new RPi model B planned soon

I'd not venture yet to say that the model A will work better USB wise since the problem is not the LAN9512 but the limitations imposed by the USB IP on the BCM2835 which is an OTG implementation than a full featured USB HOST, plus some other issues in the driver code.

Without the LAN9512 will alleviate some power and thermal issues, still not really clear how much better things will work with an ethernet interface hanging from USB.

Still power hungry (as any other ethernet PHY that sinks current through the ethernet transformer) one experiment I'm planning is to hook up a standalone ethernet controller such as the Microchip ENC28J60 (10BaseT) or ENC424J600 (10/100BaseT) via the SPI interface. These controllers are not for heavy duty networking stuff but may work well in many applications, I know that there are some Linux drivers for the ENC28J60, didn't search yet for the ENC424J600.

Model A could be a good alternative for a headless WiFi configuration, the power budget gained by removing the LAN9512 can be used to power a decent WiFi USB dongle.

About browser speed, I didn't play with it much, but didn't notice a huge difference using Midori with the 256MB or 512MB board, it runs awfully slow in both. Still X is not hardware accelerated so I don't know if that could be the root problem.

My .02

-J

Only experience will tell, but I'm hoping (for RPF's sake, not mine) that the Model A's single USB directly plugged into an external Ethernet controller with no hubs nor any other USB device in sight will manage to stay within the narrow bounds of what the crappy USB implementation on the BCM2835 can handle without errors.

We'll just have to wait and see.

To be frank I didn't have the time to dig too much into the USB problems on the BCM2835, also I don't know how crappy Synopsys IP actually is, but what I can tell from previous experiences is that it will be very challenging to make USB OTG to fully behave as a HOST, I think that some people were working trying to put together some abstraction layer on the driver so Linux sees it as a host but I still think it is a long stretch and there will be some compromises/limitations, so Model A or B, USB host IMHO will still suck.

BTW, there is a more recent thread about USB issues on the RPF Forum http://www.raspberrypi.org/phpBB3/viewtopic.php?f=28&t=23544

And after the "pub" yada yada, there is only a single update from Gordon and it does not say much, or what it says is not much ...

Edit: Ohhh and I just noticed another post from the "friendly moderator" essentially telling people to back off reporting problems they already know exist, what a nice gesture.

Morgaine Dinova wrote:

 

As usual the fanbois tried to deny everything and paint RS as the guilty party, but coder27's links above make the situation clear.  Not even RPF can obtain more SoCs for Model A, every one produced reduces the number of BCM2835 parts for Model B.  The posts indicate that the restrictions are entirely at Broadcom's end.

 

Why are you talking as if the RPF has anything to do with the purchasing and distributing of the SoCs?  It's a very well know fact that they have 100% outsourced the production of the device to Farnell and RS.  Liz says "we" but knowing they don't actually produce the device, it would stand to reason that she is speaking for the business partners; something she's done often.  With a 30 week lead time, not purchasing enough SoCs would result in some crappy results like about 7 months of constantly running out of supplies because you didn't anticipate demand.  We've discussed this before, you certainly know a lot about general things and electronics, but you are out of your zone of knowledge when it comes to real life fabrication environments.  Not that I mean to absolve B-com from all responsibility here but RS started having backlog issues well within 7 months of the pi being released.  With the firm fact that there's a 30 week lead time for the SoC, it would be impossible for the problem too have been on B-com's end as all they were doing was filling the demand that RS had already requested.

Edit: Also the comment would seem to indicate that the parts being reduced by making model As is referring to the parts on hand, which again would be an issue of the one populating the board purchasing a large enough supply 7 months ago and not a matter of B-com not making enough SoCs today.

Oh, and I don't trust anything they say for timing.  No matter who well intentioned their estimates are something always goes wrong and we're left waiting for another 6 months.

You are completely missing the point.  RPF has said clear as daylight that every Model A that could be produced would result in one less Model B being produced.  Who orders what is irrelevant.  Not enough BCM2835 devices are being produced by Broadcom, period.

If you are trying to suggest that Broadcom would deliberately switch a fab line off BCM2835 devices or run it at less than full capacity despite knowing full well that the Pi is back-ordered around the world then you are painting Broadcom as incompetent and deliberately holding back the Pi ecosystem and not being interested in selling product.   After all they know with 100% certainty that every BCM2835 manufactured would be sold tomorrow if it were available.  I find it incredible that anyone would think Broadcom is that stupid and willfully losing out on revenue.  The mind boggles.

So no, no, no, no.  The scarcity has absolutely nothing to do with poor order planning by one partner, because that wouldn't alter the picture even if it were true (and indeed it might be true, I'm not saying it's not, but only that it's irrelevant).  You're listening too much to RPF excuses (which you should realize operate very much as an arm of Broadcom) and not thinking about what your interpretation would say about Broadcom's commercial interest.  They are in this to make money.  If they're not printing money by churning out chips faster than they are now when they know full well that they could sell every one instantly, then you can be certain that it has nothing to do with missing paperwork.  The only answer that makes commercial sense is that they don't have any more capacity.

Morgaine.

Morgaine Dinova wrote:

 

You are completely missing the point.  RPF has said clear as daylight that every Model A that could be produced would result in one less Model B being produced.  Who orders what is irrelevant.  Not enough BCM2835 devices are being produced by Broadcom, period.

 

If you are trying to suggest that Broadcom would deliberately switch a fab line off BCM2835 devices or run it at less than full capacity despite knowing full well that the Pi is back-ordered around the world then you are painting Broadcom as incompetent and deliberately holding back the Pi ecosystem and not being interested in selling product.   After all they know with 100% certainty that every BCM2835 manufactured would be sold tomorrow if it were available.  I find it incredible that anyone would think Broadcom is that stupid and willfully losing out on revenue.  The mind boggles.

 

So no, no, no, no.  The scarcity has absolutely nothing to do with poor order planning by one partner, because that wouldn't alter the picture even if it were true (and indeed it might be true, I'm not saying it's not, but only that it's irrelevant).  You're listening too much to RPF excuses (which you should realize operate very much as an arm of Broadcom) and not thinking about what your interpretation would say about Broadcom's commercial interest.  They are in this to make money.  If they're not printing money by churning out chips faster than they are now when they know full well that they could sell every one instantly, then you can be certain that it has nothing to do with missing paperwork.  The only answer that makes commercial sense is that they don't have any more capacity.

 

Morgaine.

Again, we've discussed this before but I'll say it again.  That's not how fabrication works.  There is not a single line dedicated to each chip.  You'll have a line for each genre of devices (hard drive platters, CPU, etc), but you will not have a "BCM2835" line.  This is why they have such a high production time.  You schedule a period of time to produce a specific chip in, once its turn is done you move to the next.  B-com like any other fab out there will build only to meet what has been ordered, and while that might mean they build a little extra to fill out the line, odds are they have an order increment that forces the customer to fill the line.

I'll also would like to point out again that the product availability being dicussed by Liz would quite logically being the amount they have in hand, a quantity that would have been ordered 7 months ago by the respective distributor and has nothing to do with how quickly or how much b-com could produce with an order placed today.  B-com isn't going to stop manufacture of its other chips just to alter an order for this "low cost" chip to meet the demand for it within the Rpi community.  That's just bad business practice.

I'm afraid what you say doesn't make commercial sense for Broadcom nor for RS nor for RPF, so I don't believe it holds water.  But let me try a completely different tack to demonstrate that your explanation just doesn't work.

Let's assume just for a moment that Broadcom has a mind of its own and is not just a fab contract house as you make it out to be.  Let's say that they have their eyes open and notice that the demand for BCM2835 is huge and that there is zero risk associated with producing devices not to order but for buffer stock.  And let's assume that they did so.

Now, assuming that this happened, what would be the result of Broadcom's sales department picking up the phone and dialing RS's number and saying "Oh RS, guess what, we have the N devices that you wished you had ordered 500 years ago in stock, do you want them?"  (There could even be a small markup over and above contract order pricing.)

Now don't tell me that you think RS would say "Nah, not interested mate.  We're totally happy having our reputation dragged through the mud like it is now thanks, we'll talk to you again in mid-2013 for our next batch."

It would stretch incredulity, so I can't believe that there is any possibility at all of such a response.  RS would love to slash its waiting lists to nil overnight if it could, and it would if given half a chance.  RS would win, Broadcom would win (through slightly more profit for them and a happier Pi ecosystem), and RPF would win massively by Model A not being delayed by the SoC shortage and by being able to strut around with a big smiles on their faces after proclaiming "No more delivery queues anywhere, order freely now."

When everybody would win in this picture, why theorize that Broadcom is totally inept in business and thwarted by some missing paperwork?  It just doesn't make sense.

I'm not often on the side of Broadcom because of their total dislike of openess, but I do believe that they have the business sense to print money when that option is available to them.  And it is.

Morgaine.

Morgaine Dinova wrote:

 

I'm afraid what you say doesn't make commercial sense for Broadcom nor for RS nor for RPF, so I don't believe it holds water.  But let me try a completely different tack to demonstrate that your explanation just doesn't work.

 

Let's just assume for a moment that Broadcom has a mind of its own and is not just a fab contract house as you make it out to be.  Let's say that they have their eyes open and notice that the demand for BCM2835 is huge and that there is zero risk associated with producing devices not to order but for buffer stock.  And let's assume that they did so.

 

Now, assuming that this happened, what would be the result of Broadcom's sales department picking up the phone and dialing RS's number and saying "Oh RS, guess what, we have the N devices that you wished you had ordered 500 years ago in stock, do you want them?"  (There could even be a small markup over and above contract order pricing.)

 

Now don't tell me that you think RS would say "Nah, not interested mate.  We're totally happy having our reputation dragged through the mud like it is now thanks, we'll talk to you again in mid-2013 for our next batch."

 

It would stretch incredulity, so I can't believe that there is any possibility at all of such a response.  RS would love to slash its waiting lists to nil overnight if it would, and it would if given half a chance.  RS would win, Broadcom would win (through slightly more profit for them and a happier Pi ecosystem), and RPF would win massively by Model A not being delayed by the SoC shortage and by being able to strut around with a big smiles on their faces after proclaiming "No more delivery queues anywhere, order freely now."

 

When everybody would win in this picture, why theorize that Broadcom is totally inept in business and thwarted by some missing paperwork?  It just doesn't make sense.

 

I'm not often on the side of Broadcom because of their total dislike of openess, but I do believe that they have the business sense to print money when that option is available to them.  And it is.

 

Morgaine.

If only that was the way fabrication worked!  It certainly would have made my life easier, but alas!  It isn't.  The fabrication people would get beat (sued) to no end if they called up someone and say "hey, your competitor is buying X times the amount you are.  Maybe they know something you don't, would you like to order more?"  That's both bad for business and breaking NDAs.  Now don't get me wrong, you have laid out some mighty fine logic there, extremely well thought out!  But it's obvious you've never worked in a fab plant because you keep talking nonsense in this area (no offense meant, it's just the plain truth).  Now I have real world experience in a fabrication environment, you can keep calling it crazy (which it most certainly is) and trying reject it, but it really is the way these places work! 

It also helps to keep in mind that B-com doesn't care about everyone winning, they care about b-com winning.  And B-com has most certainly won throughout this affair.  More importantly they have to keep winning by keeping their schedules with all their customers.  As crazy successful as the Pi has been, they're just a small part of their big picture their Q4 earnings were 1.82 Billion Dollars.  So far the pi has sold something in the range of .5 Million units (according to Liz's projections).  If they pay 5 dollars per chip that would only represent 2.5 Million, at 10 dollars a pop it would still only be 5 Million, heck if they made $35 of every board it will still only be 17.5 Million dollars in B-com piggy bank or what 1/100th of a percent of their Quarterly income?  B-com has bigger money makers than the pi and their version of winning is making sure that the big income makers stay on schedule.