new RPi model B planned soon

Guys, one thing that seems to be forgotten is that the RPI foundation apparently got stuck with a batch of 256Mb POP memory chips. Those are going on the model As that hopefully hit the streets soon.

When you are a supermarket, people walk in and expect to be able to buy things immediately. "1L milk, yes madam, I'll order it right away, it will be delivered here next wednesday. Goodbye madam!". Not acceptable.

Also many shops have the hopes of selling MORE if they keep stock. Someone somewhere might be looking for a product and go somewhere else if you don't have it NOW.

However in other markets, notably the BCM2835 market, are different. A chip fab is very very unlikely  (and probably contractually prohibited) to keep stock. Then there is broadcom. They apparently live in a market where they don't know in advance how many chips of an item will be sold. So they keep maybe enough stock to provide samples, but for production quanitities, they have a lead time. So they can schedule making about as many chips as have been sold, and/or shift production ratios between chips as the demand requires.

As I believe that Farnell has sold about 430k raspberries, they apparently correctly estimated such an amount of BCM2835's and ordered them in time for august/september/october/november delivery.

RS however started delivery of raspberry pi orders number 130k and up somewhere late october, early november. At the point those orders were taken, the one-per-customer rule hadn't been lifted yet (or not for long anyway).  So at that point in time, they either had sold 130k or 30k (if they started the orders at 100k).

For Farnell to order 300k extra BCM2835's before they had sold them that too is a leap of faith. A gamble that has paid out well.

Now in hindsight you might say that this was not a gamble. But it was. At first the expected salesvolume was 10k. But on launchday that number probably needed adjustment to 100k. Fine. But if you order 300k more chips will they be sold? Maybe, just maybe, someone comes out with a $30 board with better specs than the 'pi and all of a sudden sales drop. Then, what are you going to do with those extra BCM2835s?

Guys, one thing that seems to be forgotten is that the RPI foundation apparently got stuck with a batch of 256Mb POP memory chips. Those are going on the model As that hopefully hit the streets soon.

When you are a supermarket, people walk in and expect to be able to buy things immediately. "1L milk, yes madam, I'll order it right away, it will be delivered here next wednesday. Goodbye madam!". Not acceptable.

Also many shops have the hopes of selling MORE if they keep stock. Someone somewhere might be looking for a product and go somewhere else if you don't have it NOW.

However in other markets, notably the BCM2835 market, are different. A chip fab is very very unlikely  (and probably contractually prohibited) to keep stock. Then there is broadcom. They apparently live in a market where they don't know in advance how many chips of an item will be sold. So they keep maybe enough stock to provide samples, but for production quanitities, they have a lead time. So they can schedule making about as many chips as have been sold, and/or shift production ratios between chips as the demand requires.

As I believe that Farnell has sold about 430k raspberries, they apparently correctly estimated such an amount of BCM2835's and ordered them in time for august/september/october/november delivery.

RS however started delivery of raspberry pi orders number 130k and up somewhere late october, early november. At the point those orders were taken, the one-per-customer rule hadn't been lifted yet (or not for long anyway).  So at that point in time, they either had sold 130k or 30k (if they started the orders at 100k).

For Farnell to order 300k extra BCM2835's before they had sold them that too is a leap of faith. A gamble that has paid out well.

Now in hindsight you might say that this was not a gamble. But it was. At first the expected salesvolume was 10k. But on launchday that number probably needed adjustment to 100k. Fine. But if you order 300k more chips will they be sold? Maybe, just maybe, someone comes out with a $30 board with better specs than the 'pi and all of a sudden sales drop. Then, what are you going to do with those extra BCM2835s?

> Guys, one thing that seems to be forgotten is that the RPI foundation apparently got stuck with a batch of 256Mb POP memory chips. Those are going on the model As that hopefully hit the streets soon.

I don't think anyone's forgotten that.  But so far there are more questions than answers.

Why would RPF get stuck owning chips if the distributors do all the manufacturing?

Do the distributors buy their ram chips from RPF?   If so, do they buy the SoC or other chips

from RPF as well?  Do they buy unpopulated boards from RPF?  Or populated boards?

Does this shed any light on why RS boards have the Sony "made in the UK" silkscreen?

Does this shed any light on why both distributors started shipping 512MB boards on

the same date?  If RPF is in the middle of things, does RPF's limited capital restrict

the production rate?   Did the distributors balk at making Model A's?

>RS however started delivery of raspberry pi orders number 130k and up somewhere late october, early november. At the point those orders were taken, the one-per-customer rule hadn't been lifted yet (or not for long anyway).  So at that point in time, they either had sold 130k or 30k (if they started the orders at 100k).

They are reportedly up to around order number 169K.  But as Jim Manley famously

pointed out, these are order numbers, not units shipped, because there have been

quite a few reports of cancellations.  I think the Allied order numbers are separate,

although probably not very large.

The lead time seems to be different every time it is reported, and it isn't ever clear

whether the chips being delivered now were ordered 30 weeks ago, or the chips

ordered now will be delivered 30 weeks hence.   I don't think either distributor

would have any boards to sell today if they were just now getting chips ordered

30 weeks ago. If RPF starts accepting orders for Model A's in March, does that

mean they won't be delivered until October?

Roger Wolff wrote:

 

At first the expected salesvolume was 10k.

No it wasn't.  10k was merely the quantity for which RPF could afford to buy BOM components.  The interest was at fever pitch on the web site/mailing list almost from the start, and they knew very early on (long before launch) that 100k+ would be wanted immediately under one-per-person rules, and more once the restriction was lifted.  That wasn't speculation, but direct feedback.  No hindsight needed.

In addition, all rational analyses were predicting that a vastly greater pool of interest existed beyond the techies that typically inhabit online venues.  That was speculation, granted, but very well informed speculation, and the way the Pi was continually being promoted was ensuring that this extra interest would just grow and grow.  The promotion was done in ordinary news channels after all, not just techie ones, and media centre functionality was so frequently mentioned that the promotion was clearly targeting people outside of techdom as well.

So no, it's not reasonable to claim that only with hindight could the level of demand we see today have been known.

And of course, a few weeks after the chaos of launch day, any doubt that might have been held about the levels of demand disappeared completely, so suggestions that the level of demand was unknown at any time after the first weeks of March are completely erroneous.

coder27 wrote:

 

The lead time seems to be different every time it is reported, and it isn't ever clear

whether the chips being delivered now were ordered 30 weeks ago, or the chips

ordered now will be delivered 30 weeks hence.   I don't think either distributor

would have any boards to sell today if they were just now getting chips ordered

30 weeks ago. If RPF starts accepting orders for Model A's in March, does that

mean they won't be delivered until October?

Indeed.  The information we've been given just doesn't hang together at all.  And now it has to be reconciled with:

Lizwrote:

 

Farnell told us yesterday that they've shipped 429,000 so far, and I don't have current figures for RS, but they'll be only a little lower)

which of course means that RS could not have been ordering a lot less than Farnell on average, since at the current date they've shipped nearly as many as Farnell (+/- Liz's phrase "little lower").  Any claim that RS were underordering initially by a huge amount has to be met with acceptance that RS were ordering vastly more than Farnell later on, otherwise they could not have shipped roughly similar amounts by today.

I would take "little lower" with a few grains of salt, given that order #169xxx,

ordered on July 21, was reportedly shipped yesterday.

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=23&t=22508&start=181

[sorry if the formatting comes out badly. I tried editing it in the "full editor", couldn't do what I wanted,

I then edited some parts in the HTML, and got what I wanted, according to the "full editor". But

the final result shows it quite badly again. Sigh. ]

coder27 wrote:

 

I don't think anyone's forgotten that.  But so far there are more questions than answers.

Why would RPF get stuck owning chips if the distributors do all the manufacturing?

The foundation made the first batch. It is unlikely that the minimum order quantity of all chips exactly line up at 10000 units.

Or, someone at the Foundation had the smarts to think ahead and order some long-lead-time memory chips.

Do the distributors buy their ram chips from RPF? 

No.

 If so, do they buy the SoC or other chips from RPF as well?  

No.

 Do they buy unpopulated boards from RPF?  

No.

  Or populated boards? 

No.

 
Does this shed any light on why RS boards have the Sony "made in the UK" silkscreen? 

No. I suspect that Pete Lomas provided the new 2.0 CAD files to Farnell (who were going to make them in UK) and expected to redo them (just the silkscreen) for RS when they needed a new batch. He forgot, didn't have time, didn't bother.

Does this shed any light on why both distributors started shipping 512MB boards 
on the same date? 

No. I think that's a contractual obligation between RPI foundation, Element14 and RS. Remember when the "one per order" limit was lifted? RS still had an order queue of about 5-6 months, while Farnell was down to a few weeks. 

 If RPF is in the middle of things, does RPF's limited capital restrict the
production rate?  

No. They have been offered capital and refused: we have enough of that. Recently Liz has said they have plenty of money.

  Did the distributors balk at making Model A's? 

No. They might be a bit reluctant to take it up. Especially if their markup can be a bit higher on the  $35 version. Even from the beginning I've thought that is a steep markup: $10 for an ethernet/usb chip. Right now the ethernet USB adapters (including PCB and connectors!) sell for $1 on ebay. The Lan9512 sells for EUR 3.41 at mouser if you buy 1000. So price pressure is bigger on the model A. 

>> RS however started delivery of raspberry pi orders number 130k and up

>> somewhere late october, early november. At the point those orders were

>> taken, the one-per-customer rule hadn't been lifted yet (or not for long

>> anyway).  So at that point in time, they either had sold 130k or 30k (if

>>> they started the orders at 100k).

>They are reportedly up to around order number 169K.  But as Jim Manley famously

> pointed out, these are order numbers, not units shipped, because there have been

> quite a few reports of cancellations.

I skimped over: Any multiple-units-per-order would probably cancel the cancelations. It's a rough estimate anyway. And has the 100k uncertainty if nobody can remember wether or not someone reported an order number of under 100k.

 
  I think the Allied order numbers are separate,
although probably not very large.
 
The lead time seems to be different every time it is reported, and it isn't ever clear
whether the chips being delivered now were ordered 30 weeks ago, or the chips
ordered now will be delivered 30 weeks hence.   I don't think either distributor
would have any boards to sell today if they were just now getting chips ordered 
30 weeks ago. 

I expect that the lead times vary according to demand.  So somewhere may/june, farnell snatched all the BCM2835s that were available for delivery august/september. 8-12 weeks lead time. And the next available batch of BCM2835s would be delivered october.

If RPF starts accepting orders for Model A's in March, does that mean they won't be delivered until October?

Nah! I don't think so. First: the "march" date comes from a post from Liz saying you'd surely be able to buy them in March. Implying March 2012, as she said that in Feb 2012. Secondly I have the impression that the BCM2835 situation is clearing up. Reasonable stock levels are now in place in the places where they are needed.

There is this "game". There are several teams (1 to 5 people per team). Preferably about 5 or 6. The first one is the factory. They have to decide how many goods they are going to produce. Next there is an importer, wholesaler and finally a chain-of-retailers. Everybody is told not to tell their suppliers how much they sold, everybody pays cost on keeping stock etc. The factory is started off with "Last month you sold 1000, how many are you going to produce for the next month?". Similarly the importer is started with: "Last month you sold 1000, how many are you going to order for the next month?". And so on down the chain. But nobody knows what the others have been told. Everybody submits their order for the next (virtual) month, and the game continues.

From the outside with full information everybody would benefit from everybody ordering 1000 units for the next month. But people want to "bet" on sales increasing and shy away from having to say no to a customer. So they want to build a small stock, so first the demand exceeds production. Then to prevent that from happening again, they order more from their suppliers, who start thinking sustained demand is higher than it really is. You get enormous dynamics before things settle down.

In the "real world", the actual numbers are dynamic as well. Demand is not the stable 1000 units as in the game. So people in control have been playing the game for a while already and at the factory they know that they will obtain the best profit if they keep low stocks. Chips go out of fashion really quickly.

I'm really struggluing to understand why you all find this so interesting and keep going on about it. Analysing the Foundation and their distributors sales, orders and all that fluff is all rather pointless I've decided. SO many conspiracies theories being banded about, so many almost certainly rubbish  assumption, when if people just took a quick look, took everything at face value, its all seems fairly bloody obvious. Trying to attach all the other nonsense it really a waste of time. - perhaps people time woudl be better spent actually doing something interesting, like using the damn thing?

It all comes down to...

Unexpected popularity, bad order forcasting by the manufacturers, long chip lead times. Add on a few minor contractual arguments behind the scenes (I thinking the Foundation having the 256's for the Model A) , and you have explained everything.

There, I've simplified it all for you. No need to thank me, although a tinny wouldn't go amiss.

Bill

Roger, a formatting hint.  The way I overcome the idiosyncracies of both forum editors that you've experienced is as follows.

Use the Quote Previous Message (QPM) button to bring in the whole post, and delete everything below the line to which you want to respond.  Add your comment below it.  (This is what you did, once.)

Then for each new section of the post to which you want to reply, use QPM again, delete the top section (including the "XXX writes:") which you already included earlier, and delete the section below the next block to which you want to respond, leaving only the part you want to reference, and write your stuff below it.  Rinse and repeat.

The referenced blocks then come out nice and tidy, which seems to be impossible if you just use QPM once.  The quoted blocks don't react well to being split.

IMHO: that's a workaround which works for me if you need to quote two or three sections. I thought I had found another workaround. I had: I showed fine in the "full editor".

coder27 wrote:

 

I would take "little lower" with a few grains of salt, given that order #169xxx,

ordered on July 21, was reportedly shipped yesterday.

 

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=23&t=22508&start=181

I really don't know what to make of it, coder.  The only way forward that comes to mind is to ask for the apparent discrepancy to be explained, and for an actual units shipped figure to be given for RS as they did for Farnell.  Whether it's worth risking your RPF account to obtain a response, well, that's up to each individual to decide.

Really, is it that important? We can guess at the numbers a bit. 30k shipped, 60k shipped, 160k shipped. Or even more. But my guess is it's one of the lower numbers. This would be an embarresment for RS to admit. So they won't. Farnell on the other hand can boast about their number. So they do.

Raspberry pi foundation probably should know about the numbers (because they get a percentage), but contractually they probably don't have the right to tell anybody.

So, RPF won't tell you. RS won't tell you. And my best guess is 60k shipped.