This is the way.

Exactly. If anything, signs like these drag the attention away from the actual issue.

Just to clarify, do you want multiple remote users connecting to the same desktop session or to separate desktop sessions on the same computer (like a terminal server)?

They can and they are making their own chip designs to do the job.

The cloud part of Apple Intelligence runs on their own designed hardware.

I just use whatever is included with the desktop environment. On KDE and GNOME launching an application involves pressing the Super (“Windows”) key, typing the first couple of letters of the application I want to launch and pressing the return key.

I might be missing something here but I don’t know how other launchers could possibly make this a simpler process.

Show 'em, that’ll teach these nasty fanboys! Reads like writing that got you a big dopamine rush.

I agree, commenting “Use Firefox!!!1!11” on every post remotely related to (other) browsers doesn’t help anybody, just like commenting “Use Linux!!!1!11” on every post about a vulnerability in Windows doesn’t contribute anything meaningful at all.

Look, I also disagree with what Mozilla is doing here and yes, they 100% deserve the flak they are getting for it. But - like most things in life - it’s not black and white. Firefox could still be less intrusive to your privacy than Chrome (I’m not saying it necessarily is, but it could be that way). A different example: your mail provider could track every time you login to your account, or it could analyze and track the content of every email you receive. One is clearly worse than the other, right?

Which browser(s) do you recommend/use?

There is no definitive roadmap.

Let’s see if this really affects all Linux systems or if the stars need to align for this to actually be exploitable.

65 watts will be fine without the dGPU module. If anything, the CPU will be limited to a slightly lower wattage (see here). EDIT: this seems to be a bug.

The “Apple TV” is Apple hardware.

I’m not sure how that would help in letting lost people go.

How do you not do that? It’s all in your local network, how would it not work offline…?

That’s why it’s there.

What I mean by that is that they will take a huge disservice to their customers over a slight financial inconvenience (packaging and validating an existing fix for different CPU series with the same architecture).

I don’t classify fixing critical vulnerabilities from products as recent as the last decade as “goodwill”, that’s just what I’d expect to receive as a customer: a working product with no known vulnerabilities left open. I could’ve bought a Ryzen 3000 CPU (maybe as part of cheap office PCs or whatever) a few days ago, only to now know they have this severe vulnerability with the label WONTFIX on it. And even if I bought it 5 years ago: a fix exists, port it over!

I know some people say it’s not that critical of a bug because an attacker needs kernel access, but it’s a convenient part of a vulnerability chain for an attacker that once exploited is almost impossible to detect and remove.

That’s so stupid, also because they have fixes for Zen and Zen 2 based Epyc CPUs available.

Intel vs. AMD isn’t “bad guys” vs. “good guys”. Either company will take every opportunity to screw their customers over. Sure, “don’t buy Intel” holds true for 13th and 14th gen Core CPUs specifically, but other than that it’s more of a pick your poison.

We have yet to see what other AI 300 SKUs support when they release then I guess, and I usually take whatever LTT says with a big grain of salt. That being said, Linus might have some insider information because of his Framework investment. Framework will very likely know by now what their next mainboard supports in terms of memory.

It’s a weird timeline. CAMM2 modules are basically right around the corner, and Intel’s Core Ultra 2 only supports on-chip memory, and now AI 300 seemingly only supports soldered memory? This is all a bit weird to me.

I’d be fine transitioning to CAMM2 modules and making SODIMM obsolete, but a Framework mainboard with soldered on memory wouldn’t sit right with me.

DDR5 is supported up to 5600 MT/s (that’s as high as JEDEC spec currently goes for DDR5 SODIMMs afaik) while I’ve seen Ryzen AI notebooks using LPDDR5X with 7500 MT/s. Latency should also be lower on LPDDR5X as it can be physically closer to the CPU.

There are some performance comparisons on older Ryzen CPUs/APUs showing that the iGPU benefits a lot from faster LPDDR compared to DDR.

It’s definitely supported (although many manufacturers will use LPDDR5 for faster speeds at the cost of upgradability/repairability), see https://www.amd.com/en/products/processors/laptop/ryzen/300-series/amd-ryzen-ai-9-365.html

My guess would be that Framework releases a model/mainboard featuring Ryzen AI 300 (Zen 5). It’ll probably be a while as they are in the process of releasing Core Ultra Series 1 right now, and these CPUs have been out since mid-December 2023.

So I wouldn’t expect a new AMD model to be released before 2025. Maybe it’ll be announced at the end of this year though.

This. Even I would be too paranoid to keep using a phone (or other device for that matter) that the police confiscated before.