On the next version of the Zero (without adding cost)

[alphanumeric]

Would people be willing to pay $10, or would that put the Zero too close to the other models pricewise, for somewhat lesser features.

I'd pay $10. It also has attributes the other Pi's simply don't have, particularly lowest size and low current draw.

One-off price is pretty immaterial for most people, just a note of a different colour, for multiples it's not that much more expensive because there is some saving in shipping. Buying ten individually at £4 each with £2.50 shipping costs £65, buying ten at £9 each with £5 shipping is £95. It's more but not excessively.

Not sure how much it could save but a Pi Nought-Point-Five, a Zero without HDMI, without CSI, without DSI, just one USB power plus data socket, might do well because many people won't be using those in micro-clusters, embedded applications, or in USB gadget mode. Basically a SoC breakout board.

One thing to bear in mind is how many sales are being lost because it's not readily available in quantity.

The Zero without HDMI, without CSI, without DSI already exists. Its called the Compute Module. And there is a new version 3 just out. https://www.raspberrypi.org/blog/#compu ... e-3-launch Costs a bit more than 5 dollars though.

[hippy]

The Zero without HDMI, without CSI, without DSI already exists. Its called the Compute Module. And there is a new version 3 just out. https://www.raspberrypi.org/blog/#compu ... e-3-launch Costs a bit more than 5 dollars though.

Cost is not the only issue, it is usability by hobbyists and home users; that is probably the background to wanting additional pads to make user's lives easier or to reduce costs of use to a bare minimum.

The Pi Zero has multiple target markets. Some want 'everything' on a $5 board but are willing to pay a bit more. I think we would all like that but the trouble is that adding more doesn't increase the profitability of the thing unless price go up correspondingly and profitability is the main factor in lack of availability.

Pragmatically the solution is to increase profitability. Either by increasing prices ( and possibly including things which are cheap but make the product more appealing to justify that increase ) or to shave component costs to maximise profit at a lower price.

A further thought I had was on total cost. A single Zero currently costs £4 plus £2.50 and probably £4 for an SD Card; ten would cost £105. If Zeroes had an easy to use boot-over USB option ( allowing SD or USB ), with Zeroes at £9 each, using USB boot, it would cost about the same. That won't help all cases but could be good enough for most people to find a $10 priced Zero acceptable.

[i486]

I don't have much visibility, but I suspect that it would need a minimum price of $10 to start to increase production levels, ie it would need at least a $5 hike to get large scale manufacturers interested. $5 simply doesn't have the margins for large scale production. Would people be willing to pay $10, or would that put the Zero too close to the other models pricewise, for somewhat lesser features.

Maybe the manufacturers can equip Zero with good plastic case and sell for $10.

[mattmiller]

by jamesh » Wed Jan 18, 2017 10:16 am

I don't have much visibility, but I suspect that it would need a minimum price of $10 to start to increase production levels, ie it would need at least a $5 hike to get large scale manufacturers interested. $5 simply doesn't have the margins for large scale production. Would people be willing to pay $10, or would that put the Zero too close to the other models pricewise, for somewhat lesser features.

Add in full sized USB socket holes and throw in some header pins and yes please

[alphanumeric]

by jamesh » Wed Jan 18, 2017 10:16 am

I don't have much visibility, but I suspect that it would need a minimum price of $10 to start to increase production levels, ie it would need at least a $5 hike to get large scale manufacturers interested. $5 simply doesn't have the margins for large scale production. Would people be willing to pay $10, or would that put the Zero too close to the other models pricewise, for somewhat lesser features.

Add in full sized USB socket holes and throw in some header pins and yes please

That's why I used an A+ in one of my projects. Full sized HDMI and a full sized USB port. No messing with adapters. It's twice as big as a Zero but that wasn't a problem for me. It's mounted to my sense hat which has the same dimensions. Win, win, for me in that situation. I didn't need that added bulk of 4 USB ports and the Ethernet Jack that comes with a 2B or 3B. I'm hoping to swap out the A+ with the rumored 3A at some point. Assuming it will have the same footprint as the current A+.

[alphanumeric]

The Zero without HDMI, without CSI, without DSI already exists. Its called the Compute Module. And there is a new version 3 just out. https://www.raspberrypi.org/blog/#compu ... e-3-launch Costs a bit more than 5 dollars though.

Cost is not the only issue, it is usability by hobbyists and home users; that is probably the background to wanting additional pads to make user's lives easier or to reduce costs of use to a bare minimum.

The Pi Zero has multiple target markets. Some want 'everything' on a $5 board but are willing to pay a bit more. I think we would all like that but the trouble is that adding more doesn't increase the profitability of the thing unless price go up correspondingly and profitability is the main factor in lack of availability.

Pragmatically the solution is to increase profitability. Either by increasing prices ( and possibly including things which are cheap but make the product more appealing to justify that increase ) or to shave component costs to maximise profit at a lower price.

A further thought I had was on total cost. A single Zero currently costs £4 plus £2.50 and probably £4 for an SD Card; ten would cost £105. If Zeroes had an easy to use boot-over USB option ( allowing SD or USB ), with Zeroes at £9 each, using USB boot, it would cost about the same. That won't help all cases but could be good enough for most people to find a $10 priced Zero acceptable.

Yeah I know. I just wasn't sure if he knew of the Compute Module. I probably should have just let it go. The Compute Module is likely the least user friendly PI in the bunch. No jacks of any type, and no GPIO pins. I had absolutely no interest in it until I saw what Western Digital was offering for it. Now I'm looking to buy one of the new CM3's and put it in a Compute Stick.

[i486]

Yeah I know. I just wasn't sure if he knew of the Compute Module.

And this price of Compute Module is? Much higher than Zero.

[alphanumeric]

Yeah I know. I just wasn't sure if he knew of the Compute Module.

And this price of Compute Module is? Much higher than Zero.

Yes, it is. It's a different animal though. Onboard flash RAM etc. I shouldn't have brought it up, it's a bit off topic, sorry.

[mattmiller]

Assuming it will have the same footprint as the current A+.

mm - if they could squeeze it down to PiZero footprint I think I'd prefer it because then we could choose to buy one PiZero/day or order in a load of A3+ (as long as keenly priced of course - $25 would't really be worth paying when pi3 = $35

I think a $15 A3 (with WiFi/BT) would be a massive seller as the item to use for deployed projects if RPi can't move on Zero availabilty

[ejolson]

I don't have much visibility, but I suspect that it would need a minimum price of $10 to start to increase production levels, ie it would need at least a $5 hike to get large scale manufacturers interested. $5 simply doesn't have the margins for large scale production. Would people be willing to pay $10, or would that put the Zero too close to the other models pricewise, for somewhat lesser features.

The Zero is uniquely smaller and more energy efficient. Maybe a series of special editions with colored circuit boards that sell for $10 would work. The plain $5 version could still fulfill the role of a one-per-person giveaway to promote computer literacy. The blue, yellow, red, purple and black versions would be available in larger quantities at a price of $10 each.

Admittedly this is "the same version of the Pi Zero but increasing the cost" which, being the opposite of the original idea, is a bit off topic.

[laurent]

On the top of my wishlist for a Pi Zero upgrade is there a BCM2836 or 2837 single core CPU instead of the BCM2835 old ARMv6 one, in order to have a better efficiency.
The actual Pi Zero draws a little power making it a good choice for solar or battery powered system. But drawing a little power for a long time is not that efficient.
A BCM2836 or BCM2837 with a single core and a frequency reduced to keep the same power consumption of the actual BCM2835 should offer a nice horsepower for battery or solar powered projects, for the same consumption.

I'm currently waiting for the incoming Pi 3 model A for that purpose.

[davidcoton]

On the top of my wishlist for a Pi Zero upgrade is there a BCM2836 or 2837 single core CPU instead of the BCM2835 old ARMv6 one, in order to have a better efficiency.
The actual Pi Zero draws a little power making it a good choice for solar or battery powered system. But drawing a little power for a long time is not that efficient.
A BCM2836 or BCM2837 with a single core and a frequency reduced to keep the same power consumption of the actual BCM2835 should offer a nice horsepower for battery or solar powered projects, for the same consumption.

I'm currently waiting for the incoming Pi 3 model A for that purpose.

BCM2836 is dead, that's why we have a Pi2B2 using 2837.
And the BCM2837 is just too power hungry for Pi0 type applications.
There is also a problem with RAM chips that can be easily and cheaply soldered -- POP works, the current 2836/7 packaging doesn't.

So, we need the next gen (28nm) SOC, and knowing how clever the folks at RPF and Broadcom can be, they will be designing (by now, possibly even testing) something that will work for both Pi4 and a low-speed low-power option (maybe even software selectable), or possibly two versions, single core for Pi0v2 (512MB POP RAM) and multi-core for high performance Pi4 (probably >1GB, non-POP RAM).

[mattmiller]

Admittedly this is "the same version of the Pi Zero but increasing the cost" which, being the opposite of the original idea, is a bit off topic.

Don't worry - the original idea is just an obvious win/win for everyone so that's a done deal and we should expect them out in June I'd thought

Love the diff coloured idea for the $10 one sin quantity

[W. H. Heydt]

I think a $15 A3 (with WiFi/BT) would be a massive seller as the item to use for deployed projects if RPi can't move on Zero availabilty

The A+512MB is $20. Upgrading the the '2837 SoC, (presumably) 1GB RAM, and (possibly) WiFi/BT would make it harder to drop the price. Of course, Dr. Upton does seem to be able to pull rabbits out of his hat on a regular basis (witness the lower prices on the CM3 compared to what they were--still are in some places--for the CM1), so you never know.

[W. H. Heydt]

So, we need the next gen (28nm) SOC, and knowing how clever the folks at RPF and Broadcom can be, they will be designing (by now, possibly even testing) something that will work for both Pi4 and a low-speed low-power option (maybe even software selectable), or possibly two versions, single core for Pi0v2 (512MB POP RAM) and multi-core for high performance Pi4 (probably >1GB, non-POP RAM).

You are somewhat more optimistic than I am. By now, I would expect that whatever SoC is going to wind up in the Pi4B is probably in Silicon and being tested. Much as I think a single core Cortex-A53 SoC for the Pi Zero would be an excellent idea, I seriously doubt that one is in the works, UNLESS working up such a chip was done as an exercise to gain familiarity with the 28nm process node. Even then, it's an extreme long-shot that that was done, and an even longer stretch that it would wind up as a production SoC.

The other factor here that would make a Cortex-A53 (or, actually, *any* ARMv8 core) Pi Zero desirable would be to unify all Pi product lines with the ability to run a 64-bit OS. Note that this is NOT a call to actually move to 64-bit, at least not in the short (i.e. 5 years) time frame. But it would enable people to create 64-bit OSes that would run on all Pis. With the release of the CM3/CM3L and the pending Pi3A, all other Pi lines will have that ability.

[alphanumeric]

Assuming it will have the same footprint as the current A+.

mm - if they could squeeze it down to PiZero footprint I think I'd prefer it because then we could choose to buy one PiZero/day or order in a load of A3+ (as long as keenly priced of course - $25 would't really be worth paying when pi3 = $35

I think a $15 A3 (with WiFi/BT) would be a massive seller as the item to use for deployed projects if RPi can't move on Zero availabilty

I don't see it being Pi Zero sized. Not and have CSI, DSI, a full sized HDMI and one full sized USB port. I'll be happy if its the same size as the current A+ but with WIFI and BT added. And quad core with 1 gig of RAM. =)

[i486]

And quad core with 1 gig of RAM. =)

... with overheat

[laurent]

On the top of my wishlist for a Pi Zero upgrade is there a BCM2836 or 2837 single core CPU instead of the BCM2835 old ARMv6 one, in order to have a better efficiency.
The actual Pi Zero draws a little power making it a good choice for solar or battery powered system. But drawing a little power for a long time is not that efficient.
A BCM2836 or BCM2837 with a single core and a frequency reduced to keep the same power consumption of the actual BCM2835 should offer a nice horsepower for battery or solar powered projects, for the same consumption.

I'm currently waiting for the incoming Pi 3 model A for that purpose.

BCM2836 is dead, that's why we have a Pi2B2 using 2837.
And the BCM2837 is just too power hungry for Pi0 type applications.
There is also a problem with RAM chips that can be easily and cheaply soldered -- POP works, the current 2836/7 packaging doesn't.

So, we need the next gen (28nm) SOC, and knowing how clever the folks at RPF and Broadcom can be, they will be designing (by now, possibly even testing) something that will work for both Pi4 and a low-speed low-power option (maybe even software selectable), or possibly two versions, single core for Pi0v2 (512MB POP RAM) and multi-core for high performance Pi4 (probably >1GB, non-POP RAM).

2836 is dead because the demand is too low to maintain the production. With a new product with it, the demand could raise up to justify to produce them again. I don't absolutely want to see a 2836, I just want to say that nothing is impossible.

2837 is too power hungry ? According to the number of cores and frequency of the Pi 3's one, indeed.
But if the cortex A53 has a better efficiency compared to the ARM11, it means that for a given equivalent power consumption, the A53 will beat the ARM11. So, a equivalent power consumption may just mean a reduced number of cores and frequency !

Like any other manufacturers, Broadcom may have some BCM2837 (or 2836) aren't capable of running all their 4 cores and/or their full frequencies (those are simply rejected).
Keep in mind that the Pi Zero targets the lowest price ! Harvesting previously rejected 2836/37 batches may be a low cost solution.

I remember to have already discussed about the POP package here and it wasn't such a problem. Technically it wasn't if I'm right, but after, it's harder to find a huge choice of POP memories.

[W. H. Heydt]

Like any other manufacturers, Broadcom may have some BCM2837 (or 2836) aren't capable of running all their 4 cores and/or their full frequencies (those are simply rejected).
Keep in mind that the Pi Zero targets the lowest price ! Harvesting previously rejected 2836/37 batches may be a low cost solution.

While clever, and it *has* be done by both Intel and AMD, you are relying on at least two conditions that are by no means certain to be true. The first is the assumption that the BCM2837 has a way to disable any power going to bad cores. The second is that the chips that test bad have their failures in 1 to 3 CPU cores. Statistically, this is unlikely because the overwhelming majority of the circuitry on the SoC is part of the VC4, so one would expect something like 98% of bad chips to have perfectly good ARM cores and a bad VC4. As a result, even if the first issue would permit the idea to work, the second would mean that the "yield" would be far too low to sustain production of Pi Zero 3 boards.

So, plus points for a reasonably clever idea, but I really don't think it would work.

[alphanumeric]

And quad core with 1 gig of RAM. =)

... with overheat

https://www.adafruit.com/?q=heat%20sink& I have heat sinks on all my Pi's, even my Zero's. I just use a low profile for my Zero's and a tall one for my 3B's.

[laurent]

Like any other manufacturers, Broadcom may have some BCM2837 (or 2836) aren't capable of running all their 4 cores and/or their full frequencies (those are simply rejected).
Keep in mind that the Pi Zero targets the lowest price ! Harvesting previously rejected 2836/37 batches may be a low cost solution.

While clever, and it *has* be done by both Intel and AMD, you are relying on at least two conditions that are by no means certain to be true. The first is the assumption that the BCM2837 has a way to disable any power going to bad cores. The second is that the chips that test bad have their failures in 1 to 3 CPU cores. Statistically, this is unlikely because the overwhelming majority of the circuitry on the SoC is part of the VC4, so one would expect something like 98% of bad chips to have perfectly good ARM cores and a bad VC4. As a result, even if the first issue would permit the idea to work, the second would mean that the "yield" would be far too low to sustain production of Pi Zero 3 boards.

So, plus points for a reasonably clever idea, but I really don't think it would work.

You're right, I forgot this.
I can't wait to make some test with the upcoming 3A with some parameters tuning to achieve a comparable consumption to see the efficiency difference (even if some test may probably be done with a 3B with some tricky peripherals deactivation tough).

[ejolson]

You're right, I forgot this.

My understanding is that some ARM-based SBC's will reduce clock speed first when overheating and disable cores second if the overheating persists. This means, at least for thermal regulation, some SOC's can disable cores. While the Pi 3B could be tested by placing it in a toaster and running cpuburn, bread tastes better. Does anyone know how and if it is possible to disable cores in a running Pi 3B to reduce heat and power consumption?

[gregeric]

the overwhelming majority of the circuitry on the SoC is part of the VC4

You can just about make out what i think are the arm cores in this video: https://www.youtube.com/watch?v=V4idnxE5AbE - two pulsing white dots towards the GPIO header, and maybe two on the opposite edge. There was another video I've seen showing this more clearly, but of course I cannot find it right now. I seem to recall seeing four in a row in that one, but I may well be wrong in that.

When four cores were added, the package jumped from 12x12mm to 14x14mm, hence the pop ram no longer fits on top - another reason you couldn't repackage faulty 2836/7 dies for the Zero.

[jamesh]

You can try maxcpus=1 on the kernel command line to see if there are any power savings by only running one core.