I really wish they would stop making newer, faster, more expensive Pi's and just focus on making the existing Pi's cheaper and more available. Out of the dozen or so RPi's I have, not a single one is even connected to a monitor. I don't need dual 4k HDMI ports. I'd love a widely available and in-stock $20 Pi. It's like they've forgotten what the RPi is all about.
Their mass production of the Pi keyboard should clue you into at least one direction that they believe "the Pi is all about"; education, and increasing access to modern computing hardware. Enhancing the performance envelope and connecting to a monitor do, to me, seem like worthy steps toward that end.
Everyone uses general purpose computers differently. I feel your statement "they've forgotten what the RPi is all about" isn't just ignorant; its hurtful. Maybe their direction isn't parallel with what you want out of the products they make, but you should at least have the empathy to recognize that you aren't the main character in this play.
How are two niche ports requiring adapters more accessible than one full-size HDMI port? Are two ports on such constrained hardware really worth it? If my adapter cable breaks (which already happened) I have to order a new one and can't use the device while sitting on a pile of unused but unsupported HDMI cables. Meanwhile iirc it took until after the launch of the Pi 4 to get USB boot support without first flashing and booting a microSD card, another piece of hardware not known for its longevity.
The RPi project is amazing and I'm glad it exists, but some decisions are just weird considering their goals. Though to be fair the RPi 4 & 5 improved the situation significantly.
Famously the devices have been difficult to acquire because RPi was putting all their manufacturing output into fulfilling commercial market needs (1,2,3). Filling a market need -- well-specced hardware with long support horizons -- is perfectly fine, but Upton et al. have made it very clear that this is the priority and pretending otherwise is weird. Mass producing generic USB keyboards in corporate livery does not meaningfully communicate corporate ethics here.
Meanwhile, most of the competition focuses on beating the Pi on specs, availability, or price -- and none of them seem to have noticed that long support life as being the differentiator which leads to business adoption of the Pi hardware. So I suspect the status quo will persist for the foreseeable future.
Fortunately we don't need belief, since the foundation has years of volume-based sales data on market segments, e.g. for the last two years they have prioritized industrial business users over education, during supply chain shortages.
Is there a published market segment breakdown for each generation of RPi?
Their existing contracts and focus should clue you more into their desired direction than an accessory that they can produce for consumers that is tangentially related to raspberry pis.
The pi foundations focus(be it intentional or just out of a desire to stay open) has been to have prototypes created with raspberry pis that go into production without replacing the raspberry Pi so that device manufacturers are locked into large long-term contracts.
The inability for consumers to find devices available and the odd io choices(IoT video devices like kiosks would benefit from the double ports rather than a single full size HDMI) are IMO the intention of the pi foundation. Their focus is on generating contracts and fulfilling them, consumer devices are secondary.
That's to say they've "forgotten what the pi is about" insofar as they've made choices to grow the business/contract side of the pi foundation more than the education and consumer availability
While I admit that I am a bit envious of the large 4k monitors that my coworkers use, I have managed to make it through life without buying anything larger than a 1440p monitor.
What exactly is the educational advantage (for the demographic that the raspberry pi is supposed to be educating) of the inclusion of 2x 4k ports and does that advantage override the opportunity cost of including the ports?
In other words '1080p ought to be enough for anybody learning how computers work from an SBC.
You can get $60 android tablets brand new on Amazon that don't need a monitor, keyboard and mouse and can use the vast amount of education apps in the play store.
Not sure why you would need dual monitor outputs for education use. I seem to remember at the time of release them talking about how it would actually be useful in telephone call centres.
If they want to increase access to computing hardware, their 'industrial customers first, everyone else can pick over whatever odds and end are left over' sales policy is a really strange way of showing it.
Parent should know that the hurtfulness of having a different opinion about the best use of a computer cannot be overstated. Imagine if my great grandfather who survived the holocaust read this comment--it would kill him even worse than hitler tried to.
I wish they took performance more seriously. I migrated a bunch of my stuff off of a Pi 4 to an Intel N100 NUC because the Pi 4 just didn't cut it. I have a bunch of other use cases as well that have struggled with the Pi 4s and competing RK3588 based systems look compelling.
Point is, your use case doesn't match my use case. I hope they continue to make things more performant AND cost effective.
I haven't bought any in a while, but looks like the Pi Zero W is in stock and available for $10 at Microcenter. Other sellers seem to have them in stock for $15. So isn't there already a widely available Pi in stock for under $20?
I have found that a good chunk of random opensource projects won't run on the zero 2. The ARM core is ARM7 and doesn't support modern instructions, so plenty of software just immediately dies with 'Illefal Instruction'.
Also, it's 32 bit, and a lot of software is starting to depend on the massive address spaces offered by 64 bit.
And there is no high ram model. Some stuff just gobbles so much ram that 512M isn't practical.
Totally fair desire, though some people definitely do want these to be desktop or HTPC replacements.
I'm usually more impressed with RockChip stuff anyway, though, so if the Raspberry Pi folks wanted to focus their efforts on affordable dev boards, I think there'd still be plenty of options for those who do want to play back 4k video on a monitor.
They've done an excellent job with the Pi Pico, keeping it in stock and dirt cheap. I've gotten a dozen or more of those at like $4 each. So they're still catering to the "maker" types with those.
I'm excited to get a better version of something like the Pinebook Pro someday, based on a newer SoC with more RAM and a bit more GPU power for video playback and some simpler games like Minetest and Xonotic.
I agree, they could make two models that cater both the markets they are gong for which are "mini servers" for lans and iot where no one needs to look at the monitor and you access it with ssh/vnc, and "mini desktops" for education where an active user sits in front of a monitor.
So:
- Pi Server: without all the stuff that not belong in it -- for a cheap price like you said
- Pi Desktop: with all the stuff that belongs there -- for a premium price.
Of course this means double design, double production line, double testing,. support, etc. And will most likely reduce revenue. This is probably why they go with the all in one approach.
Different people have different needs though. I have a smart mirror that runs a 4K display driven by a Pi 4. I'm sure there a lot of other kiosk-like use-cases for Pis.
I want less powerful pis with more hardware features.
Everyone else is getting onboard battery management and AI coprocessors and stuff, the Pi is busy trying to run on expensive SSDs because for some reason they don't just fix their software to run better on SDs(And lots of apps seem uninterested in fixing SD write issues).
There are other flavors of Pi, like the Orange Pi or Banana Pi. Have you looked into those? The newest Orange Pi is really expensive but maybe there's others that are cheaper
Yes, for context the Pi 1-4 all had H264 hardware encoder/decoder support, which could comfortable encode at least 720p @ 30Hz in realtime. The die space argument makes sense for why they don't have AV1/HEVC encoding, but it does not explain why H264 was dropped. The fact that the CPU is now powerful enough to encode H264 at better quality and frame rates than the old hardware encoder is a better argument, but still a step backwards for folks who need lower power consumption or need the CPU for other things, and doesn't explain why the hardware support needed to be dropped. It really does sound like something else (like licensing) drove the decision, and this is a post-facto attempt to sell/justify that decision.
Encode is always a bit messier; perhaps you need a tool which is not GPU aware, maybe your needs go beyond what the encoder can do.
But dropping the hardware h.264 decoder is a horrible thing to do. H.264 might as well be the lingua franca of online videos. Think of all the kids in classrooms loading videos on Youtube now constantly hammering the CPU for decode. It's such a weird decision that it can only be caused by business issues.
It is a bit frustrating how married the Raspberry Pi foundation is to a company that couldn't care less if they live or die. Broadcom has always been at best indifferent and at worst hostile to the open source community.
Broadcom is one of the few options out there for ARM SoC. Rpi could dump a bunch of money into making their own SoC, but that would really ballon the costs.
Are you saying the hardware actually has encode on die but was "fused" off?
If so, Raspberry Pi has an organizational culture of outright lying, as opposed to simply speaking plainly. I know this is not Eben Upton speaking, but I've found him many times in the past being either evasive with the truth, or simply lying. It's one thing to not step of the toes of partners, or to not Osborne a product by suggesting a successor is nearing, but to be outright dishonest. Yuck.
> Are you saying the hardware actually has encode on die but was "fused" off?
I don't think that is what they are saying, and I haven't seen any evidence of that. It makes the most sense to me that either (as GP said) they got rid of the H264 hardware to save on paying Broadcom for the rights to include that hardware block in the chip, or to save on the die area they take up (as RPi said), or both.
"In future we’ll have to do something, but for Pi 5 we feel the hardware encode is a mm^2 too far."
Sounds reasonable, given a fast cpu & less-than optimal hw-accelerated encoding options. As for that "something", maybe:
1) Drop hw-accelerated encoding and decoding entirely, and use the freed up silicon for much beefier cpus (like ones including -bigger- vector units, more cores etc. Cortex X?). That would be useful for any cpu heavy applications.
2) Include hw encoder for a common (1), relatively 'heavy' codec. And hw decoder for same + maybe others.
3) Only include decoder(s?), like they seem to have done for RPi5.
4) Include some kind of flexible compute fabric that can be configured to do the heavy lifting for popular video codecs.
Combined with:
5) Move to newer silicon node to obtain higher efficiency or transistor budget.
Whatever route a future RPi would go, imho hw-accelerated decoding is much more useful than encoding.
That’s somewhat understandable, but still a bit disappointing. Not all playback devices support HEVC, and additionally HEVC will likely remain patent-encumbered much longer than e.g. H.264.
I really wish there was at least one other and open hardware-accelerated format. It’s probably a bit too early for AV1, but VP9 would work with modern iOS and Android devices, for example.
I also wonder how much licensing costs were a concern here, although past RPis had software-unlockable codecs for that exact reason.
The rk3588 and rk3588s (i.e. orange pi 5) both support hardware H264/HEVC encoding @ 8k30fps, as well as hardware HEVC/VP9 decoding @ 8k60fps, H264 decoding @ 8k30fps, AV1 decoding @ 4k60fps.
Surprisingly, AV1 is easier to implement a decoder for than VP9 (by design). Some of the VP9 transforms were axed with AV1 purely to make the stream easier to decode.
Why AV1 hardware decoding has taken so long seems to be an issue with hardware manufactures not wanting to support it. HEVC and VVC seem to have more hardware support.
HEVC was built from the ground up to be implemented in hardware to stem the tide of the open-source codecs and thus keep the license money train running.
Ah, yes, I do agree on the quality – it certainly can't compete in any way with high-quality software or dedicated encoders.
But arguably it doesn't have to; many real-time applications (e.g. surveillance cameras) have local bandwidth to spare and/or don't care too highly about quality, and compatibility with older viewing devices without re-encoding is a priority.
Last time I tried to encode a video from H.264 to HEVC (hw-accelerated) on Linux it was such a pain to get to work that I eventually gave up and simply accepted the performance hit. I'm sorry, but I'm not gonna recompile ffmpeg so that it works on my machine. Considering that most RPi-users probably use a Linux-based OS this is IMHO a sensible decision.
There are a ton of pi derivatives that offer a pretty broad range of configurations. I always think of the Pi as the general consumer flagship. I was still pretty impressed with the Pi 4B... I just wish it had broader availability.
Yeah, there's a huge variety of SBCs that exist, most of which have a better specs/price ratio than RPi. If you buy a RPi, you're spending money on the hardware, you're spending it on the mountains of support/tutorials/standardization. I suspect that most people on HN can handle the reduced knowledge base that exists for BPi, OPi, ROCK, etc.
Jetson Orin Nano, the new devkit from Nvidia, surprisingly does not have a hardware video encoder either, you'll need use CPU to encode video instead, which is extremely odd considering video-encoding is a common use case on lots of video applications.
Wow, I was shocked to see that. They don't even bother using CUDA cores to encode (less of an issue if you have unified memory like in Jetsons). Anyway it seems you can get four 1080p streams at 30fps encoded to h264 in CPU if you set it up right: https://www.ridgerun.com/post/jetson-orin-nano-how-to-achiev...
Only removed from the tegras, not the gaming cards.
But yeah if you were doing OBS streaming on your jetson, then the newer hardware won't have NVENC. Normally I'd say "but nobody is doing that" but someone (was) indeed doing exactly that, lol.
As mentioned it's probably a ton more relevant to embedded and robotics/computer-vision use-cases, which NVIDIA is really trying to push for (automotive etc). So it's still a surprising change and will be a pain for some people.
h264 is the most available but probably not the most used. Every streaming service today either seeds AV1, VP9, or HEVC content first since it saves bandwidth[0] and every client from the past 5 years supports one of these newer formats (phones, GPUs, smart TVs, streaming boxes, etc).
H.264 is effectively the MP3 of video today: It provides neither the best quality, nor is it patent free (yet), but if an SoC supports any hardware accleration at all, H.264 is usually on the list.
Applications/services that can afford the overhead of multiple format encoding will do so, but it's not always an option.
Most devices can indeed most likely handle software decode of more common resolutions, codecs and bitrates. But I'd really hope they'd pick the one that won't suck up all the battery, so H264. This line of thought is supported by the fact that YouTube still provides an H264 option with most if not all videos.
With higher bitrate things, HEVC seems to grow in popularity but even software decode support is not everywhere. Netflix for example requires the installation of HEVC support on Windows to play 4K content.
Actually hardware-accelerated video decode is even spottier and more unreliable across most platforms. The JS API for codec support (canPlayType) literally returns "maybe" and "probably". It's quite bad.
So far the best compatibility I've seen has been Edge with flags on Windows (MPEG-2, H264, HEVC, AV1, VP8, VP9 with most also supporting accelerated encode). It still fails with some content (Dolby Vision P5 colors are incorrect, HEVC Rext doesn't play - more info about HEVC is available here https://github.com/StaZhu/enable-chromium-hevc-hardware-deco...). Chrome on macOS is a close second in terms of codec support.
The worst in terms of HW acceleration being all the browsers on desktop Linux-s, few and fragile combinations that offer limited and janky support. But it's slowly improving. This combined with the not-the-latest hardware many use, means things like VP9 or AV1 tend to stutter.
I'd love to see some more generic stats, but considering the APIs aren't sufficient to determine actual support, these might be difficult to gather.
But all things considered, I heavily doubt dropping H264 HW decode support was a good idea considering how often its still used.
True, but hardware support does wonders for power consumption. I'm blown away that I can watch 3 hours of h264 content on my flights and only 5% of my iPad battery is used. i.e. more factors to consider than simply whether it can/can't with brute force.
Remember you can get an HP Prodesk G3 400 or some such for $60 refurbed with far more hardware capabilities if you don't need the portable form factor, gpio, or power consumption.
Right, I agree, but electricity prices are variable and location dependent, for example, it doesn't cost me 30c/kwh, it's more like 5 cents and 3 cents depending on the time.
With that logic, the $20 in difference of $30 per year vs $10 per year of Pi is fairly meaningless for performance boost I am getting.
If your electricity is expensive, sure, but I imagine the performance boost would be worth it for an actual homeserver that a G3 can make vs a tinkering/fun project you can make with a Pi.
You're comparing a CPU that became generally available in 2017 or earlier (i.e., the 7th-gen or 6th-gen Core CPU in the Prodesk G3) with a more much recent CPU.
Readit News
Everyone uses general purpose computers differently. I feel your statement "they've forgotten what the RPi is all about" isn't just ignorant; its hurtful. Maybe their direction isn't parallel with what you want out of the products they make, but you should at least have the empathy to recognize that you aren't the main character in this play.
The RPi project is amazing and I'm glad it exists, but some decisions are just weird considering their goals. Though to be fair the RPi 4 & 5 improved the situation significantly.
Meanwhile, most of the competition focuses on beating the Pi on specs, availability, or price -- and none of them seem to have noticed that long support life as being the differentiator which leads to business adoption of the Pi hardware. So I suspect the status quo will persist for the foreseeable future.
1 - https://www.jeffgeerling.com/blog/2022/you-cant-buy-raspberr... 2 - https://www.elecrow.com/blog/why-is-it-so-hard-to-buy-raspbe... 3 - https://www.tomshardware.com/news/raspberry-pi-ceo-eben-upto...
Fortunately we don't need belief, since the foundation has years of volume-based sales data on market segments, e.g. for the last two years they have prioritized industrial business users over education, during supply chain shortages.
Is there a published market segment breakdown for each generation of RPi?
The pi foundations focus(be it intentional or just out of a desire to stay open) has been to have prototypes created with raspberry pis that go into production without replacing the raspberry Pi so that device manufacturers are locked into large long-term contracts.
The inability for consumers to find devices available and the odd io choices(IoT video devices like kiosks would benefit from the double ports rather than a single full size HDMI) are IMO the intention of the pi foundation. Their focus is on generating contracts and fulfilling them, consumer devices are secondary.
That's to say they've "forgotten what the pi is about" insofar as they've made choices to grow the business/contract side of the pi foundation more than the education and consumer availability
What exactly is the educational advantage (for the demographic that the raspberry pi is supposed to be educating) of the inclusion of 2x 4k ports and does that advantage override the opportunity cost of including the ports?
In other words '1080p ought to be enough for anybody learning how computers work from an SBC.
Their little 'breadbox' pi4* certainly points in that direction very heavily.
Their cheaper 'embedded' version (the zero series, and the pico) is doing what this guy wants.
* https://vilros.com/products/raspberry-pi-400-vilros?variant=...
Deleted Comment
If they want to increase access to computing hardware, their 'industrial customers first, everyone else can pick over whatever odds and end are left over' sales policy is a really strange way of showing it.
Deleted Comment
https://www.microcenter.com/product/486575/Zero_W
Wow...
Point is, your use case doesn't match my use case. I hope they continue to make things more performant AND cost effective.
The pi was never meant to be a server.
Also, it's 32 bit, and a lot of software is starting to depend on the massive address spaces offered by 64 bit.
And there is no high ram model. Some stuff just gobbles so much ram that 512M isn't practical.
https://thepihut.com/products/raspberry-pi-zero-2
I'm usually more impressed with RockChip stuff anyway, though, so if the Raspberry Pi folks wanted to focus their efforts on affordable dev boards, I think there'd still be plenty of options for those who do want to play back 4k video on a monitor. They've done an excellent job with the Pi Pico, keeping it in stock and dirt cheap. I've gotten a dozen or more of those at like $4 each. So they're still catering to the "maker" types with those.
I'm excited to get a better version of something like the Pinebook Pro someday, based on a newer SoC with more RAM and a bit more GPU power for video playback and some simpler games like Minetest and Xonotic.
So:
- Pi Server: without all the stuff that not belong in it -- for a cheap price like you said
- Pi Desktop: with all the stuff that belongs there -- for a premium price.
Of course this means double design, double production line, double testing,. support, etc. And will most likely reduce revenue. This is probably why they go with the all in one approach.
Everyone else is getting onboard battery management and AI coprocessors and stuff, the Pi is busy trying to run on expensive SSDs because for some reason they don't just fix their software to run better on SDs(And lots of apps seem uninterested in fixing SD write issues).
not $35, but not what seemed to be $185 either
Dead Comment
is corporate speech for "broadcom decided not to let us use their video IP cores for low cost/no cost any longer"
But dropping the hardware h.264 decoder is a horrible thing to do. H.264 might as well be the lingua franca of online videos. Think of all the kids in classrooms loading videos on Youtube now constantly hammering the CPU for decode. It's such a weird decision that it can only be caused by business issues.
Broadcom is one of the few options out there for ARM SoC. Rpi could dump a bunch of money into making their own SoC, but that would really ballon the costs.
Deleted Comment
If so, Raspberry Pi has an organizational culture of outright lying, as opposed to simply speaking plainly. I know this is not Eben Upton speaking, but I've found him many times in the past being either evasive with the truth, or simply lying. It's one thing to not step of the toes of partners, or to not Osborne a product by suggesting a successor is nearing, but to be outright dishonest. Yuck.
This is very, very disappointing.
I don't think that is what they are saying, and I haven't seen any evidence of that. It makes the most sense to me that either (as GP said) they got rid of the H264 hardware to save on paying Broadcom for the rights to include that hardware block in the chip, or to save on the die area they take up (as RPi said), or both.
"In future we’ll have to do something, but for Pi 5 we feel the hardware encode is a mm^2 too far."
Sounds reasonable, given a fast cpu & less-than optimal hw-accelerated encoding options. As for that "something", maybe:
1) Drop hw-accelerated encoding and decoding entirely, and use the freed up silicon for much beefier cpus (like ones including -bigger- vector units, more cores etc. Cortex X?). That would be useful for any cpu heavy applications.
2) Include hw encoder for a common (1), relatively 'heavy' codec. And hw decoder for same + maybe others.
3) Only include decoder(s?), like they seem to have done for RPi5.
4) Include some kind of flexible compute fabric that can be configured to do the heavy lifting for popular video codecs.
Combined with:
5) Move to newer silicon node to obtain higher efficiency or transistor budget.
Whatever route a future RPi would go, imho hw-accelerated decoding is much more useful than encoding.
I really wish there was at least one other and open hardware-accelerated format. It’s probably a bit too early for AV1, but VP9 would work with modern iOS and Android devices, for example.
I also wonder how much licensing costs were a concern here, although past RPis had software-unlockable codecs for that exact reason.
Dead Comment
Why AV1 hardware decoding has taken so long seems to be an issue with hardware manufactures not wanting to support it. HEVC and VVC seem to have more hardware support.
> HEVC and VVC seem to have more hardware support.
Yes to HEVC, but which common SoC already supports VVC? AV1 hardware support seems to be much more widespread.
But arguably it doesn't have to; many real-time applications (e.g. surveillance cameras) have local bandwidth to spare and/or don't care too highly about quality, and compatibility with older viewing devices without re-encoding is a priority.
Can't find the datasheet for the BCM2712 right now.
Deleted Comment
Just a handful of examples:
Banana PI M5: https://www.banana-pi.org/
Odroid C4: https://wiki.odroid.com/start
Odroid N2+: Odroid C4: https://wiki.odroid.com/start
Libre "Le Potato": https://libre.computer/
Libre "Renegade": https://libre.computer/
Orange Pi 3 LTS: http://www.orangepi.org/
Orange Pi 5: http://www.orangepi.org/
Rock Pi 4C+: https://rockpi.org/
Nano Pi M4B: http://nanopi.io/
What will OBS do then?
https://www.nvidia.com/en-us/geforce/guides/broadcasting-gui...
But yeah if you were doing OBS streaming on your jetson, then the newer hardware won't have NVENC. Normally I'd say "but nobody is doing that" but someone (was) indeed doing exactly that, lol.
https://blog.fosketts.net/2020/09/10/introducing-rabbit-i-bo...
As mentioned it's probably a ton more relevant to embedded and robotics/computer-vision use-cases, which NVIDIA is really trying to push for (automotive etc). So it's still a surprising change and will be a pain for some people.
0: https://www.etcentric.org/netflix-switching-from-vp9-codec-t...
Applications/services that can afford the overhead of multiple format encoding will do so, but it's not always an option.
With higher bitrate things, HEVC seems to grow in popularity but even software decode support is not everywhere. Netflix for example requires the installation of HEVC support on Windows to play 4K content.
Actually hardware-accelerated video decode is even spottier and more unreliable across most platforms. The JS API for codec support (canPlayType) literally returns "maybe" and "probably". It's quite bad.
So far the best compatibility I've seen has been Edge with flags on Windows (MPEG-2, H264, HEVC, AV1, VP8, VP9 with most also supporting accelerated encode). It still fails with some content (Dolby Vision P5 colors are incorrect, HEVC Rext doesn't play - more info about HEVC is available here https://github.com/StaZhu/enable-chromium-hevc-hardware-deco...). Chrome on macOS is a close second in terms of codec support.
The worst in terms of HW acceleration being all the browsers on desktop Linux-s, few and fragile combinations that offer limited and janky support. But it's slowly improving. This combined with the not-the-latest hardware many use, means things like VP9 or AV1 tend to stutter.
I'd love to see some more generic stats, but considering the APIs aren't sufficient to determine actual support, these might be difficult to gather.
But all things considered, I heavily doubt dropping H264 HW decode support was a good idea considering how often its still used.
Dead Comment
It's actually a better deal for home servers.
Prodesk G3: 24 hours * 365 days * 35W * 1000W/kW * $0.30/kWh = $91
Raspberry Pi 5: 24 hours * 365 days * 12W * 1000W/kW * $0.30/kWh = $31
It's likely that 12W over-counts for the Raspberry Pi 5 and 35W under-counts for the Intel chip, so it might be even worse than this.
With that logic, the $20 in difference of $30 per year vs $10 per year of Pi is fairly meaningless for performance boost I am getting.
If your electricity is expensive, sure, but I imagine the performance boost would be worth it for an actual homeserver that a G3 can make vs a tinkering/fun project you can make with a Pi.
yes its double the price, but in terms of power usage its ~4 watts at the plug at 50% load, (vs 30) and more over a million times faster.