System76 Will Start Designing and Building Its Own Linux Laptops in January 2020

The Intel offerings for mobile are very well developed and cost effective. Apple alone pushes quite a bit of development in this segment, in part because lots of us nerds out there with a commit history are running that hardware. Intel can supply powerful, integrated, shrink wrapped subsystems, etc already tailored for mobile form factors. Their marketshare here is crushingly dominant.

Conversely, the AMD mobile platform relies on realtek and others, eliciting howls from grumpy users with certain numbers of greying beard fibers. AMD doesn't supply the ethernet, wifi, storage in this fashion, requiring the integrators to do more work. Making a change will likely be costly, perhaps even riling the, now merely sated, Intel naga from its E-Commerce/Big Data fed slumber, into a veritable springtime of onerous contract terms and volume requirements.

Then there is of course the video subsystem wicket. I have encountered a bunch of anecdotal people that I mentally associate as BSD users who simply prefer the Intel UHD graphics. They like counting on the low power, and simple functionality, as do I. I recognize that AMD has improved by vast geologic scales in the last decade on the drivers and opensource front. I don't fully understand why the graphics haven't enveloped the world. Seems like everyone is happy with Intel UHD-whatever and Mali, or just take what comes and complain if it is buggy.

Additionally there is a mentality of "At least Intel works properly!" that is still floating around a bit, which I don't exactly understand fully. It seems useful to myself, now, to pause and reflect on Intel's ongoing hardware security vulnerability uh--management. Certainly a masterclass of disclosure and mitigation.

Finally, wrapping my afghan around these poor old bones here on my Internet enabled armchair, I wonder what life would be like if the CPU market were more like the automotive industry, and Intel were producing Taxis and Uber vehicles. Would this kind of systemic defect not have been completely ruinous? Much different industry of course, but one thing makes me chuckle--maybe those banks shelling out for those 30 year old Mainframe support contracts had the right idea all along.

https://system76.com/laptops/darter

https://system76.com/laptops/galago

I have one of the chunkier ones. The roominess of its insides make me happy whenever I open it up (e.g. to clean it up).

I hear ya, but aren't those efforts kind of like very high-performance controllers for other equipment? Surely the Cavium stuff isn't processing frames directly on the host cpu, instead they leverage the ISA to add support for specialized subsystems, (At requisite cost) right? Plus the advantages that the datacenters are chasing down with this architecture are density related, which is to say a lot of little things doing little jobs.

How much of the datacenter push is replacing general purpose, hot cores, with specialized lower power devices? Who's running The Cloud on arm yet? How many ARM cores are we gonna cram into a laptop? (I would like many, please.)

Wrt the Fujitsu effort, it seems like an interesting development, but I don't envision that platform hitting desktops. Its possible, but where are the units? You can already buy a Talos II today.

>The whole point of "SoC" is cramming everything onto the one package. It's the most efficient way to do things, and everything from the Raspberry Pi's Broadcrap chip to the biggest AMD EPYC is a SoC.

Yes of course. But this bloats the power and heat budget of the SoC by more than a little.

The biggest AMD EPYC SoC (I really disagree with this generalization of SoC but I understand what you mean) is a huge piece of silicon and not the same kind of ARM ISA lego game. Growing an arm die with specially designed HPC subsystems to that size, also including 128 PCIE lanes, and 10GbE, probably means some really expensive ISA purchases or in-house development, if looking at Fujitsu gives us any clues.

AMD and Intel are already experts at managing the power and heat budgets. Reproducing these efforts looks like another steep engineering cost to my eyes.

One additional mention, and of coursethank you for your time: Seems you might be interested by some other ARM ISA datacenter efforts, check out the RockPro64 Arm cluster that Pine Microsystems has created to host their online services. Ctrl+F "cluster" in the following link. I don't know where to find more detailed information yet. https://www.pine64.org/2019/11/05/brave-heart-edition-pineph...

Thankfully the Thelio is absolutely beautiful, so I still have high hopes for their in-house designs.