There's a fundamental issue that is dependent on physics and how atoms and molecules move about inside of materials that has to do with CPUs. And that is that the only way to get more power is either raise the clock speed, or recompile software with higher CPU flags turned on which pushes more of the computational work into giant opcodes inside of the CPU, or add more CPU cores and rewrite the software to be multithreaded.
And of these 3 things, guess which one is the easiest? Yep, #1, increasing clock speed so your inefficiently coded software runs lickety split. And the faster you switch electrons in the CPU and the denser the CPU core is - the more heat is produced and the more power is sucked. The inefficient code that makes up Microsoft Windows is too bloated to run optimally on anything other than a high power CPU which means massive sucking power drain and more massive sucking power drain for the cooling apparatus for the CPU. And that same code bloat is also affecting Open Source software. We see it everywhere. For example Chrome now on Ubuntu is a SNAP package not native binary. And the Debian binaries are all compiled with very low CPU flags so they can run on the crappiest antique Core2 CPUs out there that might run Linux. And more code is python, perl, php, and other bloated, slow inefficient interpreted languages that are not compiled. The code bloat MIGHT make it easier to quickly code something - although frankly I can't see any real difference in the complexity of, say, Visual Basic (which can be compiled into actual binaries not p-code with an older version of Visual Studio) and Python or PHP or the learning curve needed. But the price of that "ease" is requirement for high power CPUs and those are NOT going to sip power. And the OTHER fundamental issue that is dependent on atoms and molecules is that electrical energy storage in batteries is a horribly inefficient process of converting electrical energy into chemical energy during charging and then back again during use. And on top of that - the larger and fatter and heavier the battery the more chemicals it can hold and thus the more electricity it can store. I have told people constantly that you CANNOT have high power in a lightweight, compact case. But they see movies like Star Wars with the Jedi Knights and their lightsabers which emit enormous amounts of power (honestly, do you really know how much power is required for melting a steel door that is 4 feet thick like they did in Phantom Menace) See estimates like this one: Stardestroyer.net/Empire/Tech/Beam/DeathStar.html Which estimates at minimum 1E32 joules of energy to destroy Alderran which of course is far greater than would be physically possible by anything short of a Sun. People insist that they can buy a laptop in the store that has the output of the Boardman coal fired plant encased in it then get mad when that doesn't happen. You cannot argue with the laws of physics, folks. If you MUST have real computer power, your mobile device needs to be nothing more than a terminal to a REAL computer tied into main power. Otherwise you are just kidding yourself. Ted -----Original Message----- From: PLUG <[email protected]> On Behalf Of Eldo Varghese Sent: Sunday, September 15, 2024 9:38 PM To: [email protected] Subject: Re: [PLUG] Seeking something between Raspberry Pi and NUC I will mirror the previous recommendations of chromebook (Ted & Ben) and other tablets/laptop lite. If you want a separate dev 'box', run docker/kvm/virtbox etc. for some degree of separation from your local desktop environment. If you are still sure these chromebooks/tablets dont meet your requirement I would recommend the Radxa x4 [0] (the heatsink is a must apparently). You can go a bit bigger with (and more expensive) with UP boards latest offering [1] No matter which SBC you go with I highly recommend going with alder lake [2] or higher. -Eldo [0] https://radxa.com/products/x/x4/ [1] https://up-shop.org/default/up-7000-series.html [2] https://en.wikipedia.org/wiki/Alder_Lake#Alder_Lake-N On 9/14/24 3:20 PM, Ben Koenig wrote: > For a while intel was making mobile-class x86 processors. These popped up in > some chrome books and the Pixel Slate. It's discontinued, but I have a pixel > slate and it fits what you are looking for. > > You might be able to find that class of processor floating around in > obscure products. I would check all the shops that make > consumer-oriented devices with Linux pre-installed since there are > some interesting form factors. Maybe take a look at Starlabs.systems > https://us.starlabs.systems/pages/starlite > > TBH that tablet looks like someone saw Google discontinue the Slate and said > "OMG we have to do that!" > > If the firmware/bootloader on a chromeOS device is getting in your > way.. there are solutions for that https://docs.mrchromebox.tech/ > > After a firmware swap my Slate runs a slackware installation with NO > special modifications to the OS. x86 tablet with 8GB RAM :) > > -Ben > > > On Saturday, September 14th, 2024 at 10:14 AM, Eric House > <[email protected]> wrote: > >> I do a lot of traveling where a laptop is too much: won't fit in the >> airline seats I'm willing to afford, or won't survive weeks in a >> bicycle saddlebag. And so I've cobbled together a setup that lets me >> do terminal-based development anywhere: Raspberry Pi 5, powered by a >> USB battery pack, sits nearby. On table (or airline seat tray) sits >> an Android tablet running a terminal app, and in my lap sits a >> keyboard. The keyboard talks to the tablet using bluetooth, and the >> tablet connects to the Pi over wifi. >> >> The weak link is the Pi5: I'd love to have more than 8g of RAM (to >> run Android dev tools in an x86 emulator, for example) and >> faster/more reliable storage than a memory card. And so my question: >> >> Is there a class of computers out there low-power enough to run for >> hours on a USB battery pack but significantly more capable (and >> perhaps more "standard", e.g. able to run GRUB2 and generic Debian) >> than the Raspberry Pi 5? >> >> I figure a fanless and USB-C-powered NUC clone might be a starting >> point, but they don't seem to exist, which has me thinking power >> requirements and heat generation are still too high when using Intel >> chips. So maybe it has to be ARM-based? Anyway, I figure this group will >> know. >> >> Thanks, >> >> --Eric House
