In other news today...



  • 138364e8-13b9-4226-953c-44cfdbb3b6f6-image.png



  • Wasn't aware they were still going. If they are still sailing, I think you can join them on one of the their legs as a deckhand, doing deckhand stuff.



  • @cvi I don't actually know how current that is. (Came up in my FB feed.) I just picked this as a good posting place.

    https://www.gotheborg.se/news/rescue-of-sailing-boat/


  • ♿ (Parody)

    @boomzilla said in In other news today...:

    @dkf said in In other news today...:

    @topspin said in In other news today...:

    Expect the next one soon:

    I was reading the paper; it appears to be a relatively simple material to make in the lab (if you have a temperature-controlled furnace and a decent vacuum system) as it's starting materials aren't very exotic. Expect either a reproduction of the experiment or a bunch of denials that it could possibly be true soon.

    It's definitely not an AI-written paper. The language use is slightly too awkward in places. 😉

    Yeah, here's everyone's favorite chemist's take:

    https://www.science.org/content/blog-post/breaking-superconductor-news

    Update:

    https://www.science.org/content/blog-post/room-temperature-superconductor-new-developments

    tl;dr

    Conclusion

    I am guardedly optimistic at this point. The Shenyang and Lawrence Berkeley calculations are very positive developments, and take this well out of the cold-fusion "we can offer no explanation" territory. Not that there's anything wrong with new physics (!), but it sets a much, much higher bar if you have to invoke something in that range. I await more replication data, and with more than just social media videos backing them up. This is by far the most believable shot at room-temperature-and-pressure superconductivity the world has seen so far, and the coming days and weeks are going to be extremely damned interesting.


  • 🚽 Regular

    To quote the top comment on the video about this in Sabine Hossenfelder's channel:

    The catch is that the temperature of the room needs to be near absolute zero.





  • @jinpa said in In other news today...:

    Am dissapoint.
    Did not point at several real reasons to why modern OSes are slow as fuck, but instead handwaving about malware.
    Software of today is slow, with the possible exception of some simulations and games, because "programmer time is more expensive than hardware", and there being no real lower bound for skill in the profession.
    Among other things.


  • Notification Spam Recipient

    @jinpa This reads AI generated.



  • @Carnage said in In other news today...:

    @jinpa said in In other news today...:

    Am dissapoint.
    Did not point at several real reasons to why modern OSes are slow as fuck, but instead handwaving about malware.
    Software of today is slow, with the possible exception of some simulations and games, because "programmer time is more expensive than hardware", and there being no real lower bound for skill in the profession.
    Among other things.

    Didn't mention malware. Source post this was cribbed from put some of it down to the choices of software stack (e.g. Electron) and companies wanting to save their own time-to-market with multiple target platforms without caring about their users' time in anything except the most performance-critical situations (e.g. gaming).



  • @Watson said in In other news today...:

    without caring about their users' time in anything except the most performance-critical situations (e.g. gaming).

    If they really cared about my time, they wouldn't write games. They steal so much of my time...



  • @jinpa :hanzo:
    Even in this same thread...

    @DogsB said in In other news today...:

    Tsaukpaetra amateur hour.


  • Notification Spam Recipient

    @BernieTheBernie said in In other news today...:

    @jinpa :hanzo:
    Even in this same thread...

    @DogsB said in In other news today...:

    Tsaukpaetra amateur hour.

    IMG_1486.jpeg



  • @DogsB We cannot repeat it often enough: Windows XP is great, if you are not an amateur!



  • @Tsaukpaetra said in In other news today...:

    @jinpa This reads AI generated.

    I was thinking the same while skimming it. Even if it isn't, the article is pretty crap and full of shoddy writing.

    Edit: Plus what @Carnage said - the article doesn't actually say why modern PCs "may feel" slow.



  • @cvi even if you strip out the app bloat nonsense, there’s still the likes of the indexing services and telemetry that never used to be in the older versions and these do consume resources too.

    Modern software does a lot more than old software and that rarely gets brought up.



  • @Arantor said in In other news today...:

    Modern software does a lot more than old software and that rarely gets brought up.

    The two examples you make aren't great, though. I can do without the telemetry and it seems to be more for the benefit of ${corpo}. Any benefit it may bring to me is rather intangible and far in the future. It better be unnoticeable in terms of performance.

    Same for indexing on Windows, really. I've not used their search for a while, because it's been so useless in recent versions. It's slow, it includes a lot of garbage (at least by default - I don't remember if I ever managed to turn off the online results).

    I think I might conclude that I want most tools to do less (but better). For example, [s]locate only searches file paths, but I don't remember when I last saw it take cycles for indexing and searching for a file path match takes less than a second across multiple filesystems and drives. (And best of all - no Bing results or ads in the output!)



  • @cvi What you need is SSDS.



  • @jinpa no.pnm

    OTOH, I doubt it has Bing search ads in it, so there's at least that.



  • @cvi said in In other news today...:

    no.pnm

    No Public Service Company of New Mexico?


  • Discourse touched me in a no-no place

    @Carnage said in In other news today...:

    Software of today is slow, with the possible exception of some simulations and games, because "programmer time is more expensive than hardware", and there being no real lower bound for skill in the profession.

    The core of why much software feels slow is almost certainly due to far too many copies being done by Schlemieled algorithms. Copies have real costs, and a lot of programmers have no idea what to do about avoiding them (or aren't even really aware that they are going on).

    Having more resolution and HDR and so on also adds cost to GUIs by making the sheer quantity of information that needs to be manipulated much larger. You want fancy super-scalable emoji fonts with lots of detail? You gonna pay for it.




  • BINNED

    @Arantor what @cvi said. It does a lot more, but with worse results than not doing that shit at all. At least in the case of Windows.



  • @dkf said in In other news today...:

    Having more resolution and HDR and so on also adds cost to GUIs by making the sheer quantity of information that needs to be manipulated much larger. You want fancy super-scalable emoji fonts with lots of detail? You gonna pay for it.

    IMO, this pretty much loops back to your first statement on too many copies and crappy algorithms. Chances are pretty high that if you can drive a high resolution screen with HDR, you have the hardware to render it. But doing it successfully may require you to use the hardware properly (and not do dumb stuff).


  • BINNED

    @dkf eh, the resolution is, what, ten times as high and the capabilities of a modern GPU are a bajillion times higher as back then. No valid excuse.

    Also, games aside I don't think terrible framerate is why so much feels slow. It's all the other stuff.



  • @Arantor said in In other news today...:

    @cvi even if you strip out the app bloat nonsense, there’s still the likes of the indexing services and telemetry that never used to be in the older versions and these do consume resources too.

    Modern software does a lot more than old software and that rarely gets brought up.

    So it's the same answer we've known for decades now. In the 90s we called it Wirth's Law or "Andy giveth and Bill taketh away." In the 2000s, Wirth's Law got restated as Page's law. And apparently people are still rediscovering it today!



  • @topspin said in In other news today...:

    @dkf eh, the resolution is, what, ten times as high and the capabilities of a modern GPU are a bajillion times higher as back then. No valid excuse.

    Non-garbage display resolution starts at VGA (640x480, 256 colors). My screen today runs at 1920x1080, 24 bits per pixel (but the GPU does everything in RGBA so it's really 32), and I have two of them.

    VGA: 640x480x1 = 307,200
    Modern: 1920x1080x4x2 = 16,588,800

    So that's 54x as a baseline. But it gets a lot worse because ever since Vista, we've had a compositing window manager, so now we're not just drawing all the pixels to the screen; we're drawing all the pixels for all the applications whether they're on-screen or not, in order to keep the system responsive. Let's add another order of magnitude there, just as a conservative estimate.

    Meanwhile, that "bajillion times higher GPU capacity" is geared almost entirely towards high-end 3D performance. As someone who's worked in this area, support for simple, flat 2D graphics has actually gotten notably worse, not better, over the years. You're actually not even able to directly render a simple rectangle ("quad") on modern graphics APIs at all anymore; you need to turn it into a pair of triangles first so the 3D API can understand what you're talking about! That's a textbook abstraction inversion, with all the performance costs you'd imagine. (Rendering two triangles requires 6 vertices with all their associated information, compared to only 4 for a single quad, just for starters, so add on another conservative estimate of a 50% performance penalty there.)

    There are a lot of little details like that that add up quickly as the various penalties stack multiplicatively...


  • Notification Spam Recipient

    @jinpa said in In other news today...:

    I have conflicting thoughts about this. In some ways faster tech made a lot more things viable. Git would have been a nonstarter if SSDs didn't take off. Cheaper and faster ram for caching everything made most IDEs and most browsers useful again.

    Then you have weird things like slack and teams. Why do they take two minutes to load. Why are they so slow and unresponsive? Why do they spike to 100% every now and then when I'm not doing anything with them? You could blame the webby thing underneath but vs code is actually pretty solid and probably the leading example of that tech working well.

    It's just bad developers. Browsers have to be the worst. They actually have gotten better but they're obviously not doing much QA because they break constantly, do odd things and like to eat ram. I still don't understand why six tabs needs a gig.

    One of the problems is probably that developers work on fairly good hardware and don't test on anything that hasn't been released in the last five or so years. Hate to break it to them but: most users are running hardware that is well past its sell by date. Microsoft is among the worse for this. Not so gentle nudges to update to the latest and greatest even on hardware that can barely run 10.

    Everyone is moving to their phone because it's a much better experience. The cheap end of android is terrible but you move into the 200+ bracket, you'll have a ui experience that's better than most desktops in the last 10 years. Even microsoft haven't fucked up badly on android yet.

    Speaking of, Windows 11 on their own surface tablets are a god awful experience. (Yes this again). I moved to Debian/Gnome recently and despite the ballache getting it to run on the surface the user experience is far better. But some of that is probably down to one of linux's main selling points been: It breathes life into old hardware.

    I have no idea what the devil Microsoft is up too but they need to stop it.



  • @Mason_Wheeler said in In other news today...:

    (Rendering two triangles requires 6 vertices with all their associated information, compared to only 4 for a single quad, just for starters, so add on another conservative estimate of a 50% performance penalty there.)

    This is not an useful estimate/statement.

    • The fill rate is the same, after all you're covering the same number of pixels(*).
    • You're only storing and transforming 4 points (indexed mesh, or triangle strip/fan). If you're 2D, they'll be 2D coordinates, any you can choose to use 16 bits/coordinate if that's your thing.

    Besides even a garbage GPU these days can easily push a few 100k triangles at 60 Hz. (With a mediocre GPU, you can probably push more triangles than you have pixels.) Fill rate is more likely to be the limit.

    Additionally, you do have support for stuff like blitting rectangles. If that's a bottle-neck for you (not likely), you have options (all the way to compute shaders, which are now very standardized).

    Finally, with DDR4 = 20GB/s - 50GB/s, you can fill that amount of memory (~16MB) about a thousand times per second from the CPU at the lower end.

    (*) There is a small overhead along the diagonal due to raster stamp size. You can avoid that easily in some cases (full screen), and if it's a problem, you can work around it in the general case too (e.g., single triangle + clip, not sure if that will be faster).

    Edit: Above estimate ignores caches. Ryzen 5600 has 32 MB L3$, so it basically fits the whole framebuffer (2xfull HD) into it. With a bit of cleverness (e.g., tiling), you can make parts fit into L2 or even L1 while they're being drawn.


  • BINNED

    @Mason_Wheeler said in In other news today...:

    @topspin said in In other news today...:

    @dkf eh, the resolution is, what, ten times as high and the capabilities of a modern GPU are a bajillion times higher as back then. No valid excuse.

    Non-garbage display resolution starts at VGA (640x480, 256 colors). My screen today runs at 1920x1080, 24 bits per pixel (but the GPU does everything in RGBA so it's really 32), and I have two of them.

    VGA: 640x480x1 = 307,200
    Modern: 1920x1080x4x2 = 16,588,800

    So that's 54x as a baseline. But it gets a lot worse because ever since Vista, we've had a compositing window manager, so now we're not just drawing all the pixels to the screen; we're drawing all the pixels for all the applications whether they're on-screen or not, in order to keep the system responsive. Let's add another order of magnitude there, just as a conservative estimate.

    And yet games manage to push an insane number of graphics primitives at 60 fps while the Windows start menu takes between ~0s (as it should be) and ~5s just to appear, and then all the ads aren't even loaded yet. So, really, it's not that.

    Meanwhile, that "bajillion times higher GPU capacity" is geared almost entirely towards high-end 3D performance. As someone who's worked in this area, support for simple, flat 2D graphics has actually gotten notably worse, not better, over the years. You're actually not even able to directly render a simple rectangle ("quad") on modern graphics APIs at all anymore; you need to turn it into a pair of triangles first so the 3D API can understand what you're talking about! That's a textbook abstraction inversion, with all the performance costs you'd imagine. (Rendering two triangles requires 6 vertices with all their associated information, compared to only 4 for a single quad, just for starters, so add on another conservative estimate of a 50% performance penalty there.)

    Your GPU really doesn't care if you blit a window-sized texture using 4 or 6 verts, but anyway, you don't need 6 either. Considering the actual pixel data involved, that's not a 50% penalty but an ε one.



  • @DogsB said in In other news today...:

    I still don't understand why six tabs needs a gig.

    To make it fast. Very specifically, to make it fast. Chrome engineers have talked about this. They use memory-hungry data structures like hash tables all over the place, taking the old "speed vs. memory" engineering tradeoff to the extreme on the "speed" setting. When people started complaining about Chrome eating so much RAM, they tried dialing it back a bit, but every time they found that the result was noticeably sluggish browser performance.

    The real problem is that the underlying technologies are a terrible fit for what people are actually using browsers for these days. The HTML/DOM model was designed essentially for rendering static Word documents. JavaScript is a toy language designed for ad-hoc cute animations. CSS is... CSS. None of it was created with "web apps" in mind.



  • @topspin said in In other news today...:

    And yet games manage to push an insane number of graphics primitives at 60 fps

    Yeah, see that bit I wrote about how all that hyper-optimized GPU performance is specifically for high-end 3D, at the expense of flat 2D rendering which is actually getting worse. You're making precisely the apples-to-oranges comparison I already said was invalid.


  • Notification Spam Recipient


  • Notification Spam Recipient

    @Mason_Wheeler are they keying everything in the dom to everything else in the dom.

    Don’t answer. I’m not going to like where it goes.


  • BINNED

    @Mason_Wheeler said in In other news today...:

    @topspin said in In other news today...:

    And yet games manage to push an insane number of graphics primitives at 60 fps

    Yeah, see that bit I wrote about how all that hyper-optimized GPU performance is specifically for high-end 3D, at the expense of flat 2D rendering which is actually getting worse.

    Yeah, I saw that and the response to that was right below: the part you snipped.

    You're making precisely the apples-to-oranges comparison I already said was invalid.

    So you claim that being able to push a huge number of textured 3d primitives at 60fps does not imply that the hardware is also more than capable to push a few hundred textured 2d quads?
    Because that is demonstrably wrong. It works just fine in all the non-shitty apps.


  • ♿ (Parody)

    @DogsB said in In other news today...:

    The cheap end of android is terrible but you move into the 200+ bracket, you'll have a ui experience that's bettermore responsive than most desktops in the last 10 years.

    Still a much worse experience to not have mouse an keyboard and stuck with a tiny screen.


  • ♿ (Parody)

    @topspin said in In other news today...:

    And yet games manage to push an insane number of graphics primitives at 60 fps while the Windows start menu takes between ~0s (as it should be) and ~5s just to appear, and then all the ads aren't even loaded yet. So, really, it's not that.

    What gets me is stuff like putting in a PIN for a smartcard using the built in Windows interface. You hit enter or click ok and there's zero indication that anything is happening for several seconds. This is 100% shit programming by people who hate their users (OK, I can respect that).



  • @Mason_Wheeler said in In other news today...:

    Yeah, see that bit I wrote about how all that hyper-optimized GPU performance is specifically for high-end 3D, at the expense of flat 2D rendering which is actually getting worse.

    Your specific example with the rectangles has been true for at least 15 years and probably closer to 20. 2D rendering performance hasn't gotten worse in any way -- actually, the reverse, it's gotten way better as hardware has improved. There's nothing magic about 3D. At the end you're still drawing 2D shapes and/or pixels in a 2D image.

    You're making precisely the apples-to-oranges comparison I already said was invalid.

    You're basing this on assumptions on your side that simply aren't true.

    Besides, if you want, you can always implement a specialized software renderer. You can do all the stuff you did 25+ years ago, except that you now have modern hardware. In that context, the extra copy of the framebuffer to the GPU to display it, isn't going to be a bottleneck. (And you can very likely bypass it for integrated GPUs that share system memory anyway.)



  • :facepalm: I never said that was the only factor. In fact I specifically said the opposite, that:

    There are a lot of little details like that that add up quickly as the various penalties stack multiplicatively...



  • @Mason_Wheeler said in In other news today...:

    :facepalm: I never said that was the only factor. In fact I specifically said the opposite, that:

    There are a lot of little details like that that add up quickly as the various penalties stack multiplicatively...

    Granted, it's been more than a decade since I actually poked at 3D now (I think) but the only heavy part of doing a 2D UI on 3D hardware is generating the texture, so the speed of the card is kind of irrelevant, the CPU and data bus will be the bottle neck either way, because even internal gpus should be able to handle tens of triangles with 4k textures at an fps where it will not take a second to render a new window because of the 3D rendering, nevermind actual 3D hardware.



  • @Mason_Wheeler Ok, but what other penalties?

    The one that you listed wasn't true (keep it at 4 vertices, or if you want to be clever about it, an origin plus an extent, equivalent to 2 vertices, since it's most likely axis aligned, or better, don't bother because it's a minuscule amount of data and your profiler hasn't told you it's a bottleneck), and is in practice pretty much the opposite of a penalty.

    At the "cost" of having to split it into 2 triangles somewhere in the pipeline, you get a hardware (like, actual silicon) unit that does rasterization for you. Plus, you get efficient clipping and related operations for free. And, it'll automatically schedule it on whatever SIMD-like hardware setup you have and run everything in parallel.



  • @topspin said in In other news today...:

    the Windows start menu takes between ~0s (as it should be) and ~5s just to appear, and then all the ads aren't even loaded yet

    Click - and here it is. Complete, without ads.
    Windows 7, and animations cräpps / aero switched off.



  • @BernieTheBernie even Win7 with Aero and animations on my 2010 era box felt faster than Win10 doing the equivalent needful in 2023.

    The terrifying part is when I break out the Amiga emulation and load entire applications in less time (even though it’s emulating an entirely different set of architecture including the speed of the mechanical drive)



  • @DogsB said in In other news today...:

    are they keying everything in the dom to everything else in the dom.

    The dom has the key and the sub doesn't, but that thread is :arrows:.



  • @topspin said in In other news today...:

    all the non-shitty apps.

    E_DIV_BY_ZERO



  • @Mason_Wheeler said in In other news today...:

    (Rendering two triangles requires 6 vertices with all their associated information, compared to only 4 for a single quad, just for starters, so add on another conservative estimate of a 50% performance penalty there.)

    You'd have to go out of your way to turn what can be a 4 vertex triangle strip (or maybe something even more optimized) into two separate triangles



  • A funny story. A hungarian man worked for an austrian company. His salary was paid to his austrian bank account. Well, there is a small issue: austria uses the euro, hungary has still its forint. And 1 Euro is at 367 forints. And the austrian company paid him the number of moneys in euro instead of forints, and now try to get their money back.
    :nelson:
    https://www.faz.net/aktuell/karriere-hochschule/ungarn-mann-erhielt-lohn-irrtuemlich-in-euro-und-verweigert-rueckzahlung-19076548.html



  • @cvi said in In other news today...:

    the article doesn't actually say why modern PCs "may feel" slow.

    Yes it did. Windows 11 :half-trolling:



  • @BernieTheBernie What salary does his contract say he's entitled to? If what they paid him was the wrong amount, then they should be able to reclaim it. But if they screwed up on the contract, they should have done a better job writing it.



  • @Mason_Wheeler It also depends on what the laws there are. The U.S. uses the UCC, which basically specifies the defaults, and then you can put most anything in the contract. Hungary and Austria may have a different split between what can be overridden in a contract and what cannot be as well as different defaults.

    But I think it's doubtful that his employer is just out of luck unless it was as you say.



  • @DogsB said in In other news today...:

    @Mason_Wheeler are they keying everything in the dom to everything else in the dom.

    Don’t answer. I’m not going to like where it goes.

    Remember that React and co specifically create a “shadow DOM” so they go out of their way to minimise changes to the “real DOM” for “performance”.


Log in to reply