[HN Gopher] We built a cloud GPU notebook that boots in seconds
___________________________________________________________________
We built a cloud GPU notebook that boots in seconds
Author : birdculture
Score : 82 points
Date : 2025-11-03 00:50 UTC (4 days ago)
(HTM) web link (modal.com)
(TXT) w3m dump (modal.com)
| zkmon wrote:
| When did booting time has become a problem to solve?
| mcemilg wrote:
| From my (just a user) perspective, GPUs are expensive, they
| shouldn't be left standing if they're not being used.
| jack_tripper wrote:
| I was there Gandalf, 3000 years ago, when people used
| folding@home to donate idle CPUs.
| krige wrote:
| and SETI@home too!
| embedding-shape wrote:
| > From my (just a user) perspective, GPUs are expensive, they
| shouldn't be left standing if they're not being used.
|
| How much does a idling GPU actually take when there is no
| monitor attached and no activity on it? My monitor turns off
| after 10 minutes of inactivity or something, and at that
| point, I feel like the power draw should be really small (but
| haven't verified it myself).
| ukblewis wrote:
| I honestly can't believe that the only top level comment right
| now is this kind of "I can't be assed to read the linked
| article, I came just to shit on it" kinda comment
| hrimfaxi wrote:
| When did low-effort comments become acceptable here?
| doctorpangloss wrote:
| The problem to solve is low cost, secure multitenancy.
| ukblewis wrote:
| This looks awesome!
| binaryturtle wrote:
| I remember visiting a computer exhibition (CeBIT) in the very
| early 90s. In one booth they had some of the big Amiga systems
| (2000, I think) and at some point on of the booth's staff did the
| 3 finger salute (press 3 specific keys on the keyboard to force a
| reboot) on one of the machines. The machine was back up in what
| felt like an instant. I was amazed by that. They probably had
| setup the whole boot process via RAM (see "RAD" disk on the
| Amiga), but I hadn't any idea about that back in the days.
|
| Still to this day I think this is how it should be. You want to
| switch ON your computer and it should be ready for use.
|
| But what do we get? What feels like minutes of random waiting
| time. My Raspberry PI with Linux which probably eats 10 of those
| Amiga 2Ks for breakfast shifts through through a few 1000 lines
| of initialising output... my Mac which probably eats like 50 of
| those Amiga 2Ks for lunch... showing a slowly growing bar doing
| whatever... Why didn't this improve at all in the last 30 years?
| vachina wrote:
| Because we still carry data over coppers and wires.
| davemp wrote:
| Windows prioritize phoning home and data collection over UX. If
| you have a corporate install you'll also have negligent EDP
| software killing your boot times.
|
| You can get fast boot times on linux if you care to tweak
| things.
| embedding-shape wrote:
| > They probably had setup the whole boot process via RAM (see
| "RAD" disk on the Amiga), but I hadn't any idea about that back
| in the days.
|
| > Still to this day I think this is how it should be. You want
| to switch ON your computer and it should be ready for use.
|
| Don't we already kind of have this? It's setup to be dynamic,
| and we'd ended up calling it "sleep", but it basically does
| what you're talking about, but dynamically and optionally,
| basically chucking the entire state into RAM (or disk for
| "hibernate") then resumes from that when you wanna continue.
|
| Personally I've avoided it for the longest of times because
| something always breaks or ends up wonky when you resumes, at
| least on my desktop. The PS5 and the Steam Deck handles this
| seemingly even with games running, so seems possible, and I
| know others who are using it, maybe Linux desktop is just
| lagging behind there a bit so I continue to properly shut down
| my computer every night.
| sheepscreek wrote:
| Macs on the other hand are extremely stable. In my 4 years of
| using my MacBook Pro M1 Max, I've only restarted during OS
| updates. There were maybe a handful instances where it froze
| and I forced restart. Other than that, I only put it to sleep
| every time and it works like a charm. I use it for heavy duty
| software development and experimentation with local models,
| so it's even more surprising!
| embedding-shape wrote:
| The hardware Apple makes is incredible, bar none, which is
| why is such a shame the OS and application UX is absolutely
| horrible and continues to get worse with each iteration. If
| Apple would publicly support Linux efforts on Apple
| hardware I'd probably switch back in an instant. But until
| then, I guess I'll continue turning off my desktop at
| night, and waiting a whole of 15 seconds for the startup in
| the morning _oh the horrors_.
| __mharrison__ wrote:
| Really? I tend to reboot a lot. OBS, monitors, USB hub all
| trend to flake out after a few days of sleeping.
| vel0city wrote:
| I'm using an M4 Macbook right now and I _constantly_ have
| issues with USB devices (especially hubs) failing to work
| properly after sleep. Its very unpredictable too, I can 't
| seem to make it happen.
|
| Its actually kind of funny, because while people talk about
| how unreliable Bluetooth is, moving a few of those devices
| from USB to Bluetooth (like my trackball mouse) made the
| situation _far more reliable_. Sleep has been _that bad_.
| Grazester wrote:
| I have used Windows hibernate since Windows XP and never had
| an issue with devices after resuming Windows. Within recent
| years on Windows 10 I have gone months without a restart,
| only hibernating my pc. In the early days I used a custom
| built pc. In the later years(post 2005) I have only used
| laptops, mostly Dells with a sprinkling of Lenovos; if that
| matters.
|
| I don't know why Windows now hides it from the power menu by
| default now.
| embedding-shape wrote:
| I think it's mostly us with lots of external gear (mostly
| audio related) that things get a bit wonky, and if you're
| running graphic-heavy applications that you're trying to
| resume at the same time. For example, Ableton for the
| longest of times couldn't handle resuming from hibernation
| for me, seems to work today (Windows 11), but still having
| the same issue with a running Houdini window, resuming from
| hibernation does something with the communication with the
| GPU (my hunch) and the window freezes when resuming.
| threeducks wrote:
| It's the OS. About 10 years ago, I had an Asus EeePC, which was
| an underpowered piece of trash with a 32 bit Intel Atom CPU,
| but it cold-booted in less than 3 seconds. And by "booted", I
| mean completely booted, i.e. not like Android, where you have
| to wait a few more minutes until all the background services
| settle and the UI stops lagging.
|
| Unfortunately, the MeGoo OS was discontinued shortly after.
| https://en.wikipedia.org/wiki/MeeGo
| dagmx wrote:
| Because a modern OS is much higher fidelity and there's limits
| to how fast all those components can load.
|
| You may not care about the newer features , or think you don't
| at the least, but there's a limit to how fast they can be
| loaded.
|
| More than just loaded, they're also often checked for integrity
| as well.
| darthShadow wrote:
| Just curious, was something like
| https://github.com/containerd/stargz-snapshotter
| considered/evaluated before designing your own lazily-loaded
| container FS and if so, any pros/cons for the same?
| hhthrowaway1230 wrote:
| Also curious! I was also wondering if criu frozen containers
| would help here. I.e. load the notebooks, snapshot them, and
| then restore them.
| amitprasad wrote:
| This is notoriously hard when you start to involve GPUs
| amelius wrote:
| Seconds ... how many?
|
| I remember reading that if a webpage takes more than 4 seconds to
| load, 50% of users will have closed the page.
| stronglikedan wrote:
| > Seconds ... how many?
|
| Right? Any process that eventually completes successfully takes
| seconds, even if it's a million of them.
| htrp wrote:
| is this the fruits of the jamsocket deal?
___________________________________________________________________
(page generated 2025-11-07 23:02 UTC)