Asking people to wait for a device society has normalized slamming shut and tossing in a bag the millisecond a clock reads you're done for the day

Yeah, if it wasn't clear from my previous comments, putting liquid metal on laptops is a crazy thing to do, but if it gives a boost on benchmarks or a slightly thinner laptop or both, manufacturers of gaming laptops will go for it.

is like developing a new car tire that requires drivers to slow down to 10MPH for the last mile of driving to keep the tires from exploding in the garage.

The Dodge Demon was such a product. It comes with "cheater slicks" from the factory, which are basically slicks that have just the minimum amount of tread pattern mandated by law (which isn't much) and with a compound that's unsafe for driving on the wet, with warnings from the manufacturer not to drive the car on the wet unless you absolutely have to, and if so, only at the lowest possible speeds, 10mph or less. So yes, if rain happens during the last mile drive to your garage and you are rich enough to own such a car, you have to slow down to 10MPH for the last mile. Impractical designs exist in the automotive world too. The problem with gaming laptops is that liquid metal is appearing on mainstream models because it's cheap and gives a massive performance boost, and laptop manufacturers don't care about how risky it is for daily use because, well, they don't care about longevity.

Most people would just run Windows Update troubleshooter to "reset" Windows Update if it took hours during every update, instead of throwing an entire system away...

Yeah obviously performance is gonna be capped in battery mode. Also, 99Wh batteries exist.

As an aside, the battery won't explode even if there is liquid metal leakage in laptops that have liquid metal as their thermal interface material, lithium batteries have overcurrent protectors that protect from short-circuits happening outside the battery enclosure (which is plastic and doesn't amagalmate with liquid metal), what happens is that the liquid metal ends up all over the board and makes the board almost unrepairable: https://www.youtube.com/watch?...

This is like watching McDonalds advertise their fresh hot coffee by showing a copy of the lawsuit they settled after burning the shit out of a customer.

If this happens, you were holding it wrong.

or is the term "melt down" about to get really explosive being near all that lithium ion

If this happens, you were holding it wrong.

Ok, jokes aside, liquid metal is a popular thermal interface in high-performance yet thin laptops, and it comes with warnings to let the laptop cool down before you put it in a vertical orientation (for example, in a backpack or briefcase/laptop bag) so the liquid metal has a chance to cool down and become less viscous and to not flex the chassis because that may cause the liquid metal to leak out of the seal covering the heatsink.

But again, we don't know whether the Surface Laptop Ultra uses liquid metal thermal interface material. With the 80W an RTX Spark N1X pulls, it's possible to have sustained full power on a thin laptop with just a well-designed cooling solution and somewhat aggressive fan curves.

Knowing 21st Century marketing, they did achieve 1 petaflop. Measured for approximately 0.18 seconds. Right before the entire laptop exploded, with performance experts standing on the other side of 6" of bulletproof plexiglass. Not that they expected that result or anything. We're told the Guinness adjudicator was just there for show.

Again, 80W is not that much even for a thin laptop. There are thin gaming laptops pulling twice as much. Depending on the choices made in the cooling system, you can easily run an 80W SoC at full speed sustained.

The TDP of an RTX Spark N1X SoC is 80W, so that's 0.00000008 gigawatts. Plus a bit more if you use the screen and play music on the speakers.

Again, it's an RTX Spark N1X SoC, the TDP is 80W. There are gaming laptops consuming two or even three times that, and they last 2 hours under full load.

It's an RTX Spark N1X, the TDP is 80W.

FYI their x86-64 Surface offerings are standard PCs, there is nothing "magical" about them: https://www.makeuseof.com/put-... So, if you like their hardware, you aren't stuck with Windows. Their ARM offerings should be assumed to be stuck with Windows on ARM, much like ARM boards intended to run Linux should be assumed to be stuck with the Linux images the manufacturer provided. Ahh... I can feel all that ARM openness hitting me in the face...

Wait? 80W is a lot for you guys?

/hides away 400W Area-51m that came with two power supplies that must be used simultaneously

All I want is excellent compatibility with my existing x86-32 and x86-64 software, aka in-hardware backward compatibility (when it comes to CPU instruction set architecture), which is something you don't get from ARM CPUs and the various OS-provided emulators and recompilers. If a developer goes as far as to recompile their software to take advantage of the new instructions in Nova Lake or whatever, so much the better, but it's not a necessity. It's been that way since the first Pentium introduced some nifty new instructions not found in the 486.

That said, the RTX Spark SoC is heavily tailored towards users running local LLMs (don't ask what they do with them, not a clue), compatibility with existing Windows software isn't a priority here.

FYI the 1 petaflop is in the GPU, not in the CPU. The SoC is an RTX Spark N1X, so the CPU is 10+10 or 9+9 ARM cores (Cortex-X925+Cortex-A725), depending on the specific variant. And Windows Update uses the CPU to put its KBs in a row and install them, not the GPU.

That said, if your Windows installation takes hours to Windows Update every month (not just after a clean install), something is wrong with your Windows installation.

That one petaflop spec is going to last a tenths of a second I expect.

You can't know that yet. Yes, this obsession with thinness is unhealthy, but you don't know how aggressively the fans are programmed to ramp up and whether a liquid metal thermal interface was used or not. Some OEMs can get surprising thermals out of thin laptops using aggressive fan curves and liquid metal (at the expense of the liquid metal sometimes leaking onto the board and amalgamating with the solder joints).

Also, all modern chips throttle, it's a feature called "boost", but you can't know whether the promised 1 petaflop is boost or sustained because that wasn't disclosed, you have to wait for someone to test it.

Again, this Blackwell GPU is not the kind of Blackwell GPU meant to go into a data center (or even a large desktop workstation), much like not all RTX 50-series GPUs have the same cluster config or TDP. This is a chip intended for LLM models that run locally on a laptop. Now, don't ask what people are doing with local LLMs, but the Mac Studio is selling very well to the local LLM crowd, so there is a market for them, and it's reasonable to assume that some of that crowd will want a machine that is mobile.

Just because it's a Blackwell-architecture GPU, it doesn't mean it's the same chip as the Blackwell-architecture GPUs that go into data centers. GPUs are highly parallel, so they can scale TDP-wise depending on how many clusters of unified shaders you have and the clocks you run them at. This is not new, for example, the Maxwell architecture had chips ranging from 30W TDP (GeForce 830M and 930M) to 250W beasts like the GeForce GTX TITAN X.