customers. Some of the customers, like Broadcom have more than one type of device being

made. Each customer has a contract for a certain number of all the devices it wants that the fab is expected to produce. If

a fab is running at or near full capacity as you correctly say they try really hard to do they can't just magically turn on "more

capacity" when one of their customers has a large unexpected increase in demand for one of the devices. Unless other

customer(s) can reduce their needs, more time to adjust doesn't help either. Going to another fab to produce the device

requires at least a 6 to 8 month time frame assuming of course you can find one with unused capacity that supports a

compatable fabrication process and the tools you used to design the device.

As for Broadcoms secrecy practices, they are not much different from all of the other ARM SOC sellers. There is enough

information for most of the peripherals to write drivers for them. Graphics accelerators, multimedia accelerators, and DSPs

tend to be the most highly garded because in many cases the IP for them is actually from another company and they are

under a NDA to not release it. In some cases they may just be trying to not get sued for real or imagined (as in troll) patent

violations. Finally, and what I consider the worst reason is for some perceived competitive advantage. As demand for ARM

SOCs increase this veil of secrecy is slowly being lifted. Just a couple of days ago I read about a rumor that the PowerVR

graphics accelerator was going to get oficial support for an open source driver. I have also read that there is official support

from for an open source driver for the Mali graphics accelerator although I have not been able to find it with Google.