Jump to content

UCyborg

Platinum Sponsor
  • Posts

    2,572
  • Joined

  • Last visited

  • Days Won

    28
  • Donations

    100.00 USD 
  • Country

    Slovenia

Posts posted by UCyborg

  1. I recall there was something about Mypal 68 and executing Web Assembly code...ud2 instruction was encountered and that threw crash dialog. There are legitimate uses of that instruction and a crash dialog (or kernel panic if you're in Linux kernel) are not expected (https://stackoverflow.com/questions/7268352/linux-kernel-module-to-check-memory-integrity/7268558#7268558). And another one could happen when forcing OpenGL compositing, which AFAIK Mozilla never intended to be used on Windows in the first place, but works in some browsers with glitches. Not exactly performant either, but with it, you can have video in a web browser without screen tearing, which I find to be an unusual phenomenon on Windows XP. I did notice at the later point D3D9 compositing works as well as with the old Firefox versions and roytam1's forks, just needs to be turned on manually in about:config.

  2. Aye, I suspected D3D9on12 requires higher feature level, although D3D12 in regular usage supports cards with only 11_0, but then you can't do as much with it as with more capable card, but it makes sense from game engine development standpoint to use D3D12 library to be able to utilize all available bells and whistles while leaving more limited experience for users with older cards (still better than nothing).

    NVIDIA does support their cards for a good long while. Kepler cards also get limited support until September 2024 (security updates), along with Windows 7 users in general. I think they have it as reasonable as it can get, seriously supporting a piece of hardware or software is not exactly trivial.

    BTW, found this: https://techunwrapped.com/dxvk-outperforms-d3d9on12-when-running-directx-9-on-an-intel-arc-a750/

    What caught my attention more than DXVK vs D3D9on12 results, wow, Intel makes discrete graphics cards now?

    Anyway, I didn't get around trying the older driver yet, just not a big enough issue to deter me from messing with other things, but I'm still curious about users on other cards where temperature can be normally shown in Task Manager, especially those that don't use NVIDIA.

    Heh, maybe a better question for another forum, Vistapocalypse said it right when he said even Windows 7 is too new for many people on this forum. :D

    Edit: Wrapper for enabling D3D9on12, feature level 11_0 is minimum: https://github.com/narzoul/ForceD3D9On12

  3. Made a big jump and experimenting with Win11 23H2 with the latest December 2023 cumulative update and NVIDIA driver version 546.29, the temperature still disappears in Task Manager, but no noticeable degradation in how graphics works in general. Interestingly, I have resolution 1440x1080 available out-of-the-box now, it had to be added manually before, it's the largest 4:3 resolution you can have on 1920x1080 screen, though it's been years since I used it for anything.

    I remember many years ago how newer drivers for GeForce4 MX 440 slowed games down to be practically unplayable, you already had to be extremely conservative with resolution, they looked pretty at 1280x1024, but had to use the blocky (on TFT-LCD) 640x480 for reasonable frame rates in any case.

    No such extreme issues these days that I can tell, though it's a funny contrast, having 2023 GPU driver while all other drivers are from 2010 at most.

    I read something about them running D3D9 games through D3D12, precisely D3D9on12. I don't remember specifics, not even sure I correctly understood that post, just seem to remember something along those lines was mentioned on VOGONS forum. Maybe it only applies to much newer NVIDIA GPUs, at least I don't notice any new DLL loaded related to D3D12 when running a D3D9 game.

  4. On 12/6/2023 at 8:48 PM, mockingbird said:

    Hmmm...  Maybe we should share configurations...  Have you tested this extensively in XP?  When I 'test" -- I mean to say I'm using it as my daily driver...  I have one private window for Youtube logged on to my account, and the normal window for ordinary browsing...  I did get WebGL2 working, per your instructions by disabling Angle, and I also have multiprocess mode enabled...  I think it is when Youtube is playing a video in that second window and when I am browsing in the normal window that it is triggered...  How do you recommend I generate a crash report?  I am not really adept at that kind of thing.  Thx

    Forgot to reply...while it's not a daily driver, I did use it for several prolonged sessions on XP x64, also made it consume most of my then 4 GB of RAM (I'm on 6 GB now :)). I doubt my profile is anything special, but I did have to setup a specific compatibility shim to prevent crashes, which I'm not sure is still needed with current versions. I wrote about it many posts ago, nobody commented on it back then if I recall correctly...what a surprise, no one ever knows anything about my issues I post on MSFN...I'm always one-man army tackling them.

    I don't remember ever needing to generate crash dumps on XP, I'd have to learn it from scratch as well. We have it quite straightforward these days on current systems.

  5. On 12/14/2023 at 1:22 PM, NotHereToPlayGames said:

    Agreed!

    I've been of the same thought for the past couple of years.  I heavily modify my Serpent 52 and the ability to modify-to-liking 360Chrome is why I don't see myself ditching either it or Serpent 52 for a very long time.

    Since I'm normally on Win10, I don't use Serpent there, but I've also ditched Basilisk and didn't come back to it even after its development was taken over. Just don't like having to use huge extension to revert that Australis mess of UI to how it was before, and even with CTR addon, I remember there were quirks. Didn't even Tobin say at some point Australis GUI code is a mess of spaghetti code?

    Still, regardles whether you use Pale Moon/New Moon/Basilisk/Serpent/whatever, it's still plagued by issues everyone complained about back before Mozilla radically the overhauled browser. It's like tweakability has to come with CPU usage spikes, memory leaks, sluggishness with prolonged use.

    No one suffers with these issues while browsing the web...Chromium has minimal customization as far as browser behavior goes, but basics just work. Though considering existence and performance of Vivaldi, it doesn't have to be either one or another.

  6. Trying out Win11 on real hardware for the first time, fresh initial version 23H2, using the good 'ol known tools for restoring/enhancing various functionality (ExplorerPatcher, OpenShell, 7+ Taskbar Tweaker, QTTabBar), noticed something odd with keyboard shortcut for moving windows between screens (Win + Shift + Left/Right arrow).

    The left screen has the resolution 1920x1080, the right screen is at 1280x1024. On Win10 (up to 20H2 at least), if I centered the window on any screen and moved it to the other screen with the shortcut, it would be centered on the screen it was moved to, but moving it on Win11, it's off by few centimeters.

    I didn't find anything written/discussed about it, TBH I don't even know how to word it. Wondering if it's a bug or just another intentional simplification and if there's a trick to make it work like it used to.

  7. Release notes say something about GDI rendering improving remote font rendering over Vista level DirectWrite, so I flipped that flag over in chrome://flags to force GDI. While they render then, it seems to cause some misalignment of fonts in general here, how they're positioned in 2D space.

  8. I noticed when trying out BasicThemer2, which uses functionality built into DWM to activate Aero Basic style window frames, which resources are still in stock Aero theme, that it then actually also follows *.msstyles setting for window title text placement (left, center, right). Did you use the same trick near the end of the video? I guess the text placement for normal theme rendering was always hard-coded, not sure whether it would be "correct" to take that same setting for text placement in "normal mode", though just judging by Explorer windows with ribbon, it should. I haven't checked if this basic mode also allows changing text color, either through the theme or registry settings, but it would be nice if this, along with text placement, was tweakable, in any mode, the basic theme rendering isn't compatible when the app does something with its own frame.

    spacer.png

    If I remember correctly, the trick BasicThemer2 uses is similar to what compatibility shim DisableThemes does. I came across the utility when searching for info regarding Windows 11's frame/border rendering, last time I checked, it completely ignores theme resources for rendering normal window border, it's just always invisible while Windows 10's DWM renders it as usual. BasicThemer2 was suggested as the workaround as the rendering mode it activates still has normal border around windows. While default theme in Win10 lacks those borders, they are in Aero Lite theme and certain custom themes, as also evident in your video.

    So, another thing to consider if you get into tweaking Windows 11's DWM at some point, Windows 11 is catching on as the time passes.

  9. 360Chrome on Vista (should be only GDI, which looks about the same on XP here), no extension:

    spacer.png

    While this one comes out nicer, fonts in general look all over the place when running on an old OS, some too thin, some too bold, others are edgy. Fonts below for instance look way different than on browsers running on recent Windows versions (I'll post the other image later - done):

    spacer.png

    spacer.png

    Hm, could've sworn differences on this site were more pronounced a while ago. That and it doesn't look quite right when captured on the image.

  10. Does Chromium support custom themes? GUI looks and feels like stock Chromium.

    Also, yay: ContextResult::kFatalFailure: Failed to create shared context for virtualization.

    spacer.png

    Definitely much laggier because of it and zero ability to run WebGL. Vanilla up-to-date Vista with SP2.

    Also the below happens when restoring window when using Actual Window Manager, version 8.14.1, had problems with newer versions (it just exited after few minutes), only on Vista though.

    spacer.png

    A workaround is to minimize it and restore it quickly afterwards.

    GDI font rendering looks weird to me, but at least web fonts render now.

  11. When was the last time you visited optometrist? :dubbio:

    On the unrelated note, looking at those images I posted on the site as-is, default 100% zoom, they're definitely a bit sharper in Pale Moon. I've read complaints about Chromium blurring images, but didn't pay much attention to it, especially since I don't do general browsing on Chromium (at least most of the time on PC), it's mostly reserved for resource intense web-based apps.

  12. I use Chrome Font Super Enhancer on Chromium variants that aren't Edge to improve font legibility. Still, it doesn't work for some sites/fonts for whatever reason, eg.:

    spacer.png

    Edge:

    spacer.png

    Pale Moon:

    spacer.png

    I can't get that contrast with ClearType settings on XP. I remember messed with ClearType settings in the browser, must have been Serpent 52, but they didn't do anything either. Nothing works on XP these days.

    That extension does, at least the older version, unsure about the current, would have to check, but last time I checked, it worked in Chromium 86 based browsers - OS shouldn't matter.

  13. On 12/5/2023 at 5:42 PM, VistaLover said:

    As far as script compatibility is concerned, it's true that most userscript authors first target TM, because it's more popular I guess; but VM should also do the job in the 99% of cases; incidentally, the "Restored Pagination for Google" userscript works better in VM than TM, it says so in its code: 

    // @match only allows wildcards for the TLD in ViolentMonkey, not in TamperMonkey or other alternatives

    You didn't read the code right. It's more complex because it handles domain matching on its own, it doesn't use ViolentMonkey's support for tld wildcard in @match rules. tld wildcard is however supported universally in @include rules...at least it should be. But if @include will really have to go to work in future Chromium, hopefully Tampermonkey author could add support for tld in match rules as well.

    I changed metadata block in the other Return Pagination to Google like so, just like tld better than asterisk:

    // ==UserScript==
    // @name        Return Pagination to Google
    // @description Makes Google searches break down into separate pages, rather than displaying as one continuous page.
    // @namespace   Violentmonkey Scripts
    // @include     https://www.google.tld/search?*
    // @icon        https://www.google.com/favicon.ico
    // @grant       none
    // @version     1.4.4.1
    // @author      Jupiter Liar
    // @license     Attribution CC BY
    // @description 11/16/2023, 12:40 PM
    // @downloadURL https://update.greasyfork.org/scripts/468360/Return%20Pagination%20to%20Google.user.js
    // @updateURL   https://update.greasyfork.org/scripts/468360/Return%20Pagination%20to%20Google.meta.js
    // ==/UserScript==

    Though I don't use Google search as much as I used to, the script seems to work nicely. I doubt there's a reason to fear that @include would stop working in GreaseMonkey for Pale Moon.

  14. https://www.huk24.de/zugang/registrierung/anmeldedaten uses Friendly Captcha. My results on home PC are a bit slower (compared to 32-bit St52 on 64-bit XP) on Pale Moon, both 32-bit build on Windows 10 and 64-bit build on Linux (KUbuntu 21.10 ATM), though the latter is about 2 seconds faster (13 seconds). Recent Chromium, both Ungoogled Chromium 118 and Thorium 117 are through in 8 seconds, both 64-bit builds. I use the SSE3 build of the latter, which is faster in some scenarios than plain builds, but not noticeably in this one. It's interesting that Firefox (110 ATM) fares about the same as Pale Moon. I tried disabling Web Assembly support in Pale Moon on Windows, that slows it down to about 2 minutes and 4 seconds. Maybe I should try 64-bit also without WASM. CPU is AMD Phenom II X4 920 clocked at 3 GHz, which was still quite lagging behind Intel models of the time.

    Web Assembly may not be as speedy as native C/C++, but it's significantly faster than plain JavaScript, which was one of the ideas behind it. I doubt UXP platform does Web Assembly on such old CPU as Pentium III. WebGL error is not related to Friendly Captcha, some other script calls it, but, again, a 1999 era graphics card probably only knows fixed-function pipeline, no shaders, and if it is an office PC, does it even do hardware accelerated 3D? Office PCs have come a long way since. 1999 was also the time when serious 3D games stopped offering software rendering option, at least it began in that era and when developers were weighting their options, but if you were ambitious, you either required users to get a real graphics card or play Solitaire or the other games of yesteryear.

    Not much time for testing at work laptop from 2018, but I use nothing less than AVX2 build of Pale Moon there, with (plain) Firefox and Edge on the side. Pale Moon with Web Assembly support disabled went through in about 15 seconds, with Web Assembly in 3 seconds, Edge is through in 2 seconds.

  15. I understand that, but you won't be able to change that service's configuration without becoming TrustedInstaller first. Some interesting reading:

    https://www.tiraniddo.dev/2017/08/the-art-of-becoming-trustedinstaller.html

    While you don't need to concern yourself with technicalities in that post, you will need System Informer or another utility that makes it easy to run any program in the context of TrustedInstaller (aka NT SERVICE\TrustedInstaller). It's a special service account that has full permissions over various system objects that administrator accounts or more precisely the Administrators group don't.

    I simply suggested running System Informer itself as TrustedInstaller as that program allows changing services' configuration, but if you wanted, you could run also run Windows' services.msc as TrustedInstaller, then you should be able to change the configuration of those services with said utilities as well, at least assuming the program that disabled them didn't mess with the permissions.

    System Informer also allows you to view and change permissions for each service, if you open any service on Services tab and click Permissions. There, you should see in the case of both Security Center service Windows Security Service (at least that's how they're called in US English version of Windows) that Administrators group doesn't have either Modify Configuration or Stop permission.

    spacer.png

    But TrustedInstaller does.

    spacer.png

    And the pictures from System Informer showing default startup type for both services in question:

    spacer.png

    spacer.png

    At least that's how it should be in Win10 builds from the last 3 years or so...sometimes they change some services' configuration with newer builds.

    On 12/5/2023 at 9:15 AM, NotHereToPlayGames said:

    For some reason, GPU% doesn't work for 3.0.6040 and higher, technically only tried on this one computer so far.  :(

    Including the most recent one? It's updated quite often so oddities can come and go.

  16. On 12/6/2023 at 8:10 PM, AstroSkipper said:

    Thus, your comment is not very logical, rather absurd or provocative and certainly not helpful.

    I don't see it that way. if anything, it's the waiting time that is absurd.

    The waiting time does increase to about half-minute when forcing affinity to 1 core, but I'd have to boot into XP to get the exact seconds as it slipped off my mind. That's on some version of St52 from first half of October 2023. It's done in 15 seconds on Pale Moon 32.5.0 on Windows 10. Single-core here increases waiting time to about 85 seconds.

    It does manage to load all 4 cores as-is, which is a rare sight in UXP browser in my experience, when it's stuck with some websites' scripts, it's always one core that is working.

    21 hours ago, NotHereToPlayGames said:

    Not the first time you've cited how slow your computer is and I know I wouldn't use that "all the time" if I didn't HAVE TO, lol.

    From what I remember from his posts mentioning specs, it reminds me of my previous one, but mine was definitely slower, a Celeron instead of a Pentium, less RAM, worse GPU, it didn't even do DirectX 8.

    23 hours ago, AstroSkipper said:

    Here are some facts: my car is from 1989

    23 hours ago, NotHereToPlayGames said:

    My daily drivers are a 2003 and a 1991 - so yep, I drive a very old car also

    I drove a 2001 model until autumn last year, after I switched to a 2022 model, I was the reason the old car was at the house 5 years longer than it would have been otherwise, I'm the youngest in the family, BTW. And we only got rid of the 37+ years old freezing cabinet last year, no idea how much electric bill was inflated because of it, I don't pay electric bills currently and I never asked. :P

    Regarding the car, issues were just cropping up one after another in recent times and engine shutting down mid-motorway shot my anxiety through the roof, which was the last straw. I'm on the road almost every day and the peace of mind is worth a lot to me.

×
×
  • Create New...