> At this price point, it’s easy enough to throw one in your basket when you’re already spending a decent chunk of change and be done with it, safe in the knowledge that you have a well-performing boot device. I do still think it’s a bit naff that the M.2 HAT+ doesn’t fit in the official case but I won’t re-open that can of worms.
An rpi5 plus M.2 hat plus a branded ssd and you are well into the realm of Intel N100 SBC that don’t need a hat, have a good GPU with encoding/decoding and a lot of CPU power. I really don’t understand rpi anymore beyond the Zero, 2040 and CM range for commercial use.
Power consumption is a good contender for reasons to stick with rpis. The rough, quick search numbers I'm seeing for idle power are ~6w for the N100 SCB and ~1 to 2w for the pi. Max performance is like ~15w to ~4w.
We are starting to see SBCs that bake in an RP2040 to drive some GPIO pins (without needing to tie up a physical usb port), which seems like a reasonable compromise... I do like the RPi Zero size/perf/price combo, but for anything that needs more horsepower it's hard to argue with the bigger NUC-a-likes.
Do these give you decent performance DMA-able gpio and spi like you get even on a cheap rpi4? (I'm not claiming they don't/can't, just that it isn't immediately obvious to me that they would.)
For example, spi is an easy hack to bit sample a digital level and stream the results into memory for later processing. You can effortless do this at 20MHz+ plus on an rpi4. When you want to stream multiple digital levels simultaneously (in sync), DMA on gpio is a nice generalisation. (Could perhaps abuse I2S inputs too, which might be present on a more generic x86 board?)
I always wonder about a nice PCIe board supporting good DMAable gpio/spi/i2c/etc on standard x86 machines. I'd probably pay a fair amount of money for a quality one. But it's a product I've never seen on the market. Bodging a USB gpio interfaces with one of the relatively low spec chipsets seem like a pale, inefficient and high-latency imitation.
Actually, given the RP1, maybe the rpi5 has slow/non-DMA gpio too? I haven't been able to check it out yet as these boards still don't have mainlined linux support and I'm not up for buying toys that don't. (Am I imagining it, or did they used to be a lot better at mainlining support for their boards promptly?)
It’s not quite as seamless though, sadly, it’s just the same as having a Pico plugged in via USB but it’s a good step, even if the X4 (assuming that’s what we’re referring to) has some other flaws!
The N100 SBCs have a much higher peak power and heat, but it's also fine to let them throttle down.
Really though, I think most hobbyists would be perfectly fine with a base model 2GB RAM Raspberry Pi and a cheap USB to NVME enclosure. It's not optimal, but it works and you're not spending money to hot-rod something that will never be all that fast anyway.
Wait, which raspberry pi are you talking about? Max performance on a rpi5 is closer to 20 watts technically, and a bit less than that usually because of thermals. But it's not 5 watts, at least nothing I've seen indicates that. The official rpi5 charger provides 5v5a, so 25 watts. And a lower power mode is used when a "standard" 15w charger is used.
Most things I see rpi being used for would be better done at 1/10 of the power with an ESP32, STM32 or arduino.
The things that need an rpi and therefore must use power from the wall anyway, I always find myself thinking would be better done by a minipc, eg one using a laptop cpu like a 7730U.
Costs more, but you get like 3-4x the speed at a similar power use, and have much more flexibility on what OS you can use because it's a "proper" PC.
Don't get me wrong, I like the idea of rPi. I have a Pi Zero that I've used exactly once because I wanted to do something and I came across a project that did exactly what I needed that used rPi.
I just find there's always a better option than rPi for every project I find myself doing.
Is it really 1.5× longer in practice? The N100s are efficient, and you can simply choose a lower power profile by sending a different energy/performance preference to the CPU, and it finishes tasks quicker.
I suspect you are describing a large part of consumer behavior.
Someone gets a sporty-car, then starts working on the engine, the suspension, the exhaust, gets new rims and tire, new seats...
and in the end they could have bought a porsche
Thing is, it all depends on your use case.
For the pi, it is lots better for hardware projects than a nuc style machine, via the gpios/csi/dsi. and specialized hats. There is also a huge community of forums/documentation/people to help you do new things, solve problems, get unstuck.
But if you are doing linux PC kinds of things, like a nas or a media server, you need to do the math.
We've replaced most of our RPi units with the Beelink SER MAX 5 or whatever at $300ish price point; just too good of a deal (and all of its hardware works with Linux, unlike some other MiniPCs).
But we still use RPi4/5s when we want the ecosystem of the RPi, particularly around the GPIO stuff. Typical things include environment monitoring such as lighting and air quality - super simple plug-in devices that easily interface with Raspbian and the GPIO system.
Power draw is much lower on the RPis which I think is relevant for some people; not really us. But that's a pretty valuable use case for those who are shipping battery-backed devices.
Yep rpi5 does not make sense anymore. N100 goes for $110 on sale with 512gb nvme/16gb of ram/psu/case all included. It completely roasts rpi5 while staying with x64 Linux.
I’d argue they were never a particularly good option as a server, especially a file server, because of their poor IO. You’d often get better performance by installing Linux on a NAS. Ironically the Pi5 is the only one that is actually ok at being a file server. Pis are very good at being lots of other things though.
The Pi 4 still makes sense for 35$ if your use case doesn't require power but still requires something more featured than a esp32 or rp2040. Most of my Pis are either video displays or 3d printer controllers, so I'll keep getting Pi 4s for these. Maybe the 3A+, I like the form factor of that one.
I don't mind spending a bit more for the 3A+ to not have to deal with the adapters like microUSB to USB-A host ports for my 3D printers. Or for video displays having Ethernet and full-sized HDMI. I do have a Zero 2W just in case, but for my current use cases I just like the 3A+ more.
Ah I see that makes sense! The adapters are indeed a big annoyance with the zeros. The 3A+ doesn't have ethernet either though (like none of the A models).
While there is some commercial/industrial use of RPi, that seems like an odd consideration when generally considering how they've positioned the brand. Meanwhile you can write and run very similar code across everything from the zero all the way up to the 5—which is enormous flexibility that x86 simply doesn't offer.
Plus, my understanding is that the N100 still tends to have poor driver support issues that the Pi likely doesn't with such a large community.
> Meanwhile you can write and run very similar code across everything from the zero all the way up to the 5—which is enormous flexibility that x86 simply doesn't offer.
Every raspberry pi, from Zero, Zero W, Zero 2 (great name guys), Pi 1, Pi 2, Pi 3, Pi 4, Pi 5 and every variation are all general purpose linux computers that can run pretty much any code you can run on a normal linux computer. They are all the same compilation target IIRC. If you wrote a program for the original Pi, you could have run it on every other Pi without modification or recompilation, depending on peripheral specifics I guess.
it's a known quantity with known good support for various things and a large community.
People and particularly businesses want to buy something they know works, rather than trying to change and figure out new configuration nuances every time something with a slightly better looking spec sheet comes out.
I agree - I've got a couple of N100 micro PCs that have replaced Raspberry Pi devices at home. One of them is running Home Assistant and a bunch of addons, and the other is my Plex server. Both of them have more than enough power and super easy to add more RAM or storage.
One issue I have with a lot of the N100 systems is there seems to be a lot of variants from different manufacturers, slightly different form factors, etc. The PI may be worse but at least you have a pretty good idea what you are actually getting most of the time.
Back in college we used a Pi (not the latest gen because their code was out of date but whatever) and being a local with a car, I volunteered to get them for classmates who couldn't. This was back when you could really only buy them in person or pay a 3-4x markup. Anyway, the prof let us know we'd need a pi, card, maybe a case, and maybe a power supply. A lot of people just wanted everything so they didn't have to worry about it. I think most beginners don't want to deal with this stuff and don't care if it's a marked up Samsung drive or marked up whoever uSD card. And now that I'm in industry, I agree. I wouldn't use a Pi but first party accessories are usually worth it vs the debugging hours to find out you bought garbage.
desire for a low resource consumption web browser.
What we really want is low resource consumption web content, right?
When you pull up the browser's debugger and look at what a "modern" mainstream web page/app needs to deal with... you can see there's little hope for a low power device. Megabytes upon megabytes of obfuscated javascript, from multiple sources, nearly all of it needing decompression and decryption.
Using my Raspberry Pi 4
The Raspberry Pi 5 (8GB) I recently got is actually very close to tolerable for web browsing and running VSCode.
(I realize VSCode is kind of a pig. I'm just kind of experimenting to see what might be viable)
I run Firefox with uBlock, which seems to help somewhat. Total RAM usage with FF and VSCode according to htop is close to 4GB so I suspect that your 1GB Pi is running into swap once you start tooling around the interwebs with a web browser.
Upgrading from an A1 SD card to a A2 SD seems to make something of a palpable difference in "desktop usability." Certainly the benchmark scores for the A2 card blow the A1 away.
I've only had the Pi for a few weeks so I haven't experimented with more aggressive web optimization stuff, like switching my user-agent to request mobile versions or running a Pi-hole, etc. I'm also running at 4K native, so.... changing that certainly might help too....
Long solved, though some of the new stuff tends to be out of stock at launch for a bit at least.
But it's still all produced in the UK and shipping + customs fees easily rack up another 30% so it's at least as expensive as tickets for Taylor Swift.
But what about the heat sinks / fan when you mount the m2 on top? Especially /w Pi 5, to get more perf out of it, it presumably should have a sizeable heat sink complete with the fan?
In what way is thermally throttling at 86.5 C quite enough cooling? "Active Cooler in Case with HAT" idles near the same temperature "Active Cooler" reaches under max, unthrottled, load.
I think you misread the graph. "Active Cooler in Case with HAT" is pretty much the same temperature (74.8) as "Active Cooler" (no case, no hat, 70.6). None of the configurations with the cooler throttled.
Ah shit, you're 100% right. Looking at it on my computer monitor I think I was looking at the "Heatsink in Case with HAT" line color by mistake. My error and apologies!
The official Pi 5 NVME hat doesn’t work with the active cooler and doesn’t fit full sized SSDs. I think the Pimoroni NVMe Base is a better choice since it doesn’t have those issues.
The Pineboards PoE + NVMe HAT will be released next week and will allow you to have both, and some other 3rd party ones do the same. The official Pi M.2 HAT does block the PoE headers though sadly!
eMMC is not a slot. It's a form of built-in storage. It conforms to the old MultiMediaCard interface... which was obsoleted by SD cards a LONG time ago. The eMMC variant is still used to cheaply bung in some usable flash storage into rinky-dink devices like thin clients, cellphones, and such. But for something like the Pi, if you want performance and reliability over SD cards you need a grown-up solution: NVMe or SATA SSD.
The problem of Raspberry Pi today is they need to be relevant as a cheap innovation platform in the AI space.
They would need to do with AI what they did 10 years ago in the IoT and self-hosting space.
I have no idea if their recently AI Camera or Hailo based AI HAT is.
But my guess is they would need to offer something good enough for a fraction of the price of a Nvidia Jetson.
> At this price point, it’s easy enough to throw one in your basket when you’re already spending a decent chunk of change and be done with it, safe in the knowledge that you have a well-performing boot device. I do still think it’s a bit naff that the M.2 HAT+ doesn’t fit in the official case but I won’t re-open that can of worms.
An rpi5 plus M.2 hat plus a branded ssd and you are well into the realm of Intel N100 SBC that don’t need a hat, have a good GPU with encoding/decoding and a lot of CPU power. I really don’t understand rpi anymore beyond the Zero, 2040 and CM range for commercial use.
Power consumption is a good contender for reasons to stick with rpis. The rough, quick search numbers I'm seeing for idle power are ~6w for the N100 SCB and ~1 to 2w for the pi. Max performance is like ~15w to ~4w.
Also, a bunch of GPIO if you want to control some hardware, not just run a Linux desktop or a media server.
We are starting to see SBCs that bake in an RP2040 to drive some GPIO pins (without needing to tie up a physical usb port), which seems like a reasonable compromise... I do like the RPi Zero size/perf/price combo, but for anything that needs more horsepower it's hard to argue with the bigger NUC-a-likes.
Do these give you decent performance DMA-able gpio and spi like you get even on a cheap rpi4? (I'm not claiming they don't/can't, just that it isn't immediately obvious to me that they would.)
For example, spi is an easy hack to bit sample a digital level and stream the results into memory for later processing. You can effortless do this at 20MHz+ plus on an rpi4. When you want to stream multiple digital levels simultaneously (in sync), DMA on gpio is a nice generalisation. (Could perhaps abuse I2S inputs too, which might be present on a more generic x86 board?)
I always wonder about a nice PCIe board supporting good DMAable gpio/spi/i2c/etc on standard x86 machines. I'd probably pay a fair amount of money for a quality one. But it's a product I've never seen on the market. Bodging a USB gpio interfaces with one of the relatively low spec chipsets seem like a pale, inefficient and high-latency imitation.
Actually, given the RP1, maybe the rpi5 has slow/non-DMA gpio too? I haven't been able to check it out yet as these boards still don't have mainlined linux support and I'm not up for buying toys that don't. (Am I imagining it, or did they used to be a lot better at mainlining support for their boards promptly?)
Yes there are N100 SBCs including a RP2040, so best of both worlds.
It’s not quite as seamless though, sadly, it’s just the same as having a Pico plugged in via USB but it’s a good step, even if the X4 (assuming that’s what we’re referring to) has some other flaws!
Especially when you want to work with PIO as well. Well, either way Raspberry Pi gets paid.
> Max performance is like ~15w to ~4w.
Peak power from the Raspberry Pi 5 is 12W, according to Raspberry Pi themselves: https://www.raspberrypi.com/news/introducing-raspberry-pi-5/
The N100 SBCs have a much higher peak power and heat, but it's also fine to let them throttle down.
Really though, I think most hobbyists would be perfectly fine with a base model 2GB RAM Raspberry Pi and a cheap USB to NVME enclosure. It's not optimal, but it works and you're not spending money to hot-rod something that will never be all that fast anyway.
Wait, which raspberry pi are you talking about? Max performance on a rpi5 is closer to 20 watts technically, and a bit less than that usually because of thermals. But it's not 5 watts, at least nothing I've seen indicates that. The official rpi5 charger provides 5v5a, so 25 watts. And a lower power mode is used when a "standard" 15w charger is used.
Depends on where you live then. My home state of Minnesota has electricity for 11 cents per kWh, with only 19% coal.
Very different than, say, Germany. Power efficiency becomes a much lower priority.
How often is that the case? Are you really saving so much more money with a 10W board vs a 15W system to offset the purchasing price?
It's like 1.5x longer run time on battery power :shrug:
Most things I see rpi being used for would be better done at 1/10 of the power with an ESP32, STM32 or arduino.
The things that need an rpi and therefore must use power from the wall anyway, I always find myself thinking would be better done by a minipc, eg one using a laptop cpu like a 7730U.
Costs more, but you get like 3-4x the speed at a similar power use, and have much more flexibility on what OS you can use because it's a "proper" PC.
Don't get me wrong, I like the idea of rPi. I have a Pi Zero that I've used exactly once because I wanted to do something and I came across a project that did exactly what I needed that used rPi.
I just find there's always a better option than rPi for every project I find myself doing.
Is it really 1.5× longer in practice? The N100s are efficient, and you can simply choose a lower power profile by sending a different energy/performance preference to the CPU, and it finishes tasks quicker.
How many PIs did you see in battery powered applications?
It's purely about battery runtime in these cases.
How many battery powered applications did you see use a pi?
I suspect you are describing a large part of consumer behavior.
Someone gets a sporty-car, then starts working on the engine, the suspension, the exhaust, gets new rims and tire, new seats...
and in the end they could have bought a porsche
Thing is, it all depends on your use case.
For the pi, it is lots better for hardware projects than a nuc style machine, via the gpios/csi/dsi. and specialized hats. There is also a huge community of forums/documentation/people to help you do new things, solve problems, get unstuck.
But if you are doing linux PC kinds of things, like a nas or a media server, you need to do the math.
Most hardware projects I tend to look at dont need the power of a Pi5 and are fine with a 2040 or zero.
The niche of the Pi5 just has gotten so much smaller, imo.
We've replaced most of our RPi units with the Beelink SER MAX 5 or whatever at $300ish price point; just too good of a deal (and all of its hardware works with Linux, unlike some other MiniPCs).
But we still use RPi4/5s when we want the ecosystem of the RPi, particularly around the GPIO stuff. Typical things include environment monitoring such as lighting and air quality - super simple plug-in devices that easily interface with Raspbian and the GPIO system.
Power draw is much lower on the RPis which I think is relevant for some people; not really us. But that's a pretty valuable use case for those who are shipping battery-backed devices.
Yep rpi5 does not make sense anymore. N100 goes for $110 on sale with 512gb nvme/16gb of ram/psu/case all included. It completely roasts rpi5 while staying with x64 Linux.
Really? Where?
Currently I’m seeing $120 for n100/16gb/512 on both eBay/aliexpress but I’ve seen it get cheaper.
Totally true. If you want a small server just get a N100.
The pi makes sense for something where you need power (eg more than esp32 or 2040) and gpio. Not as a small server. Those days are gone.
I’d argue they were never a particularly good option as a server, especially a file server, because of their poor IO. You’d often get better performance by installing Linux on a NAS. Ironically the Pi5 is the only one that is actually ok at being a file server. Pis are very good at being lots of other things though.
The Pi 4 still makes sense for 35$ if your use case doesn't require power but still requires something more featured than a esp32 or rp2040. Most of my Pis are either video displays or 3d printer controllers, so I'll keep getting Pi 4s for these. Maybe the 3A+, I like the form factor of that one.
Oh yeah if you like the 3A+ you might even be able to get away with the Zero 2W, it has the same chip. And it's even cheaper.
I don't mind spending a bit more for the 3A+ to not have to deal with the adapters like microUSB to USB-A host ports for my 3D printers. Or for video displays having Ethernet and full-sized HDMI. I do have a Zero 2W just in case, but for my current use cases I just like the 3A+ more.
Ah I see that makes sense! The adapters are indeed a big annoyance with the zeros. The 3A+ doesn't have ethernet either though (like none of the A models).
While there is some commercial/industrial use of RPi, that seems like an odd consideration when generally considering how they've positioned the brand. Meanwhile you can write and run very similar code across everything from the zero all the way up to the 5—which is enormous flexibility that x86 simply doesn't offer.
Plus, my understanding is that the N100 still tends to have poor driver support issues that the Pi likely doesn't with such a large community.
>While there is some commercial/industrial use of RPi
I thought the whole reason for the shortages the last few years was due to commercial applications buying out most of the Pi's available?
> I thought the whole reason for the shortages the last few years was due to commercial applications buying out most of the Pi's available?
Sure, but that's not why people find raspberry pis interesting or valuable.
> Meanwhile you can write and run very similar code across everything from the zero all the way up to the 5—which is enormous flexibility that x86 simply doesn't offer.
This doesn't make sense, can you elaborate?
Every raspberry pi, from Zero, Zero W, Zero 2 (great name guys), Pi 1, Pi 2, Pi 3, Pi 4, Pi 5 and every variation are all general purpose linux computers that can run pretty much any code you can run on a normal linux computer. They are all the same compilation target IIRC. If you wrote a program for the original Pi, you could have run it on every other Pi without modification or recompilation, depending on peripheral specifics I guess.
Are you getting confused by the 2040?
I am not at all confused by a microcontroller. I am confused because what you said about the pi can be said about x86.
it's a known quantity with known good support for various things and a large community.
People and particularly businesses want to buy something they know works, rather than trying to change and figure out new configuration nuances every time something with a slightly better looking spec sheet comes out.
I agree - I've got a couple of N100 micro PCs that have replaced Raspberry Pi devices at home. One of them is running Home Assistant and a bunch of addons, and the other is my Plex server. Both of them have more than enough power and super easy to add more RAM or storage.
One issue I have with a lot of the N100 systems is there seems to be a lot of variants from different manufacturers, slightly different form factors, etc. The PI may be worse but at least you have a pretty good idea what you are actually getting most of the time.
One example: https://www.omgubuntu.co.uk/2024/07/radxa-x4-cheap-intel-n10...
I started down this thought path and ended up with a pair of reconditioned DL369 G9s.
> it’s a bit naff that the M.2 HAT+ doesn’t fit in the official case
Pineboard have a hat that fits in the official case. Only £9.
https://thepihut.com/products/hatdrive-nano-for-raspberry-pi...
Did you mean this[1]?
[1]: https://thepihut.com/products/hatdrive-nano-for-raspberry-pi...
Just to clarify, the M.2 HAT+ does fit in the official case, you just can't put the lid on.
Doesn't it also block any fan you've got? Probably a good thing you can't close it.
Back in college we used a Pi (not the latest gen because their code was out of date but whatever) and being a local with a car, I volunteered to get them for classmates who couldn't. This was back when you could really only buy them in person or pay a 3-4x markup. Anyway, the prof let us know we'd need a pi, card, maybe a case, and maybe a power supply. A lot of people just wanted everything so they didn't have to worry about it. I think most beginners don't want to deal with this stuff and don't care if it's a marked up Samsung drive or marked up whoever uSD card. And now that I'm in industry, I agree. I wouldn't use a Pi but first party accessories are usually worth it vs the debugging hours to find out you bought garbage.
Using my Raspberry Pi 4, I find that interestingly enough, even just using the SD card slot on 1 GB of RAM, device usability is perfectly good.
As soon as I need to use a web browser, performance goes out the window, and you're nearly stranded in terms of usability.
I think that easily qualifies as "desktop usage," and it leaves this desire for a low resource consumption web browser.
When you pull up the browser's debugger and look at what a "modern" mainstream web page/app needs to deal with... you can see there's little hope for a low power device. Megabytes upon megabytes of obfuscated javascript, from multiple sources, nearly all of it needing decompression and decryption.
The Raspberry Pi 5 (8GB) I recently got is actually very close to tolerable for web browsing and running VSCode.(I realize VSCode is kind of a pig. I'm just kind of experimenting to see what might be viable)
I run Firefox with uBlock, which seems to help somewhat. Total RAM usage with FF and VSCode according to htop is close to 4GB so I suspect that your 1GB Pi is running into swap once you start tooling around the interwebs with a web browser.
Upgrading from an A1 SD card to a A2 SD seems to make something of a palpable difference in "desktop usability." Certainly the benchmark scores for the A2 card blow the A1 away.
I've only had the Pi for a few weeks so I haven't experimented with more aggressive web optimization stuff, like switching my user-agent to request mobile versions or running a Pi-hole, etc. I'm also running at 4K native, so.... changing that certainly might help too....
This is why I install netsurf on all my Pis. "But bitwize," you say, "netsurf doesn't support JavaScript!" To which I reply, "Oh no! Anyway..."
Web browsers are merely platforms to run web apps in. Does the same happen if you visit a couple of "plain HTML" web sites?
Is supply chain still an issue or is it finally possible to buy a pi without making it like buying tickets to Taylor Swift?
Long solved, though some of the new stuff tends to be out of stock at launch for a bit at least.
But it's still all produced in the UK and shipping + customs fees easily rack up another 30% so it's at least as expensive as tickets for Taylor Swift.
Looks pretty good per: https://rpilocator.com/?cat=PI5
Extremely available. E.g. pishop.us alone has 947 in stock to buy instantly.
But what about the heat sinks / fan when you mount the m2 on top? Especially /w Pi 5, to get more perf out of it, it presumably should have a sizeable heat sink complete with the fan?
The Active Cooler part (that contains a heatsink too) fits well under the HAT.
The fan won't do much if there a second PCB stacked right on top of it though, unless there's cutouts for it.
It does quite enough. Some test results are here: https://bret.dk/official-raspberry-pi-m-2-hat-review/#Temper...
And for any moderate use cases the need for any cooler at all could be disputed. I bet most RPi use cases don't include heavy computing.
In what way is thermally throttling at 86.5 C quite enough cooling? "Active Cooler in Case with HAT" idles near the same temperature "Active Cooler" reaches under max, unthrottled, load.
I think you misread the graph. "Active Cooler in Case with HAT" is pretty much the same temperature (74.8) as "Active Cooler" (no case, no hat, 70.6). None of the configurations with the cooler throttled.
Ah shit, you're 100% right. Looking at it on my computer monitor I think I was looking at the "Heatsink in Case with HAT" line color by mistake. My error and apologies!
A covered fan is not very clever and the rp5 need active cooling.
The official Pi 5 NVME hat doesn’t work with the active cooler and doesn’t fit full sized SSDs. I think the Pimoroni NVMe Base is a better choice since it doesn’t have those issues.
Does anyone know if these will work with the PoE HATs?
The official NVMe hat blocks access to the PoE header, but there are third party options that combine NVMe and PoE into one hat:
https://www.youtube.com/watch?v=x9ceI0_r_Kg
Alternatively you could get one of the NVMe boards that sits under the Pi, and put a standard PoE hat on top.
As an example of the latter, Pimoroni offer an NVMe “base” that would avoid this problem and leave the PoE header free: https://shop.pimoroni.com/products/nvme-base?variant=4121958...
The Pineboards PoE + NVMe HAT will be released next week and will allow you to have both, and some other 3rd party ones do the same. The official Pi M.2 HAT does block the PoE headers though sadly!
"Are we seeing the beginning of the Cambridge Cartel? Where they corner the market of each of the accessories needed for your Pi?!"
What is this hysteria?
It’s a joke about how many new products they’re releasing releasing recently
It would be cool to have a nex gen RPi (perhaps even the Zero edition) that does away with the micro SD slot and replaces it with a built in NMVe one.
In addition, replace most of the other ports (microUSB, etc...) with a number of USB-C ports.
Well, not an NVMe maybe (yet), but at least a built-in eMMC slot.
eMMC is not a slot. It's a form of built-in storage. It conforms to the old MultiMediaCard interface... which was obsoleted by SD cards a LONG time ago. The eMMC variant is still used to cheaply bung in some usable flash storage into rinky-dink devices like thin clients, cellphones, and such. But for something like the Pi, if you want performance and reliability over SD cards you need a grown-up solution: NVMe or SATA SSD.
OK, "socket". (ODROID-N2+, for example, uses eMMC as a faster, better alternative to an SD card.)
https://www.hardkernel.com/shop/odroid-n2-with-4gbyte-ram-2/
Why not. Looking at the prices 128GB NVME is $15 on amazon. That's also the price of various micro SD cards.
The problem of Raspberry Pi today is they need to be relevant as a cheap innovation platform in the AI space.
They would need to do with AI what they did 10 years ago in the IoT and self-hosting space.
I have no idea if their recently AI Camera or Hailo based AI HAT is. But my guess is they would need to offer something good enough for a fraction of the price of a Nvidia Jetson.