It is getting progressively more difficult to set up a new Windows laptop without a cloud account. Currently it's a pretty esotaric command line prompt if you know the keybinding to get a terminal open.
The worst it can get is having to reinstall windows with an image modified to run first setup with a local account. Though with a lot of manufacturers it's a good idea to do that anyway to also avoid all of the bloatware they preinstall.
My takeaway from these complaints is there were 5th graders in the 90s that were more skilled at installing Windows. If you can't find the way around it (10 sec search) the natural reaction is not "I'll install Linux and complicate my life further".
In fact if getting stalled by a single screen in the installer is a showstopper, then you're in for a world of hurt with Linux.
To be fair, TFA lists more reasons to avoid Windows:
> Mine include the privacy concessions during setup, the years spent waiting for Microsoft to let people put the taskbar on the side of the screen again, and the advertising.
Of course you don't have to agree on all of them, but please don't assume everyone acts only because they can't find a tutorial on the net!
Windows Pro > Work or School Account > Join a Domain instead > Set password & recovery questions
The one component here that is not present on most machines straight from the manufacturer is the presence of the Pro version.
The home version of windows requires the account with increasingly limited alternatives.
As the author mentions in TFA, she's on the Arch staff as a package maintainer. I don't think there was a technical obstacle around the windows setup process.
I wonder. I sort of accidentally discovered that Rufus could edit my arm64 windows 11 install media without problem - adding a local account and bypassing the need for domain login (just as it does for amd64).
This isn't a new trick - just the regular approach to editing windows install media.
But that process still results in having Windows 11 installed on the computer. I don't think it's a stretch to assume someone on the Arch Linux staff has reasons for avoiding installing Windows which do not involve an inability to accomplish it.
Not sure there's a strong use case for one of these $6K laptops yet, considering you can get a mini PC running AMD Ryzen 495 or 395 with 96, 128 and 192 GB of memory for similar amount of money and connect remotely from a cheaper/smaller/more battery-efficient laptop. Microsoft also has this same chip in non-laptop form factor: https://www.microsoft.com/en-us/surface/devices/surface-rtx-....
Coding while flying is often the use case people bring up, but the power brick these laptops come with sometimes trip the plane's power plug in coach and you can't run local AI on battery for long. Also, over time planes will get better connectivity (Starlink etc.) and that particular full-offline scenario will go away.
For gaming, there are better options with dedicated GPUs like the Asus Zephyrus and similar.
The AMD hardware is x86_64, which means lower battery life on average for normal users, and as an architecture, it is very well supported across the Windows ecosystem.
The RTX Spark is ARM64, which means more efficient power management at the cost of a small hit to sustained performance loads. But comes at the cost of being less supported when it comes to available software packages.
Aside from this, I’m led to believe that the AMD stack when it comes to local models has more overhead related to its need to translate more NVIDIA native instructions and therefore can be less performant than Nvidia.
Not sure how true this is anymore, but ive yet to hear otherwise.
What? I have spent $10k on a computer but would never fly business or first class, that shit is way too expensive. Two trips would buy a new $10k computer!
but you're in a position to understand now, that people who do fly business everywhere, can also afford one of these laptops for a fraction of what they spend flying. that's the use case.
my style of discussing things (let's call it "give credit to your interlocuters") doesn't work on this site.
this discussion is filled with people who love high performance computing, and they are mostly saying "I wouldn't buy this". my comment was in response to a sharer of that sentiment. I picked onesuch comment since there is no way to reply to all of them, and pointed out "if you come at this from the other direction, you can see a population of people who will buy it.
>That doesn’t mean the people that buy this laptop fly business class
Last first class ticket I bought was US$13k, or a bit less than $500/hr in the air for the trip I took. A 6k laptop that I'm gonna use for 8-12 hrs/day for a couple of years is a rounding error in comparison.
The Nvidia based one uses the memory for the GPU allowing you to run AI models on the GPU, the system you're talking about has no such setup. These are Nvidia powered laptops that do what Apple does with their laptops where the system and GPU VRAM are shared as needed, except the idea here is that Nvidia designed these to perform very well for local inference, at least, that's how I understand it.
The core design is also approaching 3 years old at this point, so it’s not really even that competitive with the latest stuff out of the other chip makers. It’s fine, but nothing to write home about. Certainly not worth the crazy premium Nvidia is charging for it.
I feel like even if a next generation of chips dropped tomorrow, it'd be a while until they can scale up production and actually start selling products you or I could realistically purchase and by that time this new generation chip would already by a couple of years old with something much better and powerful in the final stages of design.
If it's anything like my DGX Spark, the "premium" is almost 100% just the memory cost, not the GPU or ARM core.
10 months ago when I got my Spark it was competitive in price with a similarly spec'd out Mac Studio. As RAM prices have gone even more parabolic, so has its price. And so has the Mac's. And the Strix Halo competitors.
I don’t even really think it’s “fine.” It’s hard to even understand who they are for. Just Josh Tech on YouTube did a great review of these laptops as they’re hitting the market.
Good battery life is probably the biggest upside.
The downsides are numerous:
- Middling GPU performance compared to discrete RTX mobile GPUs. And yes, Nvidia markets them for gaming, so we need to judge them in that context, too. Windows on ARM isn’t amazing for gaming for various reasons including lack of anti-cheat compatibility.
- Compatibility and/or performance issues on Windows ARM. You’re running a lot of stuff non-native
- As you mentioned, the CPU cores are nothing to write home about with very mid single core performance.
- Suffers greatly from “why am I not buying a MacBook Pro?” syndrome?
I think there may be a case for being a good performance value for AI training, but for local inference I’m not seeing how that makes sense. Heck, for local inference I think it’s better to run a server and then use the laptop as a client.
You’ve also got projects like Strata that are going to define trends. Why bother with $6000 massive shared RAM machines when you can get 60 tok/s on Qwen 3.8 Flash Next on a 16GB RX 9060XT with 64GB of RAM?
Want to do local inference on a laptop? Grab a Zephyrus G14, get the 64GB spec, run Strata, then as a bonus you can run AAA games at full blast.
Lack of anti cheat is rather outdated perception of Windows on Arm. Most popular competitive games like League of Legends have native arm versions now.
Even if you run games native you’re still getting worse gaming performance and spending more than a discrete RTX card.
And if your use case is gaming the question becomes why you’re spending actual thousands of dollars more than a Zephyrus G14/G16 with an RTX 5080?
The Windows on ARM issue is really only one small piece of the puzzle of why these are a bad value proposition. They just don’t have a sensible buyer persona. There’s not a lot of people who will have a good reason to buy these over something else.
Local AI folks should likely pick a MacBook Pro M5 Max for the memory bandwidth and faster token generation, and for the Unix toolset and macOS focus that the AI and industry has over Windows.
Whoever is left is just someone who must have CUDA, but can’t get enough VRAM in an RTX 5070Ti/5080. If someone knows who that buyer is and what software they run I’d be curious to know what it is.
The problem is more holistic. It’s a buyer persona problem.
Even if your game is running natively and your anti-cheat is supported, you’d have gotten better performance on a Zephyrus G14/G16 for less money.
If you want to do local AI a MacBook Pro is either keeping up or going faster (faster memory bandwidth, token output writes faster), with faster single-core CPU performance as a bonus for general purpose snappiness.
In my eyes, the only buyer left is someone who can’t stand a MacBook Pro, or needs Nvidia/CUDA and a lot of VRAM.
To be more generous to the RTX Spark platform, it does seem to be priced head-to-head with Apple. I just think it has an uphill battle.
I guess this is technically true, as the specs/prices seem to line up closely with a MacBook Pro.
Although, the thing is, for the AI tasks advertised, the MacBook Pros have faster memory, so token decode runs faster.
Then by using a Mac you also scuttle the issue of "My platform has a low rate of adoption" that Windows on ARM has had. That issue will get better over time but in the present it's still an issue.
As for the dedicated GPU gaming laptop market, you'd be surprised how it's evolved since the release date of your MSI Ge66. I mentioned the 2026 Zephyrus G16 specifically because it is a laptop with very few compromises, including heat and noise. If you don't specifically need a huge amount of GPU-addressable RAM, and you're just looking for good performance in games and graphics-intensive applications, it is a clearly better buy than these RTX Spark systems. As a bonus, being an X86 system you'll have a much easier getting Linux going on that one.
So, yes, it's good that there is another MacBook Pro alternative, but I think the issue remains that it's unclear who exactly the buyer persona is going to be. I think a lot of people in the venn diagram are going to just go with a MacBook Pro.
This part is more speculation on my part, but if Apple brings out the M7 Pro/Max chips on the rumored 2027 schedule, I think they will really blowpast RTX Spark at a time when it will take Nvidia years to have an equivalent answer.
It will also be curious to see if AMD can catch back up Medusa Halo launching next year. If that is anything close to RTX Spark it will sidestep the Windows on ARM issue entirely and look like a pretty attractive alternative. Hopefully AMD actually gets it put into a wide selection of laptops this time around.
The strong case is pretty much only if you can't wait another 3-4 years for ram prices to come back down.
Memory makers are responding with a huge build-out. Building a fab shell and cleanroom takes years, which is why the crunch has lasted so long, but many of these projects are now well underway. New capacity is what ended past memory shortages. New factories going in:
SK hynix
M15X, Cheongju, Korea: first wafers early 2026 (reported), ramping through 2027
Yongin Y1, Korea: first cleanroom Feb 2027 (company target)
Yongin Y2, Korea: groundbreaking planned July 2027, first cleanroom June 2029
Micron
ID1, Boise: first wafer output mid-2027
Tongluo, Taiwan: an existing PSMC cleanroom it bought for $1.8B, with meaningful shipments mid-2027
ID2, Boise: late 2028
Hiroshima, Japan: broke ground July 2026, with initial output late 2028
Clay, New York (up to 4 fabs, $100B): broke ground Jan 2026, with first output in 2030
Samsung
P5, Pyeongtaek: mass production 2028 (reported)
Yongin national complex: site work began Oct 2026, with the first fab targeting Oct 2029 (reported; Samsung hasn't given a date)
Nanya
Fab 5A, New Taipei: 10nm-class EUV DRAM, with wafer starts in 2H 2027 (reported)
Tesla / SpaceX (Terafab)
Research fab at Giga Texas, Austin: broke ground April 2026. There's no official output date.
Grimes County, Texas (logic, memory, packaging; Phase 1 over $16.8B): land clearing since Aug 2026. Texas filings say commercial operations in 2028, but SemiAnalysis estimates late 2029.
The first real supply relief is 2027, a second wave comes in 2028–29, and a lot of that new capacity is aimed at HBM and server memory, so consumer DDR5 and GDDR will likely ease last.
Sources: Micron FQ4 2026 call, the SK hynix newsroom, TrendForce, Texas JETI filings via ENR, and Tom's Hardware/SemiAnalysis.
And maybe 2027 is a truly optimistic relief on some level,
> a second wave comes in 2028–29, and a lot of that new capacity is aimed at HBM and server memory, so consumer DDR5 and GDDR will likely ease last.
'Consumer DDR5' has a pretty big set of knock-on effects though.
Unless you are working at a 'tech' company, the norm is that even developer machines will be constrained on memory usage for a given workload[0][1]. IOW companies trying to keep procurement in check wind up looking at procurement cost vs lost productivity.
[0] - Modern infosec and corporate software, at least where I'm at, seems to be vaguely more efficient than electron apps, IOW I can reboot, open teams and my IDE, after 4GBish of my IDE and 1GB of a fresh Teams instance I'm already around 16 of 32GB.
[1] - I had to fight hard to get 32GB as a developer baseline at my org. They wanted 16GB. Ironically the JVM based devs get 64GB but still have slowdown issues on biuld...
> Unless you are working at a 'tech' company, the norm is that even developer machines will be constrained on memory usage for a given workload
Honestly I think most of us are just spoiled kids crying when our expensive toys are being taken away. If memory cost 10x more to manufacture than it does now, we’d just have 10x less ram in our computers and our software would be written to deal with that. We’ve done it before. My first computer had 4mb of ram and it ran windows 3.1 just fine. If ram were 10x cheaper before the supply shock, I think we’d be crying that companies won’t buy us 320gb ram in our laptops. The java and rust compilers would somehow be even less ram efficient. Firefox and chrome would still happily use it all to run adware across the same 50 tabs I want to keep open.
Not everything. Theres good reason for LLMs to need that much ram. And scientific computing. And .. well I can’t think of much else.
The supply crisis is annoying. But I’m also a little glad there’s a reason to push back on fat, lazy software.
Realistically, GOG probably is trying to accommodate for the OS running the game. So it is more for the spec of the PC, not the RAM the game itself needs
Well, so its not just bloated software that makes software bloated, but bloated facts, too. Users will see specs like that and go "well, I need to upgrade my machine from 2mb to 8gigs now, d'oh!" ...
I’m not surprised! The amount of wasted memory and compute in modern software is astounding.
If you’re a developer, try promoting Claude to profile and optimise a piece of some piece of software. I routinely see 20-40x improvements with a single prompt and a few hours of opus credits.
Thanks for your work by the way. At a large scale, it is much more efficient for some software developers to do a bit of belt tightening in their code than ask millions of customers to invest in bigger computers. Globally the savings are enormous.
Oh, the irony. Claude is itself an Electron app and lags horribly.
At this point, all these agent apps—Claude, Copilot, ChatGPT—are just calling remote APIs. Why not train them on the native toolkits for each OS (Win32/WinUI3, Cocoa/SwiftUI, GTK, Qt, Jetpack Compose), and make the experience better for users?
So do you think there will be a big crash and memory will get really cheap in 2030, or do you believe that AI demand is strong enough to sustain all this buildout?
It seems like this list also doesn't account for China and state sponsored memory industry (which will be more immune to a downturn)
Maybe I'm missing something but didn't micron exit the consumer DDR market for HBM? I don't see a rollback from that. When you say DDR will ease last, in what reality would memory makers go back to a market they left behind?
You could just code on a flight.. No need for llm. Id your brain is incapable of writing code without internet/llms you could write your specs/prompts for when you land.
The author’s discussion of their personal use case starts on the seventh sentence in TFA. And frankly, it sounds like a pretty solid use case:
“Being able to run things locally and mess around with RL, quantisation and fine-tuning on a laptop would be sick. I already publish my own GGUF quants on Hugging Face, with over 160,000 downloads in the past month.”
This is similar to "Why buy powerful gaming laptop when you can build a gaming PC and a cheap laptop to remote into it". That was my setup years ago, and there are a lot of nuance things i hadn't taken into account.
For example, network is not always available and stable; File transferring is limited by connection speed, hardware passthrough is basically nonexistence; Dealing with power management (Unexpected power loss) and so on.
I bet they all can be solved(to some bearable degree), but at that point, just bite a bullet and get a powerful laptop.
The difference is that streaming zero latency 4k video is incredibly difficult and always results in serious compromises. For AI programming, the end output is very small, not latency sensitive very easy to stream over the internet.
I'm sure you can come up with some scenario where you are programming in the wilderness with no internet connection but the market for that has to be in the tens of people.
That when civilization falls, and we have to go to techno agrarian farming, like in Interstellar, the person with the RTX isn't the one that's going to be left behind, it's the person with the cloud account that owns nothing.
I don't want a cloud login telemetry of when I am using my computer, an account that can be disabled by a provider, and constant ads to backup to onedrive. I also moved to linux because of this. (was already a linux guy, but Windows is completely in my rear-view now.)
Microsoft turns on OneDrive backups by default without your knowledge or consent, sending your private files to MS servers, not only during setup but it can also happen during updates. Even when you try keep it disabled, they constantly pressure you into enabling it every time you login, and make it really easy to accidentally turn it on. Once your files are in OneDrive, MS can access them. Don't be surprised if it later turns out they're training their LLMs on people's private files.
I haven’t experienced any of this on my win 11 sim racing rig. I wonder why? I was thinking then i remembered - I ran a script I found that removed copilot, and a lot of other shit. Can’t remember the name it was on GitHub. I only use that computer to run iracing and LMU and it’s been updating fine.
Shouldn’t have to do that of course, but it can be done.
It's less reliable because a single Windows update can undo all of that. More work because you have to know to look for this non-standard setup procedure and filter it from worse similar options.
I have a DGX Spark (Ubuntu), and when Microsoft and NVIDIA announced the RTX Spark running Windows, I thought to myself: Why would I get that instead of the DGX Spark?
The OS makes a big difference for most developers, and the DGX Spark stack and experience is mature.
I have a DGX Spark, a USB-C monitor, a Logitech wireless Touch keyboard/trackpad, and a very fat USB-C battery, in the same backpack alongside a cheap MacBook Air. This portable AI/ML rig is getting a lot of work done in various local environments.
It’s not that heavy a tote - well, it’s heavy. But it’s an acceptable tote given the immense amount of local work that can be done with it, pretty much anywhere in the globe.
(Haven’t got the solar panel thing worked out yet, but yacht is my preferred data center anyway, if you know what I mean..)
Being forced to login to use your own hardware is one issue for sure…
but separately everything about Microsoft’s login experience is pure pain.
Multiple broken redirects, broken links, buttons that loop, and more just to hit a CAPTCHA AI can solve anyway.
The only thing worse is for the poor soul developers who have had to login to their dev portal to setup SSO Microsoft for an app. It’s the worst I have ever seen in my life (even using Supabase that hand holds you through it).
The recovery image of my laptop is based on an old Win11 build where Microsoft login is now broken. So it falls back to local account after failing. Best bug!
I recently tried to recover to stock on the Zenbook A16 I returned and the WiFi wouldn't connect. Had to dig up an old USB Ethernet adapter to make it do the job.
Windows is such a bugs galore it's crazy how people use it as their daily drivers.
I just had a thing today where I wanted to play Forza Horizon 4 on my handheld.
When trying to launch the game I got a "Must sign in to Xbox/Microsoft to play". So I reluctantly complied and entered my password only to be met with "password login not available" after pressing enter.
not really, it would simply be the first of many headaches that just get worse as time goes on. Windows is a festering blight, it costs you more and more in bullshit you deal with the longer you have it.
For example, I cannot currentl use the ms suite, like word or excel, because it asks me to log into an institutional account that I dont have, despite my account being the owner of the licenses for the suite running on my laptop. Make that make sense. All because i opened an institutional word document on it, i've tried fully reinstalling, no dice. i've tried nuking caches and registry entries, no dice. Window's basic functionality is indistinguishable from maleware.
I've bought some low-priced things on amazon, just to have them cancelled. Spooky since I figure someone got my name, address, phone number from amazon.
Only 30 tok/s on Qwen 3.8 Flash Next? These days thanks to those specialized engines out there I can get ~2x that speed on a 24 GB video card with RAM offload. Thermal limits, I assume...
> When rumours started circulating about a laptop with over 100 GiB of RAM the GPU could access
That was like three years ago with the release of M3, do you people not count MacBooks as laptops or something? Not that you could run Linux on it back then, but you don't even mention the existence of Apple silicon at all. :/
That's like saying Valve just released the first ever standalone VR headset because the standalones that came before had advertising and trackers (or worse). Or that TVs aren't produced anymore because they all come with similar software. Or cars.
Saying RTX Spark is the first laptop with 128 GB of unified memory just because you don't agree with the previous laptops with 128 GB of unified memory is just incorrect. You could easily make it true with qualifiers but there were none here!
Also did nobody ever release a 128GB Strix Halo laptop? I remember that being a thing around when the Framework Desktop came out? And feels like that'd have even better Linux support than Nvidia?
Since all non-dedicated GPUs work this way I'd argue this was possible even longer ago. Now one has to question whether those integrated GPUs had any usable computing power. So it's not clear where to draw the line.
I can understand disregarding Apple hardware as they are not general computing devices since they support only a single, locked down OS.
> I can understand disregarding Apple hardware as they are not general computing devices since they support only a single, locked down OS.
Mac is absolutely a general computing device but macOS is the only software that fully supports it currently (Asahi Linux is still trying to get M3 Max and newer working)
Yes, you can freely surrender your data and control to a hostile corporation. What isn't clear is why someone would want to. What you gain in freedom, control, and privacy is worth the pittance of time required to set up.
> Doing the rationale thing to take advantage of the full capacity of my overpriced hardware
Why would you ever run anything else on equipment like that, which doesn't make any sense to me when enterprise grade hardware that is stationary would be more powerful anyway...because you aren't taking that on BART (I don't steal but seeing that thing in your lap, might bend that rule for myself since no real chance Bart PD would care anyway) or probably ever moving it from the place you use it now in ways that a mobile rack in your garage would be any less convenient for you.
Or if you felt strongly about consumer facing hardware, meaning really you pay through the nose to live in an efficiency in "SoMA" aka the Tenderloin, buy the mini-pc variant is just as easy to move but much less easy to break.
i see a lot of complaining about the login part of Win11, and i dont like it either, but i have not seen anyone mention that you can build an autounattend.xml file (generator here: https://schneegans.de/windows/unattend-generator/) that allows you to bypass all this. I have a USB Key with the file and the latest copy of the 11 Pro download, and when setting up a new machine, i run a fresh install. Then once booted, i run the developer config (https://github.com/microsoft/WindowsDeveloperConfig/) and it gets everything i need installed. No Microsoft Login, no Candy Crush, WSL installed and all the tools i need.
Someone please tell the author thank you for working on that. I hope it gets into normal Arch distribution ASAP.
I recently moved to CachyOS and have been hoping to eventually save up for some kind of laptop (or more likely the cheaper N1 when it comes out, my income is very unstable) like this. I just assumed I would be able to put Linux on it.
I see this new chip/setup as being an obvious technological upgrade and more cost efficient, even though the memory bandwidth is slightly disappointing. But I see it it as a superior and more advanced architecture with the memory sharing and much more efficient integrated chip and more efficient ARM.
To me this is just clearly the logical way to build efficient laptops and PCs now. I do hope that we don't have to wait that long to get something with better memory bandwidth though.
I will admit that the description of prefill times is a little disappointing. Someone please tell the author thank you for working on that. I hope it gets into normal Arch distribution ASAP.
115 comments
[ 0.31 ms ] story [ 9.8 ms ] threadAm I being too stubborn for not wanting a cloud account tied to my desktop operating system?
In fact if getting stalled by a single screen in the installer is a showstopper, then you're in for a world of hurt with Linux.
> Mine include the privacy concessions during setup, the years spent waiting for Microsoft to let people put the taskbar on the side of the screen again, and the advertising.
Of course you don't have to agree on all of them, but please don't assume everyone acts only because they can't find a tutorial on the net!
What challenge has been overcome?
I can see that hardware was bought, and that's cool hardware and all, but consumption is rather uninspiring.
I suppose it's the arm64 part? Or the review of some new hardware other people might be interested in?
Windows Pro > Work or School Account > Join a Domain instead > Set password & recovery questions
The one component here that is not present on most machines straight from the manufacturer is the presence of the Pro version. The home version of windows requires the account with increasingly limited alternatives.
https://archlinux.org/people/package-maintainers/#josie
This isn't a new trick - just the regular approach to editing windows install media.
Coding while flying is often the use case people bring up, but the power brick these laptops come with sometimes trip the plane's power plug in coach and you can't run local AI on battery for long. Also, over time planes will get better connectivity (Starlink etc.) and that particular full-offline scenario will go away.
For gaming, there are better options with dedicated GPUs like the Asus Zephyrus and similar.
What am I missing?
Aside from this, I’m led to believe that the AMD stack when it comes to local models has more overhead related to its need to translate more NVIDIA native instructions and therefore can be less performant than Nvidia. Not sure how true this is anymore, but ive yet to hear otherwise.
this discussion is filled with people who love high performance computing, and they are mostly saying "I wouldn't buy this". my comment was in response to a sharer of that sentiment. I picked onesuch comment since there is no way to reply to all of them, and pointed out "if you come at this from the other direction, you can see a population of people who will buy it.
>That doesn’t mean the people that buy this laptop fly business class
I didn't say it did,
10 months ago when I got my Spark it was competitive in price with a similarly spec'd out Mac Studio. As RAM prices have gone even more parabolic, so has its price. And so has the Mac's. And the Strix Halo competitors.
Good battery life is probably the biggest upside.
The downsides are numerous:
- Middling GPU performance compared to discrete RTX mobile GPUs. And yes, Nvidia markets them for gaming, so we need to judge them in that context, too. Windows on ARM isn’t amazing for gaming for various reasons including lack of anti-cheat compatibility.
- Compatibility and/or performance issues on Windows ARM. You’re running a lot of stuff non-native
- As you mentioned, the CPU cores are nothing to write home about with very mid single core performance.
- Suffers greatly from “why am I not buying a MacBook Pro?” syndrome?
I think there may be a case for being a good performance value for AI training, but for local inference I’m not seeing how that makes sense. Heck, for local inference I think it’s better to run a server and then use the laptop as a client.
You’ve also got projects like Strata that are going to define trends. Why bother with $6000 massive shared RAM machines when you can get 60 tok/s on Qwen 3.8 Flash Next on a 16GB RX 9060XT with 64GB of RAM?
Want to do local inference on a laptop? Grab a Zephyrus G14, get the 64GB spec, run Strata, then as a bonus you can run AAA games at full blast.
And if your use case is gaming the question becomes why you’re spending actual thousands of dollars more than a Zephyrus G14/G16 with an RTX 5080?
The Windows on ARM issue is really only one small piece of the puzzle of why these are a bad value proposition. They just don’t have a sensible buyer persona. There’s not a lot of people who will have a good reason to buy these over something else.
Local AI folks should likely pick a MacBook Pro M5 Max for the memory bandwidth and faster token generation, and for the Unix toolset and macOS focus that the AI and industry has over Windows.
Whoever is left is just someone who must have CUDA, but can’t get enough VRAM in an RTX 5070Ti/5080. If someone knows who that buyer is and what software they run I’d be curious to know what it is.
Even if your game is running natively and your anti-cheat is supported, you’d have gotten better performance on a Zephyrus G14/G16 for less money.
If you want to do local AI a MacBook Pro is either keeping up or going faster (faster memory bandwidth, token output writes faster), with faster single-core CPU performance as a bonus for general purpose snappiness.
In my eyes, the only buyer left is someone who can’t stand a MacBook Pro, or needs Nvidia/CUDA and a lot of VRAM.
To be more generous to the RTX Spark platform, it does seem to be priced head-to-head with Apple. I just think it has an uphill battle.
Although, the thing is, for the AI tasks advertised, the MacBook Pros have faster memory, so token decode runs faster.
Then by using a Mac you also scuttle the issue of "My platform has a low rate of adoption" that Windows on ARM has had. That issue will get better over time but in the present it's still an issue.
As for the dedicated GPU gaming laptop market, you'd be surprised how it's evolved since the release date of your MSI Ge66. I mentioned the 2026 Zephyrus G16 specifically because it is a laptop with very few compromises, including heat and noise. If you don't specifically need a huge amount of GPU-addressable RAM, and you're just looking for good performance in games and graphics-intensive applications, it is a clearly better buy than these RTX Spark systems. As a bonus, being an X86 system you'll have a much easier getting Linux going on that one.
So, yes, it's good that there is another MacBook Pro alternative, but I think the issue remains that it's unclear who exactly the buyer persona is going to be. I think a lot of people in the venn diagram are going to just go with a MacBook Pro.
This part is more speculation on my part, but if Apple brings out the M7 Pro/Max chips on the rumored 2027 schedule, I think they will really blowpast RTX Spark at a time when it will take Nvidia years to have an equivalent answer.
It will also be curious to see if AMD can catch back up Medusa Halo launching next year. If that is anything close to RTX Spark it will sidestep the Windows on ARM issue entirely and look like a pretty attractive alternative. Hopefully AMD actually gets it put into a wide selection of laptops this time around.
https://www.techpowerup.com/340216/amd-medusa-halo-apu-leak-...
Memory makers are responding with a huge build-out. Building a fab shell and cleanroom takes years, which is why the crunch has lasted so long, but many of these projects are now well underway. New capacity is what ended past memory shortages. New factories going in:
SK hynix
M15X, Cheongju, Korea: first wafers early 2026 (reported), ramping through 2027 Yongin Y1, Korea: first cleanroom Feb 2027 (company target) Yongin Y2, Korea: groundbreaking planned July 2027, first cleanroom June 2029 Micron
ID1, Boise: first wafer output mid-2027 Tongluo, Taiwan: an existing PSMC cleanroom it bought for $1.8B, with meaningful shipments mid-2027 ID2, Boise: late 2028 Hiroshima, Japan: broke ground July 2026, with initial output late 2028 Clay, New York (up to 4 fabs, $100B): broke ground Jan 2026, with first output in 2030 Samsung
P5, Pyeongtaek: mass production 2028 (reported) Yongin national complex: site work began Oct 2026, with the first fab targeting Oct 2029 (reported; Samsung hasn't given a date) Nanya
Fab 5A, New Taipei: 10nm-class EUV DRAM, with wafer starts in 2H 2027 (reported) Tesla / SpaceX (Terafab)
Research fab at Giga Texas, Austin: broke ground April 2026. There's no official output date. Grimes County, Texas (logic, memory, packaging; Phase 1 over $16.8B): land clearing since Aug 2026. Texas filings say commercial operations in 2028, but SemiAnalysis estimates late 2029.
The first real supply relief is 2027, a second wave comes in 2028–29, and a lot of that new capacity is aimed at HBM and server memory, so consumer DDR5 and GDDR will likely ease last.
Sources: Micron FQ4 2026 call, the SK hynix newsroom, TrendForce, Texas JETI filings via ENR, and Tom's Hardware/SemiAnalysis.
It's been like a year-ish...
And maybe 2027 is a truly optimistic relief on some level,
> a second wave comes in 2028–29, and a lot of that new capacity is aimed at HBM and server memory, so consumer DDR5 and GDDR will likely ease last.
'Consumer DDR5' has a pretty big set of knock-on effects though.
Unless you are working at a 'tech' company, the norm is that even developer machines will be constrained on memory usage for a given workload[0][1]. IOW companies trying to keep procurement in check wind up looking at procurement cost vs lost productivity.
[0] - Modern infosec and corporate software, at least where I'm at, seems to be vaguely more efficient than electron apps, IOW I can reboot, open teams and my IDE, after 4GBish of my IDE and 1GB of a fresh Teams instance I'm already around 16 of 32GB.
[1] - I had to fight hard to get 32GB as a developer baseline at my org. They wanted 16GB. Ironically the JVM based devs get 64GB but still have slowdown issues on biuld...
Honestly I think most of us are just spoiled kids crying when our expensive toys are being taken away. If memory cost 10x more to manufacture than it does now, we’d just have 10x less ram in our computers and our software would be written to deal with that. We’ve done it before. My first computer had 4mb of ram and it ran windows 3.1 just fine. If ram were 10x cheaper before the supply shock, I think we’d be crying that companies won’t buy us 320gb ram in our laptops. The java and rust compilers would somehow be even less ram efficient. Firefox and chrome would still happily use it all to run adware across the same 50 tabs I want to keep open.
Not everything. Theres good reason for LLMs to need that much ram. And scientific computing. And .. well I can’t think of much else.
The supply crisis is annoying. But I’m also a little glad there’s a reason to push back on fat, lazy software.
You can buy a version from gog that runs on Windows 10 requiring 2 GB of RAM and an unspecified 1.8GHz CPU [2]
[1] https://www.mocagh.org/origin/u7completeaus-manual.pdf
[2] https://www.gog.com/en/game/ultima_7_complete
If you’re a developer, try promoting Claude to profile and optimise a piece of some piece of software. I routinely see 20-40x improvements with a single prompt and a few hours of opus credits.
Thanks for your work by the way. At a large scale, it is much more efficient for some software developers to do a bit of belt tightening in their code than ask millions of customers to invest in bigger computers. Globally the savings are enormous.
Oh, the irony. Claude is itself an Electron app and lags horribly.
At this point, all these agent apps—Claude, Copilot, ChatGPT—are just calling remote APIs. Why not train them on the native toolkits for each OS (Win32/WinUI3, Cocoa/SwiftUI, GTK, Qt, Jetpack Compose), and make the experience better for users?
Yes, this, 100x this.
Native-everything should just return to being the norm now.
End Cyclomatic Complexity!
It seems like this list also doesn't account for China and state sponsored memory industry (which will be more immune to a downturn)
Tailscale is the key ingredient.
The author’s discussion of their personal use case starts on the seventh sentence in TFA. And frankly, it sounds like a pretty solid use case:
“Being able to run things locally and mess around with RL, quantisation and fine-tuning on a laptop would be sick. I already publish my own GGUF quants on Hugging Face, with over 160,000 downloads in the past month.”
For example, network is not always available and stable; File transferring is limited by connection speed, hardware passthrough is basically nonexistence; Dealing with power management (Unexpected power loss) and so on.
I bet they all can be solved(to some bearable degree), but at that point, just bite a bullet and get a powerful laptop.
I'm sure you can come up with some scenario where you are programming in the wilderness with no internet connection but the market for that has to be in the tens of people.
Also that means you're forced into a manditory arbitration clase and a class action waver.
Shouldn’t have to do that of course, but it can be done.
Not defending MS btw.
The OS makes a big difference for most developers, and the DGX Spark stack and experience is mature.
It’s not that heavy a tote - well, it’s heavy. But it’s an acceptable tote given the immense amount of local work that can be done with it, pretty much anywhere in the globe.
(Haven’t got the solar panel thing worked out yet, but yacht is my preferred data center anyway, if you know what I mean..)
but separately everything about Microsoft’s login experience is pure pain.
Multiple broken redirects, broken links, buttons that loop, and more just to hit a CAPTCHA AI can solve anyway.
The only thing worse is for the poor soul developers who have had to login to their dev portal to setup SSO Microsoft for an app. It’s the worst I have ever seen in my life (even using Supabase that hand holds you through it).
Windows is such a bugs galore it's crazy how people use it as their daily drivers.
When trying to launch the game I got a "Must sign in to Xbox/Microsoft to play". So I reluctantly complied and entered my password only to be met with "password login not available" after pressing enter.
Great... guess I just won't play then.
For example, I cannot currentl use the ms suite, like word or excel, because it asks me to log into an institutional account that I dont have, despite my account being the owner of the licenses for the suite running on my laptop. Make that make sense. All because i opened an institutional word document on it, i've tried fully reinstalling, no dice. i've tried nuking caches and registry entries, no dice. Window's basic functionality is indistinguishable from maleware.
I've bought some low-priced things on amazon, just to have them cancelled. Spooky since I figure someone got my name, address, phone number from amazon.
I wonder how its performance is on dense models.
That was like three years ago with the release of M3, do you people not count MacBooks as laptops or something? Not that you could run Linux on it back then, but you don't even mention the existence of Apple silicon at all. :/
Given the reasons she lists for not wanting Windows, I can't imagine her entering Apple's closed garden.
Saying RTX Spark is the first laptop with 128 GB of unified memory just because you don't agree with the previous laptops with 128 GB of unified memory is just incorrect. You could easily make it true with qualifiers but there were none here!
Also did nobody ever release a 128GB Strix Halo laptop? I remember that being a thing around when the Framework Desktop came out? And feels like that'd have even better Linux support than Nvidia?
I can understand disregarding Apple hardware as they are not general computing devices since they support only a single, locked down OS.
Mac is absolutely a general computing device but macOS is the only software that fully supports it currently (Asahi Linux is still trying to get M3 Max and newer working)
Why would you ever run anything else on equipment like that, which doesn't make any sense to me when enterprise grade hardware that is stationary would be more powerful anyway...because you aren't taking that on BART (I don't steal but seeing that thing in your lap, might bend that rule for myself since no real chance Bart PD would care anyway) or probably ever moving it from the place you use it now in ways that a mobile rack in your garage would be any less convenient for you.
Or if you felt strongly about consumer facing hardware, meaning really you pay through the nose to live in an efficiency in "SoMA" aka the Tenderloin, buy the mini-pc variant is just as easy to move but much less easy to break.
There are already several OEMs shipping RTX Spark mini PCs with NVIDIA Linux distribution (NVIDIA DGX OS), support them, like Asus.
Instead the message keeps being the same, no value in selling pre-installed Linux devices.
I recently moved to CachyOS and have been hoping to eventually save up for some kind of laptop (or more likely the cheaper N1 when it comes out, my income is very unstable) like this. I just assumed I would be able to put Linux on it.
I see this new chip/setup as being an obvious technological upgrade and more cost efficient, even though the memory bandwidth is slightly disappointing. But I see it it as a superior and more advanced architecture with the memory sharing and much more efficient integrated chip and more efficient ARM.
To me this is just clearly the logical way to build efficient laptops and PCs now. I do hope that we don't have to wait that long to get something with better memory bandwidth though.
I will admit that the description of prefill times is a little disappointing. Someone please tell the author thank you for working on that. I hope it gets into normal Arch distribution ASAP.