I think this is the major point of modular IO. You can’t switch boards if you are stuck with a single chassis with integrated IO.
1 Like
Granted.
But the way it is setup now, the width of the device seems pretty much fixed anyway as the modules plug directly into the board.
So do they actually gain flexibility for replacing the board in the same case or reusing that board in a different case?
If the width of the device is fixed anyway, would it make much of a difference if the board extended up until the sides of the casing at least in some places?
What I was aiming at saying was more along the lines of:
If they make a model with less modules, but instead more IO overall I’ll be interested.
But that would require a different motherboard layout, breaking compatibility with the one they are launching now. So it is unlikely to happen in the near future, as that would mean maintaining multiple board-formats for upgradability.
1 Like
Good point, although there’s an obvious solution : use I/O shields. We’ve had them on ATX motherboards for over 25 years. All our desktop cases have a common large square hole and each motherboard comes with an adapted steel plate to match its I/O.
On a laptop, it would be slightly different due to the usual placement of ports : I’m thinking each edge of the laptop would have a long “slit” for the connectors to poke out, and those slits could be masked by plastic I/O shields.
Given how little plastic would be involved, this could turn into 3D-printable parts, possibly with STL files auto-generated by motherboard manufacturers. Suppose you buy a new Asus motherboard for your Gigabyte laptop, you’d go to the Asus website, enter your board’s reference and laptop chassis reference, and it would churn out the STL files you need to print.
There’s no technical limitation here, only business BS could get in the way, but if it does you can bet that lots of makers will be flooding Thingiverse with the proper STL files.
If laptops become modular, standardization will happen on its own. “Invisible hand of the market” and all that jazz…
Been waiting to see this pop up again. I specifically use older laptops because of this. I’m getting back to pentium M’s and got my childhood laptop again. Turns out the mofo perfooooorms hard core. Enough that I’m modding it and upgrading it.
I’ll look at this when its real and out. Pinebook and usb c dock interests me more tho.
2 Likes
Yeah but you gotta run Windows XP, tho… 
Seriously, though, what do you run on your laptop that a Pentium-M is performer ??? It’s very hard for me to not make jokes about Notepad and Minesweeper, right here 
.
Since when I got kernel 5.11 and an SSD what chu talkin bout willis
Linux, with some hardware work and some swapouts. Going to get an express card dock for it so it can have a gpu do some oomph work or a X8i coprocessor or something. Because I can and why would I not.
Sucks to have to pay 1000 dollars a year to have a working laptop huh
This was 20 bucks on some ebay garage sale and the parts I’m getting cost pennies and I get a solid secure spectreless laptop out of it that no one in their right mind will steal that I can coreboot.
I need to go do that whinge post.
2 Likes
I have almost bought a pinebook a hojillion times. I’m going to do it sooner or later.
Goddammit @FaunCB you’re killing me here 
Imma see what old junk I have lying around.
1 Like
LOL I like your spirit, you remind me of a good friend of mine who’s stockpiled AMD processors because he hates the inclusion of a PSP in all the new ones.
Sadly, some of us have modern problems which require modern solutions, as the meme goes. Visual Studio 2019 wouldn’t run very well on a Pentium-M. SolidWorks 2020 wouldn’t even start.
What is that ? Best Google could find is a 2009 IBM PowerXCell-8i.
Actually, my extensive experience of being the family and friends’ IT guy has taught me that a laptop costs around 300 € / year. Meaning, if you buy a 600 € machine it should last you 2 years before becoming obsolete. I have a 2014 laptop that cost 2,000 € and it broke down last year.
This 300 € / year is a rather reliable rule. They should call it Nefastor’s Law.
Yeah that
Ah see thats the difference here.
If I had to make money on anything I have a rendering workstation for that.
Personal computers are different tho ;3
Would be more interested if attachable cpu socket 
Everybody says they want that, but do you really ?
The thinnest ZIF socket I’ve ever seen (in a laptop) was 4 mm thicker than soldering the CPU on the motherboard. It also complicates the cooling solution, which, you guessed it, makes it thicker.
I’ve seen a few laptops with socketed CPU’s. No one ever upgraded their CPU’s. For starters, this would require CPU manufacturers to produce more than one generation of chip with the exact same socket and that’s rarely happened.
Closest we came to a solution was when the Pentium 2’s and 3’s and some Athlon’s were packaged in cartridges. I think that’s the only time in my life where I was able to upgrade my PC’s CPU without also changing the motherboard. But that solution makes no sense now that we’re dealing with on-CPU RAM controllers.
2 Likes
There was another time, actually. The Intel 80386 could be upgraded to hardware floating-point by installing a 80387 “math” co-processor. I still have one in storage for sentimental reasons.
I wish I still had mine. And my LAPC-1
The fact of the matter is that pretty much, 1-2 sockets per generation in a meaningful way. Even if you have something like AM4, if you bought a B350 board back in 2017, with a Zen 1 CPU, you’re still on PCI-E 2 through the chipset, you don’t have PCI-E 4 support at all, you don’t have SAM support, you don’t have Zen 3 support, etc.
It might be compatible with the next year or two, but if I’m concerned about upgrade cycles, my upgrade cycle is > 1-2 years, and so the next time I upgrade, I’m stuck getting a new motherboard anyways.
The last socketed laptop I owned had an i5 2450M in Socket G2, there was a Socket G3 after that supported Haswell mobile CPUs after that.
The sockets definitely added thickness compared to a soldered solution, with no real benefit other than if my CPU died I could replace it. Which if there was a new CPU/motherboard available, why not upgrade?
And since there are no socketed x86 mobile CPUs available since G3 it makes sense to just solder the CPU on to the motherboard than get a socket and custom fab a PGA/LGA PCB for the CPU.
1 Like
I agree. I understand the ewaste angle, but unless they modularise CPUs onto some edge connector daughterboard a la pentium 2 on the same plane as the motherboard I don’t see a way of keeping it thin.
For the record I don’t see my idea as an option either, cost complexity, not to mention cost benefit analysis.
Anyway, for the manufacturer I bet the cost for the whole unit is so low its not even worth it.
Roll on BGA.
You might be interested to learn that there’s actually a very specific, very niche technology that is used to make BGA ZIF sockets. I’ve used it a lot back when I was working for Texas Instrument on early OMAP SoC’s.
Consider the problem : you’re designing a new smartphone SoC that is both very complex and very tiny. First run of engineering samples comes in. Do you start soldering it on various test boards ? Those are expensive, and you don’t have many chips yet, and the yield isn’t dialed-in…
So you find a way to socket a BGA. The method I used relied on what’s called an anisotropic elastomer interposer.
Yeah, it’s a mouthful. We just call them “elastomers” for short. The basic idea is a squishy material that conducts electricity only in the Z axis. It’s like it has millions of microscopic wires going through an insulator without touching each other. That allows you to mount a BGA package directly onto the board footprint where it would be soldered, but without soldering it. All you need then is to press the chip onto the elastomer to ensure electrical contact.
The elastomer is around 1 mm thick. Could be thinner but then it becomes difficult to handle, plus we needed them to survive at least a thousand insertion / removal cycles each. In terms of performance, it can easily deal with signals up to 10 GHz and currents found in laptops.
Ours came from Yamaichi Electronics. Made to order since we were testing new chips. That, and the limited market, means each one cost a thousand dollars. But that’s because we were only ordering 500 per year, IIRC. If it had been millions per year I’m sure the price would have been much lower.
Now, like I said, you do need to apply pressure on the chip for this to work. And of course to retain the chip since it isn’t soldered. So we had a little machined aluminum frame on the boards, with a lid that had a butterfly screw on top. Not exactly low profile. But in a laptop, this job would be done by the heat sink.
So in theory you could make a sub-millimeter high ZIF socket for a BGA. But it would probably be as expensive as the whole motherboard and it might break easily if you use it too much.
Here’s a more modern example (mine was from 2005) :
That’s a complete BGA socket rated for 30 GHz. You can see where the elastomer is.
Let’s just remember : the technology has been around almost since BGA has been invented but it’s always remained a made-to-order bespoke product used only by chip manufacturers. There are reasons for that. I think the number one reason is that there hasn’t yet been any application where you would want to socket a BGA chip.
4 Likes
Thanks, that was really very interesting.
I almost did a degree in microelectronics engineering, totally fascinated by this stuff.
Those elastomer connectors remind me of the screen conductors you’d see in old calculators. Similar concept if not the same material.
Cool as hell!
It’s the same technology, except in “1D” it’s a lot easier to manufacture than in “2D”
1 Like
God, I can imagine.
Must be a pain in the backside, especially if it needs repositioning. I imagine they’re not intended to be user-serviceable.
I liked the holddown mechanism in the example diagram - looked like a torque controlled screw.