Want to use N2+ for long term gaming cabinets.

[lloydhussell]

Hello. My friends and I are going to develop a video game and build it into an embedded machine to sell. But we are wondering about the reliable lifespan of these devices. It would be nice if they last 30+ years without trouble.

[lloydhussell]

I don't know how to accurately calculate or predict the lifespan of SBC products.
However, it can be said that thousands of units have been shipped every month for the past five years since the launch of the ODROID-N2 series, and the cumulative average defect rate is around 0.1 to 0.2%.
Although it is far from perfect, I think it at least meets a certain quality control level for general electronic products.


But to be honest, the 30-year lifespan of N2+ seems unlikely.
Most of the Dell and HP computers I have used over the past 30 years have broken down after about 5 to 6 years of use.

[mctom]

Lifespan of electronics also depends on how they are treated. Will the cabinet have proper ventilation, will it have industrial grade PSU, what kind of solid state storage do you expect to live for 30 years, and so on.

If I were you I'd focus on serviceability instead and pick M1, which is expected to be offered until 2036 and perhaps beyond that point.

[lloydhussell]

Okay, then maybe the idea is to provide a solution that allows for easily replaceable parts in the event of breaking. And also a way to add new small form factor computers in the event the current one is no longer supported.

[mctom]

True, nobody will give you 30 years guarantee, not at this price point. In space industry we designed stuff for about 15 years with 99,7% probability it will survive, and you can imagine the cost of military / space grade design, not to mention production.
I'm not sure which conditions are worse, vacuum, rocket vibrations, temperature variations and space radiation, or a bunch of kids.. To me both are equally destructive.

After 10 years of operating a swarm of identical machines you will have a good understanding of longevity of each of their components. Surely you will replace game controllers more often than anything else. After 10 years you will be able to clearly estimate how many M1s you need to stash to ensure another 20 years of operation. I don't think M1s in storage will age significantly, there are no components prone to aging when not in use. What will vastly increase their lifespan is maintaining as low SoC temperature as possible, which is super easy with its huge heatsink and some 2W of idle power.
And of course you'll need a bunch of microSD cards once you realize humanity is irreversibly descending into "everything is in the cloud". Together with SD card writer. After 30 years, it might not be as obvious. Do you own a working 3,5" floppy disk writer? Well I do.

So yeah, you could pull this off by merely replacing parts once in a while.

If you want more effort and port your 10-15 years old project into a new platform someday, here are a few things worth remembering.

In 2053, I suspect Linux will still be a thing. Android, probably not. Also Windows will run on Linux kernel and base its business model on closed source runtime libraries, lol.

If you're designing a game, I suspect it will use 3D graphics. I'm not sure OpenGL(ES) will make it to 2053, as the very same group is developing Vulkan now. Might as well be Vulkan that will ultimately be abandoned in favor of OpenGL, it's hard to say at the moment.
As long as you use a general purpose programming language that is around since 1980s (C, C++) those will still exist, along with gcc compiler.
Will ARM architecture live long enough to see that day, or will it be substituted by a cheaper and open source RISC-V? This is why you want to stick to the most generic tools and languages.

I suppose 2053 computers will be so small and fast you may be able to simulate modern M1 inside it whatever the architecture could be. I emulated The Need For Speed (1994) in DosBOX on RasPi4 one day and it worked. Note it was full emulation and not a compatibility layer.
Computers will never achieve 10GHz clock speeds, least in silicon technology. But with 256 cores, what stops you from doing such a feat as simulating an entire legacy computer.

Now if you want to make sure new computers will fit in place of M1, I wouldn't worry at all. I fully expect USB to live up to 2053 (unless consumers insist on having everything wireless and spend half their time charging things). The display interface doesn't matter much, you'll always find a matching display. And for a form factor, M1 is already large enough so almost any other modern SBC would fit in the same spot.

Very interesting endeavor indeed. How would I even approach the problem myself? Probably stick to M1 and stock replacements.

[lloydhussell]

Really helpful and informative. Big thanks, man. Will take all this into consideration as I am building.