|
Combat Pretzel posted:Uh, if I understood things right, it's implemented somewhere in the virtual memory manager and should therefore be transparent. What's probably happening is that Chrome monitors memory usage and things don't add up anymore as it expects it, because total memory usage appears to raise even more when it starts allocating backing memory in the System process, so it starts killing tabs due to ostensibly overwhelming memory pressure. They really should lie about double-allocated (normal and being-compressed) pages. The actual pressure hasn't changed, so don't account those pages to userspace until the compression is finished. Too many applications are pressure-aware to make a sudden fake usage spike like that a good idea.
|
#
¿
Aug 28, 2015 20:47
|
|
|
|
| # ¿ Apr 24, 2024 23:30 |
|
|
Win7 EFI boot is possible, but a serious PITA. Get the newest 7 install disk you can find, and do a fresh install on some spare drive to get the full EFI partition instead of the partial stub you get from a BIOS install of 7. Copy that partition to your original disk, use the install disk to pop a command prompt and bootrec.exe /rebuildbcd Dropping this here in case it bites someone else: Cortana suggestion database corruption causes it to crash in Microsoft.Windows.Cortana_cw5n1h2txyewy\CSGSuggestLib.dll. You can recover from this pretty trivially - it's tied to your profile. There's tons of hits on google for it but nothing out there actually fixes it. Login as administrator, delete or rename your profile's copy of AppData\Local\Packages\Microsoft.Windows.Cortana_asasdasdasdas. Create a new profile and switch user to it (this stuff is created at first login) and back to administrator, then copy that profile's version of cortana to your profile. You can't just copy from administrator's profile since it's got the files locked, but you can copy from any logged-out profile. I'm not sure why it won't rebuild if you just delete the directory - that makes Cortana completely nonfunctional. Hopefully this saves someone a reinstall. As an aside - goddamn is it difficult to use W10 without that search box working.
|
|
#
¿
Jun 13, 2017 19:44
|
|
|
fishmech posted:The reason it happens is because they probably ignore all the messages Windows pops up ahead of restart and then lie that it just restarted "out of nowhere". GreenNight posted:I basically sit with my laptop every night of the week and all weekend and Windows 10 gives you ample loving warning when it wants to reboot. If you ignore it, it'll reboot on it's own over night if you leave it turned on. I can see people not saving their poo poo, leaving it on and then wondering what the gently caress. It's their own drat fault. The problem could be solved by doing the download and update automatically, and just constantly nagging you to restart until you do. You come back the next day when the simulation is finished, let it reboot, and get on with life. But hey, it doesn't impact your forums posting workflow so it must not be a real problem for anyone, anywhere. W10 is absolute poison on this, because even if you pay extra for W10 ultimate you get no say. Good luck getting every freelance petrolium consultant to setup his own IT department and domain server just to be able to render a loving flow simulation. It's also great for the guys writing Cryptolockers, because every time it happens the first loving thing people ask is "how do I make sure this never happens again?" Spoiler alert: there's tons of workarounds on google to disable forced reboots. This poo poo experience they had gets relayed to everyone they work with, and people do it "preventatively" and whoops now there's a worm running rampant that should have been patched because the patch process hosed too many people over.
|
|
#
¿
Jun 16, 2017 04:38
|
|
|
wolrah posted:As noted already an un-rebooted system is an unpatched system. The goal of the forced updates is to reduce the number of unpatched systems. If you do not force reboots at some point you end up with the same problem in a different form. That would be fine, 24 hours is way different than "Left it running overnight everything was gone in the morning." Yes, it was outside "active hours" but goddamn it'd be nice to be able to do work on Tuesday. Updates should be faster too, I've got a 4590 and a SSD, there's no reason that a system with every process shutdown should take 10-15 minutes to copy some files to backup and replace them. It's too late by now, but it'd be nice if long-running processes without user interaction had to register their restart context with the OS or be severely penalized in scheduling. poo poo happens, it'd be nice if renderers and simulators could just save their state and pick it up later. It's far too rare a feature. E: I'm nowhere near as angry about this as that sounded, cutting down the invective. Harik fucked around with this message at 07:35 on Jun 16, 2017 |
|
#
¿
Jun 16, 2017 07:25
|
|
|
wolrah posted:The 24 hours thing was a number pulled out of my rear end for my hypothetical indicator, the current auto-reboot is something like a week after the first notice AFAIK. It's not like Windows systems are all auto-rebooting Patch Tuesday night. wolrah posted:That I definitely agree with, if your render/compile/computation doesn't save its state at least every few minutes it's horribly designed and that's the actual problem. Unfortunately, almost all of this poo poo is terribly designed. Here's an example I've had direct experience with. When designing FPGAs it's not like you have a lot of choices - either use the software from the FPGA vendor or buy a $50,000/seat license for something that doesn't have full knowledge of the specific blocks in the part you're using. Neither option supports resuming a simulation of the design, so if something interrupts it you start over. It takes hours (sometimes days) to do a simulation of a complex operation. In our case, it was trying to simulate the transforms of a dozen frames of 1080p video coming in via HDMI. 50 million clock cycles through a complex part and it gets a thousand or so clocks per second done, so somewhere between 6-8 hours. Lots of industries have similar problems. Engineering software's major flaws come from the fact it was designed by engineers. You get lots of assumptions like "people will use this the way I do", "computers are an always-on resource" and "it's not like there will ever be more than one processor in a computer so we'll do everything on one thread." Fast-forward from 1996 and you're in the sorry state we're at now, because lol at a company with inertia and enough money to buy any upstarts actually changing their sorry-rear end software. poo poo sucks. Applications are awful, OSs are awful. Can we not stockholm the status quo and pretend it's flowers and butterflies?
|
|
#
¿
Jun 17, 2017 00:01
|
|
|
Double Punctuation posted:This also explains why *nixes don't usually need a full reboot. In *nix, if an update to a library isn't compatible with the old version, it gets a new name, since the version is in the name. On Windows, that only works for applications that know about side-by-side, and plenty of legacy applications don't. So *nixes can switch the link for the default version of the library over to the latest version immediately, but Windows has to wait for a reboot to prevent legacy applications from breaking. Also, *nixes use soft links, whereas Windows uses hard links, which makes it harder to tell which version is in use. No, it means exactly none of that. You don't need links to the .so, that's just a convenience for software that really needs to call a specific version by name. The linux update process works fine without it, and you could do the same on the NT kernel with what they already have. You're also confusing different compatibility: aside from some High Availibility gimmicks, no library is binary compatible in place, you have to restart the program to use it. Relocations, internal tables, etc are all in different places. Most avoid changing externally visible bits without a soname bump, though, so you can share an old & new between two different programs and they shouldn't conflict. When you do make a big enough change to change the soname, though, you can't just replace & restart. It's similar to DX9 vs DX10 - both get updates, but you can't just replace the D3D9.dll with a copy of D3D10 and expect the game to work. Same deal on linux - the old & new versions live together, each with their own filename (libfoo.so.5 and libfoo.so.6) and once all the old programs that use 5 are upgraded you can uninstall version5 entirely. I had a huge writeup about this but it was too many  . It comes down to: Because linux is installed via a distribution and 99% of your software comes from the distro, the update process knows what needs to be restarted when a library updates. Microsoft COULD track DLL usage and tell you to what to restart, but they don't support old/new coexistence of DLLs until everything is restarted because it might take programmer time away from telemetry, ad delivery and the all-important Office 365 team.E: I'm putting off Creator's for a few days until I have time to be able to do a full restore if it fucks up. Is there a way to only defer that and get the defender updates in the mean time? E2: clarified what was wrong. Harik fucked around with this message at 08:34 on Jun 19, 2017 |
|
#
¿
Jun 19, 2017 08:09
|
|
|
fishmech posted:Doing that for 20 years is why we have it hard to do now. fishmech posted:Nah, if you let people not do the update as long as they want, a bunch of them will simply never update. Citation: Literally The Last 20 Years Of Windows Updates Existing BTW this is a prime example of why your arguments are poo poo. If you reduce the question to "How to make spherical updates happen in a frictionless vacuum." then yes, this is literally the only correct answer. Just look at how much you narrowed the history we're supposed to look at. Pay no attention to genius ideas like "Hey, let's render fonts in ring0!" "Hey, let's stuff the GUI in ring0!" "Hey, let's run untrusted code from the network in ring0 with undocumented ASM traps to give it direct system access if it knows about them just so we can see what it does." Please also ignore that they've had 20 years to fix the insurmountable (to literally only themselves) problem of making updates happen without reboots. Or to rethink how they deliver services to new code so at least you could just force a reboot of the legacy sandbox instead of the whole system. Or to harden the whole OS and isolate programs from each other. So yes, you're right. The only way to get people to update when it's a process this loving awful is to try to sneak it past them when they're not watching. It's almost like the rest of us are talking about improving the garbage-fire that's the status quo and you're laser-focused on one insignificant aspect.
|
|
#
¿
Jun 24, 2017 08:05
|
|
|
Is it possible to offline/isolate/force everything to not use a core on Win10? One of my new I7 systems has a bad core and I'd like to limp along this weekend since I won't have a replacement until monday. Prime95 instantly faults threads 7 and 8, so it's pretty clear where the problem lies. 3 cores is better than 0. I haven't even overclocked it yet, just didn't have time.
|
|
#
¿
Jul 7, 2017 03:44
|
|
|
redeyes posted:I've never heard or read of any Intel procs with defects like that. I'd be more likely to blame ram memory. Agreed. I started by swapping RAM around, because whoever heard of a bad intel chip? Swaps didn't fix it, memtests are fine and it dies on the tiny FFT in prine95 (In-cache only), and always on CPU7 and 8. It's also instant, I think it fails on the first validation cycle prime95 does, it gives errors like ".4555131 > .4" and the other cores keep working. FunFact: You can browse to prime95 from an install media rescue command prompt and run it from there. Same result.
|
|
#
¿
Jul 7, 2017 04:22
|
|
|
Factor Mystic posted:Crazy idea... try pulling the CPU out and reseating it. Probably won't work, but might work. Already did - another goon suggested that maybe some pins got bent so we carefully inspected the socket. Limiting it to the first 3 cores (6 CPUs) has it up and running. It worked for a full week then started crashing at random - maybe something was super marginal and survived Intel's tests and gave out in the field. I'm as surprised as any of you, it's the last thing I'd expect to be at fault, and the last thing I checked. Swapping ram, putting NVMe in the other socket, making sure the heatsink was good and thermals were right, windows reinstall, BIOS update, checked the memory profile, pulled the CPU and checked the pins, then finally disabled the cores and it's just been running. If it weren't for P95 insta-failing the two threads on core3 immediately every time I started it I'd still suspect a flaky mainboard or something. Thanks, Fishmech. I'd found that link but disregarded it since I wouldn't have expected an old XP (95?) trick to work on 10. Harik fucked around with this message at 05:18 on Jul 7, 2017 |
|
#
¿
Jul 7, 2017 05:15
|
|
|
drunken officeparty posted:Where does Windows as an OS "begin"? What I mean is that obviously it's not down on the level of telling the processor how to flip flop individual transistors for calculations, or the monitor how to turn a pixel on. And it also doesn't make the keyboard work, because the BIOS can use it but other peripherals it is the one doing it. Do they even update that deep down anymore? Is it some 30 year old code that has been passed on, or is there a basement room with one old man using punch cards he got from Bill Gates. Pretty surprisingly low, and a lot of that is new. The old keyboard driver for PS/2 keyboards and mice doesn't help with USB, for example, and USB drivers are replaced every time the new generation of USB comes out. Disk IO is completely new (SATA and NVMe). Each CPU needs it's own special-snowflake support just to be able to task-switch and save all the new AVXXXTREME registers. Nobody uses BIOS calls at this point for anything other than ACPI/UEFI callbacks into pstate management. E: the 30 year old is stupid bullshit like COPY CON locking up your computer in 2016. Yes, ancient dos serial port drivers still hijack filenames. Plenty of other "must be backwards compatible no matter the cost" things too, although most of them try to detect the application they're being compatible for at least. Harik fucked around with this message at 06:50 on Jul 7, 2017 |
|
#
¿
Jul 7, 2017 06:46
|
|
|
Dr Snofeld posted:How can I stop Windows 10 from updating GPU drivers without my permission? I'm trying to roll back to a specific older AMD driver by: Try this? http://winsupersite.com/windows-10/stop-automatic-driver-updates-windows-10
|
|
#
¿
Jul 12, 2017 03:37
|
|
|
Trying to update to 1903 and it's telling me it's blocked by vmware workstation pro. That's nice and all, but where's the "do it anyway" because vmware isn't installed so I don't know WTF it thinks it's found. E: Like seriously, a hearty "gently caress you" to someone who thought it was a godammned good idea to check for programs that would block the upgrade AT THE loving END OF THE PROCESS INSTEAD OF THE BEGINNING WHY ARE THEY SO INCOMPETENT. Any powershell magic to make this Just Happen, drat the torpedos? E2: an exhaustive search finds no files with "vmware" in them on my system. So it's probably finding some non-vmware-named .dll somewhere and of course it doesn't do anything useful like provide details. The ever helpful error 0x1c1900209 which manages to appear zero places on the internet. Final edit: Multiple reboots later and attempts to upgrade and it worked. I had to rename the vmware directory my old images were in, remove cruft registry entries and some preferences files. I have no idea which one the updater was using to determine it couldn't proceed. This is a process that would have completed in 15 seconds if there was a "PROCEED ANYWAY, I DON'T NEED THIS APPLICATION" button. Harik fucked around with this message at 01:56 on Jun 17, 2019 |
|
#
¿
Jun 17, 2019 00:08
|
|
|
GI_Clutch posted:I'm really sick of these feature updates changing my settings. I installed updates yesterday and now my mouse cursor has changed and the sound theme changed from No Sounds back to the default theme, blaring sounds out of my speakers unexpectedly. Get your poo poo together, Microsoft. (merging settings sucks and nobody gets it right, if you end up modifying anything during an upgrade its easier to nuke to defaults rather than try to figure out what the closest equivalent is.)
|
|
#
¿
Jul 7, 2019 19:43
|
|
|
Combat Pretzel posted:Yeah, but then you start running GPU accelerated VMs, and tend to assign your main GPU, which requires shutting down your X or Wayland session, and then you start noticing how much smoother the Windows experience is, and because of various productivity you tend to spend more time in the VM instead of the Linux host, and then just start asking yourself if it is all worth it. Been there done that. You throw an older AMD card in for linux to use on an x8 or even x4 slot, possibly tell bios to boot off it but that's not required, and hand your main GPU to windows. It's not like you need a 2080ti to run linux, I just tossed an old 270x in it and it's fine. Swapping machines is as easy as hitting the input select on your monitor(s). Give them both all the cores, so whichever one you're using goes full speed. You can give a dedicated drive to windows and dual-boot to it if you absolutely need it on bare metal for some reason too, which is a nice bonus. Been my daily driver for years and having one really fast machine is much nicer than two slower machines. Total cost over one good machine is one cheap GPU, extra RAM and a dedicated windows drive.
|
|
#
¿
Jul 16, 2019 01:41
|
|
|
Apparently there's no way to get a laptop out of S-mode without piracy now? You're forced to associate a windows account to do it and absolutely gently caress that noise. e: Looks like I can just pave it with win10 home normal and the internal license will remain. If not, piracy it is. E: Lol^2 I had a throwaway phone number so I said fuckit I'll use that and delete it. Nope! They happily ingest your phone number but demand you give them full deets in order to use anything but Edge on my laptop I paid for. Waiting on the DNA sample kit... (USER WAS PUT ON PROBATION FOR THIS POST)
|
|
#
¿
Dec 10, 2019 18:58
|
|
|
Last Chance posted:what the gently caress is "S-mode? actually scratch that, I don't want to know. S-mode is "please for the love of god someone use Edge" by locking out the ability to use anything else. Or do anything, at all, with your device. That you just purchased. Basically a chromebook with less apps. So really it's "gently caress you, you WILL make a microsoft account" because the only way to get a full license is - get it from the windows store. E: I ended up making an account and promptly unlinked it but goddamn it shouldn't take this long to legitimately use my purchase,  would have been 10x faster.
Harik fucked around with this message at 20:26 on Dec 10, 2019 |
|
#
¿
Dec 10, 2019 20:16
|
|
|
|
| # ¿ Apr 24, 2024 23:30 |
|
|
Flipperwaldt posted:There's just a server side switch associated with the key in the firmware or something like that. More specifically, I don't want MY Microsoft account associated with this, as it's not my laptop. I'm setting it up for the person I'm giving it to and I can't exactly say "hey what's your microsoft password?" before I do. mystes posted:It's weird that you still have to jump through hoops like that considering that they were supposed to have abandoned the whole idea of Windows S being a separate SKU? The entire W8-W10 bluetooth low energy stack was locked behind UWP until this summer. Harik fucked around with this message at 22:22 on Dec 10, 2019 |
|
#
¿
Dec 10, 2019 22:19
|
|