The base M4 is a very small chip with a modest memory config. Don’t get me wrong, it’s fantastic, but it’s more Steam Deck/laptop than beefy APU (which the M4 Pro is a closer analogue to).
$1200 is pricey for what it is, partially because Apple spends so much on keeping it power efficient, rather than (for example) using a smaller die or older process and clocking it higher.
Yeah, that would be perfect!
Or (alternatively) they could majorly underclock the a shrunken series X chip to make it equivalent to an S.
Games are complex. Qualcomm/MS may tune it for the most popular titles, but I just don’t see how they can catch up to years of desktop GPU driver development.
Ehh, I agree in principle but have grown cynical.
The American voters clearly don’t care about actual qualifications. The presidency is an attention contest, pure and simple. Truth is relative. Even if we get a highly qualified president in (and I don’t believe that’s possible anymore) they’re going to be totally beholden to Facebook and Twitter politics memes.
I’d rather have someone that can dominate the narrative and wield actual political power to do decent things, even if their decisions aren’t always the best. Like… who else could bring millions into civic engagement?
Yeah, any framework with a “big” GPU is just so expensive.
Eh, yeah, and it’s backordered.
Ideally I’d like a full x16 slot too (or at least electrical x8), but perhaps that’s asking too much.
Also, is it even standard ITX?
I will believe it when I see it. I hope so.
Qualcomm makes a lot of hype/noise but historically tends to overpromise, and also makes some unforced blunders. But a real ARM competitor would be great.
It means emulation with pretty much every current title, and graphics driver issues and sluggish game out of the wazoo (as Qualcomm is very different than AMD/Intel/Nvidia).
ARM being more power efficient is also kind of a meme. Intel/AMD can be extremely good when clocked low (which they can do since there’s no emulation overhead), with both the CPU/GPU. Apple just makes x86 look bad because they burn a ton of money on power efficiency, but Qualcomm is more in the “budget” space. No one is paying $2K for an Xbox handheld like they would for an Apple product.
With a Qualcomm chip though… there will be some teething issues, best case.
She’ll have a full cabinet, policy crafted by an army, and practical experience under her belt. She’s clearly charismatic. AFAIK she’s not anti science or conspiratorial or anything… So what if she’s not a Harvard Graduate? As much as I’d like an STEM PhD in the White House, compared to most alternatives, I almost view that as a plus.
And again, she will have so much power. It would be like Trump, where she can swing unpopular stances through sheer force of will (like Trump is doing in the current Middle East visit: https://www.axios.com/2025/05/15/trump-israel-syria-policy-reverse-biden )
I mean, she’s managed her image and business this long. I think politics is a good fit for her skillset.
As is the power she’d wield over the Democrat Party. They would worship the ground she walks on (in an attempt for some of that celebrity to rub off).
…But the election would be gruesomely dirty. I can only imagine all the shit that would get thrown at her.
These things are awesome.
My dream is:
One embedded onto an ITX board.
An SKU with a single (8 core (ideally X3D?)) CCD but the full GPU.
Using Qualcomm chips
Oof.
Why didn’t they go AMD, or heck, even Intel? They have GPU-heavy APUs in the pipe that would mostly just work.
Intel, in particular, is not bad power-wise as long as they aren’t clocking chips to very edge like they’ve been doing, and won’t necessarily have the TSMC capacity constraint. That’s huge.
How is Trump showing up on Bluesky.
Also… is this real?
Hence she chases it down with Coca Cola ™
Ahh…
TBH I never use the apps. Of course it uses location.
I mean, there’s no way that address is really OOPs, heh, unless it got it from the IP (which could be injected into the chat I suppose).
OK, yes, but that’s just semantics.
Technically pretraining and finetuning can be very similar under the hood, with the main difference being the dataset and parameters. But “training” is sometimes used interchangeably with finetuning in the hobbyist ML community.
A federal agent injecting themselves into a random chat? I find that extremely unlikely.
It’s possibly an existing joke it found in a web search with similar coordinates? That it can do. Or maybe it got lucky and stumbled upon them in a search.
Grok is the laughing stock of the ML field. It’s horribly inefficient, performance is not good for its size/compute, and it’s “leverage” (Twitter’s userbase, Tesla cars) is objectively at risk. They’re even more closed than OpenAI, much more than Google. They only exist because Elon burned billions on a shit ton of H100s and seemingly copied what others are doing.
xAI (so far) is a joke. That could change, but unless they do something interesting (like actually publishing a neat paper or model), if they were even public, I would short them.