4 ms·

Raspberry Pi gets a serious competitor as Particle unveils its 5G-capable SBC

30 points by sharpshadow 2y ago

talldayo 2y ago
[flagged]
- itishappy 2y ago
  Come to think of it, Raspberry Pi isn't winning any awards for searchability either.
  - GloriousKoji 2y ago
    I think it's obscure enough. You don't really combine fruit, an irrational number and electronics that often. Unlike say the combination of fruit and a type of baked good.
mouse_ 2y ago
Snapdragon SoC is cool. I would buy it if it weren't an "AI chip". I'm not paying extra for that trash.
- talldayo 2y ago
  If it's any consolation, 12 TOPS of Neural engine performance isn't really that much wasted silicon. You're probably dedicating more unused hardware for video decoding/encoding and various other GPU functions most SBCs never use.
- SillyUsername 2y ago
  You do know it's just a vector processor, i.e. like graphics card internals? :facepalm:
  - kcb 2y ago
    Yea it's literally just their dsp. Not even a particularly new SOC so was probably not originally advertised for AI.
- no1089 2y ago
  [dead]
tocs3 2y ago
" Finally, it has over 20 different camera sensors," Wow.
sharpshadow 2y ago
The Kickstarter page: https://www.kickstarter.com/projects/particle-iot/tachyon-powerful-5g-single-board-computer-w-ai-accelerator?ref=discovery https://www.kickstarter.com/projects/particle-iot/tachyon-po...
beryilma 2y ago
What makes Raspberry Pis competitive is not necessarily hardware; it is software. I guarantee this is not going to be a serious competitor to Raspberry Pi, at least not for the intended audience of Raspberry Pis.
- m463 2y ago
  the pi is not only software, but mindshare. people understand it, there is a community, and many flaws have been worked around. (and it helps kids and other folks get started) There are lots of other "better than the pi" devices. The beaglebone was always better hardware than the pi (and almost the same price) and the onboard pru was interesting. The jetson nano series is pretty interesting and also does hardware well. The new banana pi does networking well and runs openwrt, which is another well-understood platform. That said, it is nice to be in an era of these interesting small devices.
SushiHippie 2y ago
> Such is the case with Particle's Tachyon, the first 5G-powered, AI-enabled SBC that the company claims is suitable "for everyone and everything." > The Tachyon also sports an octa-core Qualcomm Kryo CPU, 4GB RAM, and 64GB of built-in flash storage. They want to market it for "AI" use cases, but only include 4GB of RAM?
- no1089 2y ago
  [dead]
creeble 2y ago
Anyone want my pile of CHiP SBCs? https://www.kickstarter.com/projects/1598272670/chip-the-worlds-first-9-computer https://www.kickstarter.com/projects/1598272670/chip-the-wor...
- znpy 2y ago
  I’ve still got mines… thanks but no thanks. If you want to give away a pocket chip though (the casing with keyboard) that would be interesting
nubinetwork 2y ago
4gb of memory and no wired Ethernet? I think I'll pass.
xet7 2y ago
Running modern web browser requires 16 GB RAM. With 8 GB RAM, browser crashes. If running some AI model, like llama3.1, it requires NVidia/AMD GPU and more than 16 GB RAM. I have run it at i7 laptop, that has 32 GB RAM and NVidia GPU, using Ubuntu 24.04, with commands: curl -fsSL https://ollama.com/install.sh https://ollama.com/install.sh | sh ollama run llama3.1 Without supported GPU, it will run in slower CPU mode. Running at M1 Air 16 GB RAM, at Asahi Ubuntu arm64, it runs at CPU mode. I have not yet tried at macOS. There is some SBC that has 16 GB or more RAM, like OrangePi, but I have not yet tried can they run llama3.1 . What kind of AI can run with 4 GB RAM ?

30 points by sharpshadow 2y ago

talldayo 2y ago

[flagged]

itishappy 2y ago
Come to think of it, Raspberry Pi isn't winning any awards for searchability either.
- GloriousKoji 2y ago
  I think it's obscure enough. You don't really combine fruit, an irrational number and electronics that often. Unlike say the combination of fruit and a type of baked good.

mouse_ 2y ago

Snapdragon SoC is cool. I would buy it if it weren't an "AI chip". I'm not paying extra for that trash.

talldayo 2y ago
If it's any consolation, 12 TOPS of Neural engine performance isn't really that much wasted silicon. You're probably dedicating more unused hardware for video decoding/encoding and various other GPU functions most SBCs never use.
SillyUsername 2y ago
You do know it's just a vector processor, i.e. like graphics card internals? :facepalm:
- kcb 2y ago
  Yea it's literally just their dsp. Not even a particularly new SOC so was probably not originally advertised for AI.
no1089 2y ago
[dead]

tocs3 2y ago

" Finally, it has over 20 different camera sensors," Wow.

sharpshadow 2y ago

The Kickstarter page: https://www.kickstarter.com/projects/particle-iot/tachyon-powerful-5g-single-board-computer-w-ai-accelerator?ref=discovery https://www.kickstarter.com/projects/particle-iot/tachyon-po...

beryilma 2y ago

What makes Raspberry Pis competitive is not necessarily hardware; it is software. I guarantee this is not going to be a serious competitor to Raspberry Pi, at least not for the intended audience of Raspberry Pis.

m463 2y ago
the pi is not only software, but mindshare. people understand it, there is a community, and many flaws have been worked around. (and it helps kids and other folks get started) There are lots of other "better than the pi" devices. The beaglebone was always better hardware than the pi (and almost the same price) and the onboard pru was interesting. The jetson nano series is pretty interesting and also does hardware well. The new banana pi does networking well and runs openwrt, which is another well-understood platform. That said, it is nice to be in an era of these interesting small devices.

SushiHippie 2y ago

> Such is the case with Particle's Tachyon, the first 5G-powered, AI-enabled SBC that the company claims is suitable "for everyone and everything." > The Tachyon also sports an octa-core Qualcomm Kryo CPU, 4GB RAM, and 64GB of built-in flash storage. They want to market it for "AI" use cases, but only include 4GB of RAM?

no1089 2y ago
[dead]

creeble 2y ago

Anyone want my pile of CHiP SBCs? https://www.kickstarter.com/projects/1598272670/chip-the-worlds-first-9-computer https://www.kickstarter.com/projects/1598272670/chip-the-wor...

znpy 2y ago
I’ve still got mines… thanks but no thanks. If you want to give away a pocket chip though (the casing with keyboard) that would be interesting

nubinetwork 2y ago

4gb of memory and no wired Ethernet? I think I'll pass.

xet7 2y ago

Running modern web browser requires 16 GB RAM. With 8 GB RAM, browser crashes. If running some AI model, like llama3.1, it requires NVidia/AMD GPU and more than 16 GB RAM. I have run it at i7 laptop, that has 32 GB RAM and NVidia GPU, using Ubuntu 24.04, with commands: curl -fsSL https://ollama.com/install.sh https://ollama.com/install.sh | sh ollama run llama3.1 Without supported GPU, it will run in slower CPU mode. Running at M1 Air 16 GB RAM, at Asahi Ubuntu arm64, it runs at CPU mode. I have not yet tried at macOS. There is some SBC that has 16 GB or more RAM, like OrangePi, but I have not yet tried can they run llama3.1 . What kind of AI can run with 4 GB RAM ?