I’ll be looking forward to power consumption figures (namely for idle, but also comparing joules per job against other models). One of my use cases are battery-backed raspis.
Hey Bjorn. Vaccine fridge temperature monitoring. Battery-backed more than battery-powered.
In my locale there are certain rules about how many hours above certain temperature thresholds vaccines can be held before you have to phone it in to NSW health; and potentially dump all of the vaccines (especially if you don’t have accurate logged data of the temp profile for the event). For obvious reasons they err on the side of caution, however losing vaccines is also a pretty big thing (getting replacements is slow and $$$$). When something bad happens you want detailed and accurate data.
Current solutions involve off-the-shelf fridge monitors. Beautiful little dedicated boxes that run off CR2032 cells and use a thermocouple. Unfortunately the ones I’ve used suck in a number of ways:
Recording buffer full? Ignore the circular write option (if available) and instead stop recording data immediately.
Need to get the data off the logger? Unplug it from fridge, plug into PC.
Logger unplugged from fridge? Collected data now has spurious 200degC+ peaks and noise.
Proprietary software that can’t zoom the x and y axes separately? Why, did you want to be able to ignore those 200deg peaks or something?
A much bigger problem with these units is that they are passive, not reactive. If something goes wrong they don’t tell you. Instead you have to regularly go through the hoops to extract, archive and analyse the data yourself. Doctors and nurses are busy, the less the have to do the better they can help patients.
Also note that most of these fridges have built-in temp monitoring and min-max reporting in the form of a few buttons and an LCD display (or sim). Staff tend to be more fluent with these interfaces (to a certain degree – accidental button presses have also been known to lead to full thaws :P). Unfortunately these don’t record time periods or full profiles, so they are good indicators to regularly check but useless for further analysis of events.
The current plans involve a raspi2 + a bank of D-cell alkaline batteries + a string of temp sensors going into the fridge. Basic concept:
Automatically records temps and emails a report of them out every week (solving the menial work and archiving problems).
Sends alerts live if temp-time thresholds are breached (eg fridge door left open).
Externally monitored so alerts are also sent if site power (and hence internet) are cut, or if the unit is otherwise abused.
Able to continue running on D cells in the meantime for somewhere between 1-2 days (by my calcs).
Simple dumb lights on the front of box to say “all is good” or “there’s a problem”.
I’ll be throwing this together over the next few weeks, then running it in parallel with the existing temp monitor solution. Depending on how hard the paperwork gets it might continue to exist in parallel, but when something goes wrong I suspect it will provide much more useful data (and alerts) than the current solutions.
(I could ramble further about thermal mass, temp sensors positioning, and how I think existing solutions are somewhat biased, but this is probably already too long :D).
Back to raspis: given the low workload for a fridge monitor the primary consideration is idle current draw. As other people have mentioned the pi 3’s tend to eat a lot more than the pi 2’s, and there are differently clocked variants of the same models too. This new pi 4’s SoC is fabbed differently, so it will be interesting to see whether or not they have put any effort into idle draw or not.
PO4 chemistry is nice. Unfortunately it’s a single cell design, so I’ll get only a tiny fraction of what 6 alkaline D cells will give me:
LifePO4 18650: 1.5Ah * 3.2V nom ~= 5Wh per cell
Aklaline D-cell: 15Ah * 1.5V nom ~= 25Wh per cell (and I plan to use 6 in series)
I originally considered rechargeables, but I think it’s not as great of an idea for this usage scenario. Blackouts happen seldom (eg once or twice a year) and the cost of new D cells is minimal to the business.
Most of all: as much as I expect to be maintaining these directly myself, if I can’t then I’d much prefer to be giving instructions over the phone on how to replace D-cells versus how to fix a puffy Lipo or failed Li* cylindrical cell.
Hi Hales, thanks for your detailed reply! Sounds like good use for the pi. Have you considered the pi zero (W) since you mentioned the idle current draw? My plan is building a mostly solar powered weather station which includes UV and CO2 measurements. As far as I can tell this combination is either hard to find or none-existing.
Pi zero: I think it came down to an issue of stock and availability when I did the parts order. Otherwise yes, the only thing I’d need to confirm is whether or not the Zero has a native 1-wire interface.
UV: which UV? What sort of sensor do you have in mind?
CO2: currently liasing with a company trying to use these. Be wary of any chip that says it measures “eCO2” or is otherwise a solid-state chip TVOC chip. If you read far enough in you discover that the CO2 levels are invented based off a lookup table describing a “common indoor” relationship between TVOC and CO2; ie complete lies with a long list of assumptions. It looks like you need devices bigger than a single chip (normally little plastic chambery things a few cm big on a PCB module, IR reflection internally?) to actually measure CO2.
I’ve tried making a laptop-type device on the Pi, and the power consumption of the 3b was just brutal; I had to fall back to the 2 to have any hope of making it run. Even then I was able to get under an hour of battery life on a battery that would last a PocketCHIP for a couple hours.
Edit: the pocketCHIP also has onboard circuitry for pass-thru charging of a lipo, which is a lot harder than you’d think to handle yourself when building something like that. It seems like they’re going full-speed-ahead on making the Pi a better desktop replacement and not really concerned much about the portable use case.
I gave this as a talk at CircleCityCon, and actively use the tablet to investigate and attack my own radio assets. Obviously, it could be built and used on non-self assets. But that would be bad, so don’t do that.
Mine was an XBMC system that could handle being suddenly powered down without warning. Power failures are common where my parents live in India and there is a blip in power when battery backup from the inverter kicks in. So I only needed a few seconds till household battery backup can take over.
In the end, I ran out of time on my last visit and just cloned a bunch of SD cards so they just sub in a new one when the old one breaks on power failure.
With USB3 support and onboard GigE, this could be the first Pi that could be used as a gigabit router (with a dongle). In the past I’ve used a Banana Pi, but I had some issues with it and eventually just built a Thin ITX router.
Using a full Thin ITX board is pretty expensive, but it looks like you could setup a Gigabit router and Wi-Fi AP (if that chip onboard support AP mode) with this new Pi for <$100. Very cool.
This looks like a great upgrade! It fixes all the major issues I’ve had with Pi: low resolution, low memory, micro-usb power.
It’s really thoughtful that they’re selling their own USB-C adapters and low-cost power supplies. USB-C power supply for a laptop would cost way more than the Pi.
Do you have any insight into whether its VideoCore VI will support modern GPU features? In particular, I’m talking about compute shaders with the ability to read and write arbitrary addresses into buffers, and threadgroup synchronization.
The Raspberry Pi foundation is a nonprofit, and I suspect the margins are quite slim. They probably make more selling the accessories. They also have tremendous volume and that helps keep the prices down.
It wouldnt surprise me because that’s how it works in the cell phone shops, too. Sales people in many tell me most of their commissions come from cases.
Well, the 35 dollars is before shipping/tax/other things and the 35 dollars is the cheapest model, 1GB ram, others cost more, so maybe the make the cheapest one at more of a loss. Other than that, I am not sure.
I think at least some of the original Raspberry Pi founders were also Broadcom employees, and RasPi boards are all built around Broadcom hardware. I’m guessing there was a lot of consensus-building within Broadcom leading up to the original RasPi’s release… it would be much more difficult to approach an SoC vendor as an external third-party and try to talk them into giving you the same open documentation and design support that RasPi gets from Broadcom.
This is more than powerful enough to be a really good living room retrogaming machine.
I’m currently playing through my favorites with my kids in chronological order (so far their favorites are Final Fantasy, Link to the Past, Mega Man X and Super Mario RPG: Legend of the Seven Stars, and Earthbound) but I’ve built up Paper Mario: The Thousand-Year Door so much that they’re gonna pop if I don’t get an old GameCube or a Dolphin instance going soon…
Since they seem to be positioning this as a credible (lightweight) desktop machine, I wonder if they’ve done anything about the SD card speeds. Improving the graphics, cpu and memory but leaving it stuck at ~20 MB/s on it’s only internal storage device makes it seem… not so compelling to me.
It’s good to see that SD perf has improved, even if not by all that much. (And the USB gains look great!)
I think my issue is not only physical space, but also an aversion to the idea of having the boot drive able to be accidentally unplugged. But maybe that’s just me…
I think it’s mostly a hardware limitation on the Pi, but I’m not an expert. Lots of people report similar boards having much better IO performance (e.g. Odroid), and also that the IO rates on SD cards are much better in an adapter on a laptop/desktop machine.
Note that most of the above benchmarks, when run on a USB 3.0 card reader on my MacBook Air, show 5, 10, or 15 times greater performance in that environment
I guess I could actually run some benchmarks myself when I get home tonight…
I want to use one of these as a router, but I’m concerned that all the USB3 to Gigabit Ethernet dongles will be garbage. Anyone have any experience with those, good or bad?
I used an Intel NUC with one eth interface as a router for about a year. I just hooked it up to a small managed switch to create a couple of vlans, one for internet and one for lan, that both get pushed to the port for the router. Then you just setup a couple of virtual interfaces in the OS and it work just like a device with two physical interfaces.
The only downside then is that you’re basically making your connection half-duplex. You can still get gigabit in either direction, but not both at the same time.
I was thinking the same thing when I saw USB3 support. My current router had an on-board Ethernet port die on it and I switched it out to a USB3-Ethernet gigabit adapter. It works fine. It’s just got a standard RealTek r8152 chip in it … the USB3-Ethernet chip is typically going to be a standard RealTek, Broadcom or Atheros on most of these devices and they should function the same as their PCI equivalents.
I considered the same thing by my conclusion is that GbE dongles are a mixed bag. Some don’t work, some work but fail intermittently. (Which is really what you don’t want on a router.) There must be some that work fine for a long time but I couldn’t find anyone who already did the research on this and shared it. I ended up buying a PCEngines APU2 for my router/firewall at home. It’s much more capable than a Raspberry Pi and only costs a bit more after you include all the dongles and drives and such that you would need for a Raspberry Pi.
I haven’t tried it, but this seems like an good board for a router: https://www.hardkernel.com/shop/odroid-h2/
And its uses X86 which is a plus or a minus depending on your point of view or usage.
Is there any shop in europe that has the 4GB model in stock?
I checked all approved resellers in Belgium, Poland, Czech Republich, Germany, Austria and the Netherlands, but without any luck.
I was looking at them on reichelt.de earlier today and the 4gb model is listed as available to ship on the 28th, so just a few days wait (assuming that’s accurate)
I’ll be looking forward to power consumption figures (namely for idle, but also comparing joules per job against other models). One of my use cases are battery-backed raspis.
Hopefully Larabel will provide on Phoronix in the next week (?) or so.
Hales, I’m curious. What are your use-cases for battery-powered Raspberry Pi’s?
Hey Bjorn. Vaccine fridge temperature monitoring. Battery-backed more than battery-powered.
In my locale there are certain rules about how many hours above certain temperature thresholds vaccines can be held before you have to phone it in to NSW health; and potentially dump all of the vaccines (especially if you don’t have accurate logged data of the temp profile for the event). For obvious reasons they err on the side of caution, however losing vaccines is also a pretty big thing (getting replacements is slow and $$$$). When something bad happens you want detailed and accurate data.
Current solutions involve off-the-shelf fridge monitors. Beautiful little dedicated boxes that run off CR2032 cells and use a thermocouple. Unfortunately the ones I’ve used suck in a number of ways:
A much bigger problem with these units is that they are passive, not reactive. If something goes wrong they don’t tell you. Instead you have to regularly go through the hoops to extract, archive and analyse the data yourself. Doctors and nurses are busy, the less the have to do the better they can help patients.
Also note that most of these fridges have built-in temp monitoring and min-max reporting in the form of a few buttons and an LCD display (or sim). Staff tend to be more fluent with these interfaces (to a certain degree – accidental button presses have also been known to lead to full thaws :P). Unfortunately these don’t record time periods or full profiles, so they are good indicators to regularly check but useless for further analysis of events.
The current plans involve a raspi2 + a bank of D-cell alkaline batteries + a string of temp sensors going into the fridge. Basic concept:
I’ll be throwing this together over the next few weeks, then running it in parallel with the existing temp monitor solution. Depending on how hard the paperwork gets it might continue to exist in parallel, but when something goes wrong I suspect it will provide much more useful data (and alerts) than the current solutions.
(I could ramble further about thermal mass, temp sensors positioning, and how I think existing solutions are somewhat biased, but this is probably already too long :D).
Back to raspis: given the low workload for a fridge monitor the primary consideration is idle current draw. As other people have mentioned the pi 3’s tend to eat a lot more than the pi 2’s, and there are differently clocked variants of the same models too. This new pi 4’s SoC is fabbed differently, so it will be interesting to see whether or not they have put any effort into idle draw or not.
You may be interested in https://lifepo4wered.com/lifepo4wered-pi+.html, which I’ve found makes a pretty good UPS right out of the box.
Thanks minimax.
PO4 chemistry is nice. Unfortunately it’s a single cell design, so I’ll get only a tiny fraction of what 6 alkaline D cells will give me:
I originally considered rechargeables, but I think it’s not as great of an idea for this usage scenario. Blackouts happen seldom (eg once or twice a year) and the cost of new D cells is minimal to the business.
Most of all: as much as I expect to be maintaining these directly myself, if I can’t then I’d much prefer to be giving instructions over the phone on how to replace D-cells versus how to fix a puffy Lipo or failed Li* cylindrical cell.
Hi Hales, thanks for your detailed reply! Sounds like good use for the pi. Have you considered the pi zero (W) since you mentioned the idle current draw? My plan is building a mostly solar powered weather station which includes UV and CO2 measurements. As far as I can tell this combination is either hard to find or none-existing.
Pi zero: I think it came down to an issue of stock and availability when I did the parts order. Otherwise yes, the only thing I’d need to confirm is whether or not the Zero has a native 1-wire interface.
UV: which UV? What sort of sensor do you have in mind?
CO2: currently liasing with a company trying to use these. Be wary of any chip that says it measures “eCO2” or is otherwise a solid-state chip TVOC chip. If you read far enough in you discover that the CO2 levels are invented based off a lookup table describing a “common indoor” relationship between TVOC and CO2; ie complete lies with a long list of assumptions. It looks like you need devices bigger than a single chip (normally little plastic chambery things a few cm big on a PCB module, IR reflection internally?) to actually measure CO2.
I’ve tried making a laptop-type device on the Pi, and the power consumption of the 3b was just brutal; I had to fall back to the 2 to have any hope of making it run. Even then I was able to get under an hour of battery life on a battery that would last a PocketCHIP for a couple hours.
Edit: the pocketCHIP also has onboard circuitry for pass-thru charging of a lipo, which is a lot harder than you’d think to handle yourself when building something like that. It seems like they’re going full-speed-ahead on making the Pi a better desktop replacement and not really concerned much about the portable use case.
This is my purpose: a SigInt tablet capable of 50MHz-1200MHz SDR, Mousejack, 802.11abgn
https://hackaday.com/2019/06/05/mobile-sigint-hacking-on-a-civilians-budget/
I gave this as a talk at CircleCityCon, and actively use the tablet to investigate and attack my own radio assets. Obviously, it could be built and used on non-self assets. But that would be bad, so don’t do that.
Mine was an XBMC system that could handle being suddenly powered down without warning. Power failures are common where my parents live in India and there is a blip in power when battery backup from the inverter kicks in. So I only needed a few seconds till household battery backup can take over.
In the end, I ran out of time on my last visit and just cloned a bunch of SD cards so they just sub in a new one when the old one breaks on power failure.
With USB3 support and onboard GigE, this could be the first Pi that could be used as a gigabit router (with a dongle). In the past I’ve used a Banana Pi, but I had some issues with it and eventually just built a Thin ITX router.
Using a full Thin ITX board is pretty expensive, but it looks like you could setup a Gigabit router and Wi-Fi AP (if that chip onboard support AP mode) with this new Pi for <$100. Very cool.
This looks like a great upgrade! It fixes all the major issues I’ve had with Pi: low resolution, low memory, micro-usb power.
It’s really thoughtful that they’re selling their own USB-C adapters and low-cost power supplies. USB-C power supply for a laptop would cost way more than the Pi.
Not first of April? Wow, I feel sorry for the rest of the SBC industry. A board like this would cost 135 dollars if it wasn’t a Raspberry Pi.
Still, it was about time they offered something competitive with higher priced boards.
oh hell yeah!
Nice! This is better than the weird 4xA53+2xA72 combo RK3399 uses. Except the RK3399 clocks up to 2.2GHz, I wonder how high you can OC this.
Do you have any insight into whether its VideoCore VI will support modern GPU features? In particular, I’m talking about compute shaders with the ability to read and write arbitrary addresses into buffers, and threadgroup synchronization.
I don’t know these specific ones, but a Vulkan driver was planned and GL compute shaders in general are supported.
I still kind of like the platform. But it does have glaring issues with it:
How are they doing this? (serious question) or is other hardware seriously overpriced?
The Raspberry Pi foundation is a nonprofit, and I suspect the margins are quite slim. They probably make more selling the accessories. They also have tremendous volume and that helps keep the prices down.
It wouldnt surprise me because that’s how it works in the cell phone shops, too. Sales people in many tell me most of their commissions come from cases.
Well, the 35 dollars is before shipping/tax/other things and the 35 dollars is the cheapest model, 1GB ram, others cost more, so maybe the make the cheapest one at more of a loss. Other than that, I am not sure.
I think at least some of the original Raspberry Pi founders were also Broadcom employees, and RasPi boards are all built around Broadcom hardware. I’m guessing there was a lot of consensus-building within Broadcom leading up to the original RasPi’s release… it would be much more difficult to approach an SoC vendor as an external third-party and try to talk them into giving you the same open documentation and design support that RasPi gets from Broadcom.
This is more than powerful enough to be a really good living room retrogaming machine.
I’m currently playing through my favorites with my kids in chronological order (so far their favorites are Final Fantasy, Link to the Past, Mega Man X and Super Mario RPG: Legend of the Seven Stars, and Earthbound) but I’ve built up Paper Mario: The Thousand-Year Door so much that they’re gonna pop if I don’t get an old GameCube or a Dolphin instance going soon…
Since they seem to be positioning this as a credible (lightweight) desktop machine, I wonder if they’ve done anything about the SD card speeds. Improving the graphics, cpu and memory but leaving it stuck at ~20 MB/s on it’s only internal storage device makes it seem… not so compelling to me.
Booting from USB is supported since RPi3, I’m not sure I would bother with SD cards on RPi these days unless physical space is an issue.
Edit: This benchmark article (sorry it’s Medium) shows SD card performance has improved a little but USB 3.0 throughput is about an order of magnitude higher than SD card throughput on the RPi 4: https://medium.com/@ghalfacree/benchmarking-the-raspberry-pi-4-73e5afbcd54b
It’s good to see that SD perf has improved, even if not by all that much. (And the USB gains look great!)
I think my issue is not only physical space, but also an aversion to the idea of having the boot drive able to be accidentally unplugged. But maybe that’s just me…
Is that a constraint of SD or the Pi?
I think it’s mostly a hardware limitation on the Pi, but I’m not an expert. Lots of people report similar boards having much better IO performance (e.g. Odroid), and also that the IO rates on SD cards are much better in an adapter on a laptop/desktop machine.
From http://www.pidramble.com/wiki/benchmarks/microsd-cards:
I guess I could actually run some benchmarks myself when I get home tonight…
USB 3 SSD seems like a reasonable option
Do you think that’s a hardware limitation on the computer side or simply a desire to support every SD card, not just the newer faster ones?
I want to use one of these as a router, but I’m concerned that all the USB3 to Gigabit Ethernet dongles will be garbage. Anyone have any experience with those, good or bad?
Another option would be to use vlans.
I used an Intel NUC with one eth interface as a router for about a year. I just hooked it up to a small managed switch to create a couple of vlans, one for internet and one for lan, that both get pushed to the port for the router. Then you just setup a couple of virtual interfaces in the OS and it work just like a device with two physical interfaces.
The only downside then is that you’re basically making your connection half-duplex. You can still get gigabit in either direction, but not both at the same time.
I was thinking the same thing when I saw USB3 support. My current router had an on-board Ethernet port die on it and I switched it out to a USB3-Ethernet gigabit adapter. It works fine. It’s just got a standard RealTek r8152 chip in it … the USB3-Ethernet chip is typically going to be a standard RealTek, Broadcom or Atheros on most of these devices and they should function the same as their PCI equivalents.
I considered the same thing by my conclusion is that GbE dongles are a mixed bag. Some don’t work, some work but fail intermittently. (Which is really what you don’t want on a router.) There must be some that work fine for a long time but I couldn’t find anyone who already did the research on this and shared it. I ended up buying a PCEngines APU2 for my router/firewall at home. It’s much more capable than a Raspberry Pi and only costs a bit more after you include all the dongles and drives and such that you would need for a Raspberry Pi.
I haven’t tried it, but this seems like an good board for a router: https://www.hardkernel.com/shop/odroid-h2/ And its uses X86 which is a plus or a minus depending on your point of view or usage.
Finally, a Raspberry Pi worthy of running OPNsense!
This is almost fast enough that paired with one of the pi-laptop case/keyboard/screen combos it’d make a great ‘burner’ laptop.
If we’re going to put a press-release/product announcement up, at least tag it
release.Is there any shop in europe that has the 4GB model in stock? I checked all approved resellers in Belgium, Poland, Czech Republich, Germany, Austria and the Netherlands, but without any luck.
I was looking at them on reichelt.de earlier today and the 4gb model is listed as available to ship on the 28th, so just a few days wait (assuming that’s accurate)