Incoming wave of AI is making buying PCs riskier for businesses AI PCs due to land this year could increase businesses' risk of buying the wrong tech as there is still no current AI standard for software to work with and confusion remains over what makes up an AI PC. A slew of new systems were promised at this week's mega mobile tech show MWC, but not all of these were necessarily AI PCs …
But Borkzilla, Chipzilla & co need to sell AI PCs on the understanding that more cores, more GHz and more pixels is soo last millenium, so the only thing that marketing can think of to lure in the bait customer is now AI.
I'm sure that, by the time AI has tired itself out in marketing eyes, quantum will be the Next Best ThingTM.
But what will come after that ? I'm guessing marketing will have an apocalypse moment then.
"But what will come after that ?"
A) Quantum powered AI blockchains.
Re the article, I was just thinking the other day that if someone does jump onto the AI bandwagon, they are sure to regret whatever hardware they buy as it will be obsolete when the latest greatest chip based AI comes out weeks/months later.
It doesn't help (profit margins) that the average PC already has more power than your average user needs, even if you account for software bloat. Between that and stratospheric costs of living, people are going to wait longer in between upgrades and probably avoid doing so until something breaks or planned obsolescence hits. My newish (2023) Thinkpad running Linux is seriously something that I'll probably still be using in 8 years time. My desktop Mac is a 2019 model[1] and I'm just now starting to put money aside for a replacement in the future. I've brought out my mother as an example of the "average" user who needs a computer to deal with many daily tasks but doesn't live behind a keyboard. Her Macbook Air is going to last her until the hardware takes a shit and isn't worth fixing or it stops getting security updates.
[1] I'm far more demanding on my stuff than the average user, and still, outside of gaming[2] or having two bits of graphics software working on 35"x70" 300dpi images at once.
[2] FFXIV is the only game I regularly play, and it runs through Crossover/WINE so uses far more resources than it should. I'm getting tempted[3] to put a Linux partition on the thing for the better performance of Steamplay/Proton on there to cut down on this.
[3] Okay, I've been tempted for a long time, I'm just lazy and whinging about it is less work than actually doing it.
I only last year replaced my venerable Phenom II x4, a chip that was originally released I believe in 2009 (though purchased in 2011).
I've probably got more power demands than the average user too, though it is a GNU/Linux box.
Loads of people will end up doing otherwise unnecessary hardware upgrades just to run Win 11.
Sounds like my first build, although I did have replace the CPU in 2016 and upgraded the motherboard as well since it made sense with a NewEgg sale, and well, thermal paste failing when overclocked often does necessitate a new CPU. Both Microsoft's symbiosis with the PC industry and Apple's bottom-line being all about shifting new hardware really do mean a lot of unnecessary e-waste. Maybe if the Flathub rules changes mean that software I need gets ported to Linux, I can finally stop playing their game and just have a Desktop of Theseus again, but I'm not holding my breath.
[1] At least in terms of cash. My father's old one died a few months before he did and he wanted something shiny and rather bulletproof and didn't care too much about cost. Of course, a "free" laptop is kinda a crappy trade-off for losing a family member.
I have local NAS running off a Pi4, and had even had been running a Nextcloud server on it for remote access. Although, I ended up shutting that part down just given the glacial upload speed of an asymmetric connection. Using a zero-knowledge provider for important files is only a fraction of the cost of upgrading the home connection, and also serves as something off-site for things that I can't bear to lose. I used to have a wide variety of odd old hardware for playing around with, including a couple of SGI boxen but have reached the point in my life where getting rid of things that I don't actively use that aren't books or LPs became a priority.
That's just what Intel/AMD/Dell/HP/etc. want you to believe to force you to buy a top spec "AI PC" due to worries that to do otherwise would be akin to outfitting their whole company with Blackberries in 2018!
Smart companies will keep up the same policies they have been, and will probably slowly shift in "AI PCs" not because they feel they have to have them but because pretty quickly it will be impossible to buy a non AI PC in the price bands corporations typically buy from.
So the right PC is one that uses the huge stock of discounted PCs built with non-AI chips,
We know from previous experience, it will be highly unlikely for any of the current generation of AI-chips will be the right ones to run post 2028 AI workloads, also it will be hit and miss whether they can run AI workloads (if any are actually delivered to market) before then.
What's funny is that Apple and Qualcomm have been building "AI" circuitry into their iPhone and Android (and Mac) SoCs for over half a decade, while Intel and AMD sat on their hands ignoring that. But once ChatGPT got Wall Street to put the AI hype into overdrive they were suddenly all in!
Quite. When I buy computers for work, mostly they are required to run Word / Excel / Outlook / Powerpoint, and browse the web. Given that these requirements aren't going to magically change, so-called 'AI' will add absolutely bugger-all to the hardware which we require.
The NPU on your machine should be capable of completing the inference for co-pilot queries that otherwise would have to go back to the cloud. This improves the responsiveness of the Co-Pilot UX. The issue with the current platforms is that they just don't quite have sufficient TOPs to do this inference locally. MS have outlined an expectation of 40 TOPs being required.
Intel Meteor Lake combined TOPs for CPU+GPU+NPU = 33 TOPs
AMD Hawk Point combined TOPs for CPU+GPU+NPU = 39 TOPs
So the first genuine offering that is actually "AI capable" will likely be AMD's Strix Point CPU line being launched later this year.
As someone else mentioned, you really want all the AI work to be completed by the NPU exclusively - this will avoid wider system slow down and also will avoid killing your laptop battery, since its more efficient at completing these inference tasks than your CPU or GPU
“Buy vPro laptops this year and they will be AI PCs; buy vPro desktops and they won't”
In practical terms, apart from the obvious fact that one is a desktop and one a laptop, there is no difference! The “AI” hardware will be redundant anyway and if you have a machine cursed with such hype then the best course of action would be to find out how to switch it off to reduce power drain
Unfortunately, in the medium term at least, the “AI” elements will just be a tax on an artificially inflated new PC price tag
The most recent purchase was a quad core Ryzen5 late 2017 I think. It is more than capable of everything I want to do with it.
The other next {cough} newest is a small fanless dual core one I do my general scratching around on - like commentarding here - that was about 2014 I think.
My laptop is older still. Dunno when I got that.
Nobody needs all this preloaded bloat forced on by an upgrade. It should be available but not by default. I spend hours culling stuff forced on me every upgrade only for it come back a week later. I am not sure off the timing really but it is too often. I wonder if Microsoft is prepping us all for a miracle OS that is so efficient it will blow our minds.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
