5 ms·

Doesn't matter anyway. They shot themselves in the foot by removing hardware video encoding. I switched to the Orange Pi 5 so I can keep my video encoders and a

Doesn't matter anyway. They shot themselves in the foot by removing hardware video encoding. I switched to the Orange Pi 5 so I can keep my video encoders and also benefit from the nice TPU.

gjsman-1000 3y ago
This doesn't really make any sense - you can video encode 1080p H.264 on the Pi 5 in software without using more than ~25% of the CPU. And the quality, if you don't mind trading more CPU, will be far superior than the hardware encoders ever were.
- phkahler 3y ago
  I remember seeing an ad many decades ago for a PC ad-in board for encoding and decoding JPEG. You know, back when dinosaurs roamed the earth and that might have been a thing for desktop publishing or whatever. General purpose CPU has gotten fast enough to do a lot of things that used to benefit from hardware acceleration.
  - mhandley 3y ago
    A little over 30 years ago when I started working on multimedia conferencing over the Internet, for some reason we had a couple of hardware JPEG decoder ISA (?) cards from Bitfield Oy, a Finish company (if I recall correctly). I can't find any documentation on them now, as this predates NCSA Mosaic and the start of the web's rapid rise. Anything before about 1994 doesn't really exist online. JPEG was brand new back then, but even so it seemed a little odd, as they were actually slower than software JPEG decoding on the Sun SparcStations we were using. It might have made sense on a PC, as PCs were significantly slower than Unix workstations at the time and we looked down on the poor folks who had to use them with pity, but I'm pretty sure the market must have been small and rapidly followed the dinosaurs into extinction.
    - spicyjpeg 3y ago
      The original PlayStation's main SoC (which is incidentally about 30 years old at this point) included a trimmed down JPEG decoder [1] meant to be used for video playback. It still relied on the CPU to handle Huffman decompression [2], but it allowed that otherwise anemic 33 MHz MIPS core to push 320x240 video at 30fps. [1] https://psx-spx.consoledev.net/macroblockdecodermdec/ https://psx-spx.consoledev.net/macroblockdecodermdec/ [2] https://github.com/Lameguy64/PSn00bSDK/blob/master/examples/mdec/strvideo/main.c#L11-L27 https://github.com/Lameguy64/PSn00bSDK/blob/master/examples/...
- orangepurple 3y ago
  I'm using the CPU for other calculations. I don't need the extra overhead from removed hardware acceleration support.
  - CyberDildonics 3y ago
    Sounds like you need more than a $50 computer. Might be time to adjust these expectations.
LargeTomato 3y ago
If you know what hardware video encoding is you are not the target audience for raspberry pi.
- MegaDeKay 3y ago
  This. I think people often forget that RPi is made to a price and intended to provide a balanced suite of capabilities at that price to their target markets. The delta sales due to a lack of something like hardware video encoding will be near zero.
- redundantly 3y ago
  Yeah, completely ignoring the fact that they included both encoding and decoding before, yup! Totally not the target audience! The rPi foundation doesn't care about the individual end users anymore. They make most of their money supporting commercial vendors. They're keeping up the appearance of being good to the little guy so they can keep their support ecosystem alive and well.
  - creeble 3y ago
    I’ve yet to find a Pi-based commercial product that sold more than 10k units, ever. Do you have evidence to the contrary?
    - stonogo 3y ago
      Spin scooters, while not for sale, number approximately 50k units and are powered by a full-size Raspberry Pi 4.
    - nereye 3y ago
      Korg uses the Pi Compute Module 3 [0] in at least three of their products (wavestate, modwave, opsix), with Korg being one of the major synth manufacturers and overall annual keyboard synthesizer sales in the US alone being over 100K units. Don't have hard evidence but am assuming that Korg has sold more than 10K units of those three products. [0] - https://www.raspberrypi.com/success-stories/korg-synthesizers/ https://www.raspberrypi.com/success-stories/korg-synthesizer...
badcppdev 3y ago
It's quite all right for you to be unhappy about the lack of one feature on the newest version of the RPi but I think you're overstating it by claiming that they are shooting themselves in the foot. Also are you saying that you've already switched to the Orange Pi 5? That's amazingly fast considering the RPi5 was launched less than 2 weeks ago.
- orangepurple 3y ago
  RK3588S is the name of the game
- snvzz 3y ago
  >Also are you saying that you've already switched to the Orange Pi 5? That's amazingly fast considering the RPi5 was launched less than 2 weeks ago. Doesn't strike me as odd. Orange Pi 5 has been available for a relative long time. After seeing the Raspberry Pi 5 they were holding on for is not worth it, they finally pulled the plug with Orange Pi 5.
  - nsteel 3y ago
    Strikes me as odd to make an emotional decision over lack of hardware support for something. I'm going to wait and see if I notice a difference, I am told that I won't. And I already know from bitter experience that those other boards are great until you find the thing that doesn't work and is *NEVER* fixed.
    - snvzz 3y ago
      There's a lot of boards out there. They are not actually divided into groups like "Raspberry Pi" and "the others, which are broken." Way to talk about emotional decisions while being bitter about some bad experience with another SBC. I have personally had great experiences with other SBCs, and I'm not just talking VisionFive2, but ARM ones, over the years. There's a lot of them, and it takes some care in choice. Insist on the few ones that have SoCs with good upstream support. There's also a tendency for good ones to come in series; the already named Orange Pi 5 has many predecessors, and established reputation. And a friendly reminder below that Raspberry Pi does have its own dark history. Be it Codec licenses to use the hardware acceleration, camera DRM, or hardware bugs. Such as the USB protection dropping too much voltage on the first Pi, or the 2nd generation rebooting when exposed to light, or the third "+" models erratic ethernet behavior. Or the main serial port suddenly not working, because now it's bluetooth's, and you have to use an alternative, unreliable one which baudrate scales with cpu clock. And that broken serial port staying broken despite RPi4 added more hardware serial ports. Or RPi4 shipping w/o a heatsink, and throttling constantly until one is added. And RPi5 still not supporting voltages above 5v via usb-pd, despite demanding more peak current than ever before. Then there's people issues, like forgetting its good-for-society not-for-profit and focusing on making money. The "Pi 400", as well as prioritizing industry customers; it was really hard to get a Pi4 for a period of time spawning years. I'm saying this despite I own all models, keep most of them in use still, and will likely get a rpi5 down the line, just to mess around, despite trying to move most of my computing to the standard RISC-V ISA.
ThatPlayer 3y ago
On other hand, I've got multiple RK3588 boards and am getting a Raspberry Pi 5 for its proper Vulkan drivers. It really does depend on your use case. Panfrost team is doing good work for GPU drivers for the RK3588 boards, but I expect that to take longer
snvzz 3y ago
I am still shocked by Raspberry Pi's decision to put their weight behind MPEG's HEVC (and exclusively accelerate that codec, removing support for the others), instead of Aomedia's AV1. Then again, I feel the same way for remaining ARM, when they could very well have, with a similar timeline, moved to RISC-V RVA22+V. Many do not understand this now, but will in the coming months when VisionFive3 and other RVA22+V boards launch. Such a wasted opportunity for Raspberry Pi.
- nsteel 3y ago
  > with a similar timeline, moved to RISC-V RVA22+V. Look how long it takes them to bump their software releases. And you want to throw an entirely new architecture in the mix? Why? I don't understand how you can be shocked by the decision to stick with what works and sells (very well). Which of their goals would be furthered by taking this significant risk?

Doesn't matter anyway. They shot themselves in the foot by removing hardware video encoding. I switched to the Orange Pi 5 so I can keep my video encoders and also benefit from the nice TPU.

gjsman-1000 3y ago

This doesn't really make any sense - you can video encode 1080p H.264 on the Pi 5 in software without using more than ~25% of the CPU. And the quality, if you don't mind trading more CPU, will be far superior than the hardware encoders ever were.

phkahler 3y ago
I remember seeing an ad many decades ago for a PC ad-in board for encoding and decoding JPEG. You know, back when dinosaurs roamed the earth and that might have been a thing for desktop publishing or whatever. General purpose CPU has gotten fast enough to do a lot of things that used to benefit from hardware acceleration.
- mhandley 3y ago
  A little over 30 years ago when I started working on multimedia conferencing over the Internet, for some reason we had a couple of hardware JPEG decoder ISA (?) cards from Bitfield Oy, a Finish company (if I recall correctly). I can't find any documentation on them now, as this predates NCSA Mosaic and the start of the web's rapid rise. Anything before about 1994 doesn't really exist online. JPEG was brand new back then, but even so it seemed a little odd, as they were actually slower than software JPEG decoding on the Sun SparcStations we were using. It might have made sense on a PC, as PCs were significantly slower than Unix workstations at the time and we looked down on the poor folks who had to use them with pity, but I'm pretty sure the market must have been small and rapidly followed the dinosaurs into extinction.
  - spicyjpeg 3y ago
    The original PlayStation's main SoC (which is incidentally about 30 years old at this point) included a trimmed down JPEG decoder [1] meant to be used for video playback. It still relied on the CPU to handle Huffman decompression [2], but it allowed that otherwise anemic 33 MHz MIPS core to push 320x240 video at 30fps. [1] https://psx-spx.consoledev.net/macroblockdecodermdec/ https://psx-spx.consoledev.net/macroblockdecodermdec/ [2] https://github.com/Lameguy64/PSn00bSDK/blob/master/examples/mdec/strvideo/main.c#L11-L27 https://github.com/Lameguy64/PSn00bSDK/blob/master/examples/...
orangepurple 3y ago
I'm using the CPU for other calculations. I don't need the extra overhead from removed hardware acceleration support.
- CyberDildonics 3y ago
  Sounds like you need more than a $50 computer. Might be time to adjust these expectations.

LargeTomato 3y ago

If you know what hardware video encoding is you are not the target audience for raspberry pi.

MegaDeKay 3y ago
This. I think people often forget that RPi is made to a price and intended to provide a balanced suite of capabilities at that price to their target markets. The delta sales due to a lack of something like hardware video encoding will be near zero.
redundantly 3y ago
Yeah, completely ignoring the fact that they included both encoding and decoding before, yup! Totally not the target audience! The rPi foundation doesn't care about the individual end users anymore. They make most of their money supporting commercial vendors. They're keeping up the appearance of being good to the little guy so they can keep their support ecosystem alive and well.
- creeble 3y ago
  I’ve yet to find a Pi-based commercial product that sold more than 10k units, ever. Do you have evidence to the contrary?
  - stonogo 3y ago
    Spin scooters, while not for sale, number approximately 50k units and are powered by a full-size Raspberry Pi 4.
  - nereye 3y ago
    Korg uses the Pi Compute Module 3 [0] in at least three of their products (wavestate, modwave, opsix), with Korg being one of the major synth manufacturers and overall annual keyboard synthesizer sales in the US alone being over 100K units. Don't have hard evidence but am assuming that Korg has sold more than 10K units of those three products. [0] - https://www.raspberrypi.com/success-stories/korg-synthesizers/ https://www.raspberrypi.com/success-stories/korg-synthesizer...

badcppdev 3y ago

It's quite all right for you to be unhappy about the lack of one feature on the newest version of the RPi but I think you're overstating it by claiming that they are shooting themselves in the foot. Also are you saying that you've already switched to the Orange Pi 5? That's amazingly fast considering the RPi5 was launched less than 2 weeks ago.

orangepurple 3y ago
RK3588S is the name of the game
snvzz 3y ago
>Also are you saying that you've already switched to the Orange Pi 5? That's amazingly fast considering the RPi5 was launched less than 2 weeks ago. Doesn't strike me as odd. Orange Pi 5 has been available for a relative long time. After seeing the Raspberry Pi 5 they were holding on for is not worth it, they finally pulled the plug with Orange Pi 5.
- nsteel 3y ago
  Strikes me as odd to make an emotional decision over lack of hardware support for something. I'm going to wait and see if I notice a difference, I am told that I won't. And I already know from bitter experience that those other boards are great until you find the thing that doesn't work and is *NEVER* fixed.
  - snvzz 3y ago
    There's a lot of boards out there. They are not actually divided into groups like "Raspberry Pi" and "the others, which are broken." Way to talk about emotional decisions while being bitter about some bad experience with another SBC. I have personally had great experiences with other SBCs, and I'm not just talking VisionFive2, but ARM ones, over the years. There's a lot of them, and it takes some care in choice. Insist on the few ones that have SoCs with good upstream support. There's also a tendency for good ones to come in series; the already named Orange Pi 5 has many predecessors, and established reputation. And a friendly reminder below that Raspberry Pi does have its own dark history. Be it Codec licenses to use the hardware acceleration, camera DRM, or hardware bugs. Such as the USB protection dropping too much voltage on the first Pi, or the 2nd generation rebooting when exposed to light, or the third "+" models erratic ethernet behavior. Or the main serial port suddenly not working, because now it's bluetooth's, and you have to use an alternative, unreliable one which baudrate scales with cpu clock. And that broken serial port staying broken despite RPi4 added more hardware serial ports. Or RPi4 shipping w/o a heatsink, and throttling constantly until one is added. And RPi5 still not supporting voltages above 5v via usb-pd, despite demanding more peak current than ever before. Then there's people issues, like forgetting its good-for-society not-for-profit and focusing on making money. The "Pi 400", as well as prioritizing industry customers; it was really hard to get a Pi4 for a period of time spawning years. I'm saying this despite I own all models, keep most of them in use still, and will likely get a rpi5 down the line, just to mess around, despite trying to move most of my computing to the standard RISC-V ISA.

ThatPlayer 3y ago

On other hand, I've got multiple RK3588 boards and am getting a Raspberry Pi 5 for its proper Vulkan drivers. It really does depend on your use case. Panfrost team is doing good work for GPU drivers for the RK3588 boards, but I expect that to take longer

snvzz 3y ago

I am still shocked by Raspberry Pi's decision to put their weight behind MPEG's HEVC (and exclusively accelerate that codec, removing support for the others), instead of Aomedia's AV1. Then again, I feel the same way for remaining ARM, when they could very well have, with a similar timeline, moved to RISC-V RVA22+V. Many do not understand this now, but will in the coming months when VisionFive3 and other RVA22+V boards launch. Such a wasted opportunity for Raspberry Pi.

nsteel 3y ago
> with a similar timeline, moved to RISC-V RVA22+V. Look how long it takes them to bump their software releases. And you want to throw an entirely new architecture in the mix? Why? I don't understand how you can be shocked by the decision to stick with what works and sells (very well). Which of their goals would be furthered by taking this significant risk?