This is being produced in relatively low numbers (thousands) , so software development is a factor. Just plopping a scripted browser in kiosk mode on Debian is cheaper than ESP32 UI development.
Idk, if they’re plugging in one for each screen it sounds like a lot; there are libraries to do most of this. It wpuld only take me about a month or someone competent a couple days to write this. I kbow there’s libraries to display, but i don’t know what else this is monitoring/controlling. So that seems safe,
So there’s a computer hardware cost that goes from ~5x(4?) Per machine to ~45x4 per machine. That’s ~ 2 hours of code per machine difference that this would make, assuming you were paying ~80/hr to write it, which is reasonable.
Even assuming no code was needed for the pi, production takes twice as long as expected, and electricity costs don’t matter (which, next to the condenser; they may not) you break even at ~16 machines. 20 if you want to throw in some other random arbitrary cost.
Even if you assune pi 0’s, at, what 20/each? You still break even before 100 units.
So it would take less than a hundred machines for smaller chips to pay off. I’d believe an exec didnt (even ask someone else to) do this math, but how long have pi’s had multiple video out’s?
For an Esp32 you’d need to take a larger model which has psram. With the Pi, yes a is take a zero (Zero 2w or so). The Pi already has hdmi on board and a graphics chip and accelerator, while for the ESP32 you’d need a custom solution.
The price difference is maybe 10 Dollars per piece or so. On the PI I have 512Mb of RAM and what ever SD they put in for storage. On the Esp32 I have 8 psram or so and a tiny bit of flash.
Ah right, for the ESP i probably need to wire up a sd card, custom board, all that stuff, to just store that 24bit 1024x768bit image.
Naah, while I love my ESPs and am just build a project with one - the PI is just so more competent for this task while still being damn cheap.
A decent Esp 32 board is around Eur 5, a. pi zero 2w around 20. Compute module proably similar - customer prices.
That’s a 15 Euro difference.
Ah and my developer pool who can code for Unix is a LOT bigger than the pool who have commercial experience for the Esp32.
I can’t follow your math, at 100 units the price difference is 100x15 for me, which is 1500.- About a day of developing for a small team, if the office and hardware is free. More if you pay for those, too.
When I calculate, custom development always is more expensive.
The thing that gets me is that they seem to have a separate machine for each display
easier to buy 10 rpis than a single embedded system with 10 diplay ports
Dont need remotely a full rpi for this. A pico or esp32 might do the trick.
This is being produced in relatively low numbers (thousands) , so software development is a factor. Just plopping a scripted browser in kiosk mode on Debian is cheaper than ESP32 UI development.
Idk, if they’re plugging in one for each screen it sounds like a lot; there are libraries to do most of this. It wpuld only take me about a month or someone competent a couple days to write this. I kbow there’s libraries to display, but i don’t know what else this is monitoring/controlling. So that seems safe,
So there’s a computer hardware cost that goes from ~5x(4?) Per machine to ~45x4 per machine. That’s ~ 2 hours of code per machine difference that this would make, assuming you were paying ~80/hr to write it, which is reasonable.
Even assuming no code was needed for the pi, production takes twice as long as expected, and electricity costs don’t matter (which, next to the condenser; they may not) you break even at ~16 machines. 20 if you want to throw in some other random arbitrary cost.
Even if you assune pi 0’s, at, what 20/each? You still break even before 100 units.
So it would take less than a hundred machines for smaller chips to pay off. I’d believe an exec didnt (even ask someone else to) do this math, but how long have pi’s had multiple video out’s?
X11 can easily handle multiple screens.
Not sure about the Pi’s limitations but COTS SBCs can too.
Can’t read the text in the image but i’m informed it’s a crash.
Which would mean that machine, or that virtual machine, is not on any of the other screens. Right?
Yeah, it seems as though it’s one controlling (virtual) machine per dispenser, so 1:1 to the screens.
Im really not sure it’s vm’s though. Thats separate displays.
For an Esp32 you’d need to take a larger model which has psram. With the Pi, yes a is take a zero (Zero 2w or so). The Pi already has hdmi on board and a graphics chip and accelerator, while for the ESP32 you’d need a custom solution.
The price difference is maybe 10 Dollars per piece or so. On the PI I have 512Mb of RAM and what ever SD they put in for storage. On the Esp32 I have 8 psram or so and a tiny bit of flash.
Ah right, for the ESP i probably need to wire up a sd card, custom board, all that stuff, to just store that 24bit 1024x768bit image.
Naah, while I love my ESPs and am just build a project with one - the PI is just so more competent for this task while still being damn cheap.
A decent Esp 32 board is around Eur 5, a. pi zero 2w around 20. Compute module proably similar - customer prices.
That’s a 15 Euro difference.
Ah and my developer pool who can code for Unix is a LOT bigger than the pool who have commercial experience for the Esp32.
I can’t follow your math, at 100 units the price difference is 100x15 for me, which is 1500.- About a day of developing for a small team, if the office and hardware is free. More if you pay for those, too.
When I calculate, custom development always is more expensive.
I hope the machine is up to the job. I’d pack at least 64 gigs of RAM and a nice GPU.
At some point you realize that the slushee has melted in the face of your system’s 1000 W TDP…
That seems too low. What kind of low spec AI slush master 7000 are you using?
That doesn’t sound like enough, need at least like 128gb for the ai chat bot that you ask to change the picture for you
I totally forgot about the necessity to put AI into them, I’m so sorry!
What about some blockchain to track the usage?
If AI can’t read my bank account history and generate an NFT of the Slurpee I can afford to finance, what’s the point?
Yeah, modular systems. Buy 2, buy 20, setup time is roughly the same.