cov3rt

I'm having a hard time finding any clear explanation to how the vcache minfilecache/maxfilecache settings work on windows 9x systems and properly configure them. I have seen vague / limited or conflicting sources on how it works and / or what settings to use for "proper" function. Some sources put a minfilecache far less than the maxfilecache, others like myself put the same amount for both, then there's what appears to be a guideline on not setting the minfilecache to less than 1/24 of usable ram, which is what i also use alongside the same maxfilecache size, and this generally works in terms of not generating as much "strange" errors, but the caching inside windows significantly goes past this, and someone can correct me if im wrong, but it doesn't appear that win9x has a separate "file / system cache", in other words, minfilecache/maxfilecache and system / file cache are essentially the same thing? I also came across something interesting mentioned by rloew in a fairly old post, suggesting that if the maxfilecache is set lower than the internal assigned minfilecache that can be checked in a registry area ( HKEY_DYN_DATA\PerfStats\StatData VCACHE\cMinPages ), then it will not respect the settings. At the same time i was researching this matter, i was discussing this with chatgpt, and they said that even if i used the setting in registry for the internal assigned one for minfilecache, it wouldn't matter, because the vcache function is dynamic and constantly changing. I also came across another post here "https://web.archive.org/web/20100917110735/http://adriansrojakpot.com/Speed_Demonz/Disk_Cache_Optimization/Disk_Cache_Optimization_03.htm", that mentions "But if you set MinFileCache to the same value as MaxFileCache, Win9x would simply ignore it and use the swapfile instead.", which im not sure is accurate or not? Also even if that were correct, that doesn't seem to answer the part on caching overall and why my system's resources quickly deplete with just opening file explorer a few times just after boot, and i'll just assert that my problem is not due to memory leaks, but rather windows aggressively caching, whether dynamically or not, and i dont want aggressive caching. For more clarification on the above, the tested systems had 1/24 of system ram for minfilecache/maxfilecache, with maxphyspage=40000, on cold boot system resources were about 94 percent after a little bit sitting at idle, which makes sense if windows caches roughly 42 MB, and the OS is another 20 MB or so added on top, but as soon as i open file explorer, exit and check resources again, it goes quickly down to 8x percent, and even 7x percent after not too long, though it does go back up to 8x again, but it usually stays a few percent lower than 94 for the remaining time. And then there are programs like cacheman that modify disk cache size, presumably meant for vache settings, but is that any different than just manually editing the system.ini settings yourself, because if not, the program wouldn't really be useful for that area, other than offering some minor convenience, though there is other settings like vfat contiguous / path cache / name cache / read ahead threshold, but even here, it's not clear as to whether or not those tweaks can help if windows already dynamically adjusts these areas with a supposed "optimal" range?

Any updates? Also since we are talking about xnview, i think it's ok to ask about a related thing. Does anyone know if there's an official or unofficial method for getting rid of the file icons showing on the bottom right of thumbnails in windows explorer when xnview is installed on windows 9x / 2K? I know modern os have the option to disable this via folder options by unchecking "display file icon on thumbnails", but it's absent in 2K and older. I asked chatgpt if they know of any solutions but they have concluded that there might not be and / or that the file icon handler part might be baked into the thumbnail part itself.

Ok so fortunately i found a workable route, though the system hasn't gone through rigorous testing, with some important things missing / some apparent bugs, restart issues, etc. Basically i started fresh and used again "YSI_Win2kPro_R3.3_DEV-TEST-processr.iso" obtainable from here "http://static.skver.space/w2k/isos/", but this time before installing, i disabled "multi core" in bios, and used IDE mode instead of AHCI. on setup, i used f5 option "Advanced Configuration and Power Interface (ACPI) PC )", and wallah, it no longer got stuck at "Setup is starting windows 2000". Right after install though, it stalled out with a loading cursor, i waited a while, then manually shut off the system, and it rebooted normally into desktop to finish up some stuff. Now i won't be able to document all the stuff i did up to this point. But i did want to document this here to help others, and give the assurance that acpi support on modern hardware is more feasible, just that you need to carefully examine your bios very well. in my case, i didnt think multi core setting would matter, but it made all the difference here, though i did have to also use IDE mode as well in particular. I used videoprt.sys 5.0.2195.6833 + acpi.sys 5.1.2600.7777 to get basic functionality from the intel hd 3000 graphics. i had black/blue screen earlier on boot with other videoprt.sys versions. i installed the gpu driver using the executable, link for the gpu driver can be obtained from the following "https://web.archive.org/web/20140830162901/http://w2k.flxsrv.org/cgi-bin/dl.cgi?file=win2k_145111.exe". I do have .net framework 3.5 sp1 installed ( ran setup.cmd to install ) obtainable from blackwingcat's site ( https://win2k.org/wlu/wluen.htm ), but im not sure if that's absolutely required for the gpu to work, it seems to be only for the control panel thing which i did test working. The instability issues im having inside the os though could partially be / indirectly affected by too much usable ram at the moment ( 4 GB installed ), so i plan on limiting this to 2 GB ( boot settings ) and seeing if it helps in any way. Also for those dual booting, you can still re-enable the multi core bios setting later on, but you must immediately go to windows 2000 safe mode right after to disable the exclamation / problematic processor name with also the "unknown device" right underneath it in the same section of device manager, or else the system will become unbootable. For some reason, you dont need to disable the exclamation processor when the "multi core" setting is disabled, though im not sure if the "unknown device" that shows in safe mode is directly because of enabling "multi core" or if it had already existed there when "multi core" setting was disabled.

If you can still answer on the usbser.sys / usbhub.sys version 700x known to be problematic, that would help. yes, most of this ( both acpi and usb related ) is something i have to figure out mainly on my own at this point probably, but it would help to know if i can get another perspective before i do any more modifications. Here is the thread mentioning the usb problematic updates" "

I did look at the thread you linked before all this but it was more on general guidelines on how to integrate the related files and does not really provide any other info. The thing is on the topic of replacing files in sp4.cab / driver.cab, etc, i asked chatgpt on it ( yes, i know chatgpt can be often wrong ), but according to them, they said that it makes most sense to replace the files in all areas they are currently located in ( specifically confirming on sp4.cab/driver.cab ), as well as the "_" versions of the files for wherever they are located, However, there are things where chatgpt may not always tell you unless you ask about it, so it does not mean my approach is correct either, but rather that there may be additional steps i need to do that i haven't done yet. On the HAL switching or whatever, i dont think that is because of incorrect replacement of files, because i also used blackwingcat's iso packages without modifying them and trying the same approach ( standard pc to ACPI PC from device manager ) and it still had the same problems as i had mentioned in my earlier posts. I also tried the vanilla 2K pro sp4 iso you requested ( nothing modified at all ) in exactly the same way as the other routes, ( standard pc to ACPI PC from device manager ) and no difference.

Ok i see you didn't mention different / newer versions for hidusb.sys, usbintel.sys, and usbscan.sys, does that mean i dont need updates for these? Other than that, I do need to request for some other files, for "KB891069" ( usbaudio.sys ), "KB890202" ( usbstor.sys ), and "KB883528" (usbprint.sys ), i cannot extract the files from the package, the .sys files are not there, and only appear to be installed via a windows 2000 system. Can you link these, same for "KB838921 ( usbhub.sys ) and usbser.sys ( unless the vanilla sp4 one is ok ). i actually didnt mention KB838417 for usbser.sys, because from one site, it mentions that this update is listed as problematic, causing BSOD on usb 1 devices, so no to 7006 version?

based off my windows 2000 sp4 iso package ( only "usb.in_" (effectively usb.inf) was modified for the related files, i did not touch anything else in the iso for the following files below ): hidclass.sys - 2003-06-19 - 5.00.2195.6655 - yes includes same version, but there's a file with same name in another folder ( driver.cab ) in lowercase that is actually a lower file version at 5.0.2163.1. the reason why i find this important is if i wanted to replace these files with your updated / fixed versions, i want to be sure im replacing them correctly in all appropriate areas, and im not sure if i need to also replace that lower case version too? i know that i also have to replace the "_" versions as well, by creating an extra copy and convert it to that format, such as in the case of "hidclass.sy_", but i dont know if i also need to modify other areas. hidusb.sys - 1999-10-04 - 5.00.2142.1 - same openhci.sys - 2003-06-19 - 5.00.2195.6675 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2164.1 uhcd.sys - 2003-06-19 - 5.00.2195.6655 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2143.1 usb.inf - 2003-06-19 - 5.00.2195.6717 - appears to be the same, driver version 5.00.2195.6717 usbaudio.sys - 1999-10-12 - 5.00.2150.1 - same usbcamd.sys - 1999-09-27 - 5.00.2135.1 - same usbd.sys - 2003-06-19 - 5.00.2195.6658 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2147.1 usbehci.sys - 2003-06-19 - 5.00.2195.6709 - same usbhub.sys - 2003-06-19 - 5.00.2195.6689 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2181.1 usbhub20.sys - 2003-06-19 - 5.00.2195.6655 - same usbintel.sys - 1999-09-25 - 5.00.2134.1 - same usbport.sys - 2003-06-19 - 5.00.2195.6681 - same usbprint.sys - 2003-06-19 - 5.00.2195.6655 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2164.1 usbscan.sys - 2003-06-19 - 5.00.2195.6655 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2151.1 usbser.sys - 2003-06-19 - 5.00.2195.6655 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2153.1 usbstor.inf - 1999-12-10 - 5.00.2195.1 - could only find "usbstor.in_", once expanded, driver version said 5.00.2183.1 usbstor.sys - 2003-06-19 - 5.00.2195.6655 - same in "sp4.cab", but in "driver.cab", it reads 5.0.2138.1 I may not individually test such updates if i could somehow integrate the necessary files into another custom iso to test, because at the moment, my system isn't bootable anymore and i'd have to start fresh anyways, and so i would rather just make another custom iso to test as the next attempt. The issue would be if any windows updates / unofficial updates would override the files with older / not functional ones, but upon extracting various packages, such as update rollup 1 / uu rollup v11, only the uu rollup v11 ( Windows2000-UURollup-v11-d20141130-x86-ENU ) installs newer hidclass.sys and usbser.sys, so i dont think it would be a problem. Though i ought to check inside those updates you linked for more details. As for vanilla windows 2000 sp4 install with ACPI PC selected, yes, it does the same thing, it hangs at "setup is starting windows 2000".

Ok i have now attempted with vanilla windows 2000 sp4 ( technically slightly modified, more on this below ) with ide mode in bios enabled and installing via standard pc. It did not give that message of not being able to detect one or more mass storage devices. after os install, i had later changed to acpi, shut down pc, on reboot, the bluescreen error was exactly the same. Note, the iso was modified slightly for automatic activation + slightly modified "usb.in_ " file in an attempt to get an external usb keyboard to work, because currently the internal keyboard's keys do not all work. the modified "usb.in_" did allow my external usb mouse to work and got rid of the error of "\SystemRoot\system32\DRIVERS\usbhub20.sys device driver could not be loaded. Error Status was 0xc0000034" ( which occurred when using unmodified vanilla windows 2000 sp4 ), but external usb keyboard did not work, it had mentioned code 31 error for it ( usb composite device ) in device manager after os install. Although the external usb mouse worked, it was very laggy / skipping badly when moving, requiring very delicate strides to get any functionality both during and after setup, so it wasn't fully working. disabling usb legacy mode and entering the os however got rid of the major lag / skipping issue. I did not modify the ISO in any other way.

I have not tried the method with vanilla windows 2000 sp4 and ide in bios. I may only try that if there are no other feasible approach. You do bring up some good points on the HAL relating disk driver switching issue. I'm not sure, but it could possible relate to a "strange" problem where if i use ide mode bios in bios and then run setup with my custom iso versions, it gives the message in setup that it cannot detect one or more mass storage device controllers, or something along those lines, but after proceeding with install anyways by pressing enter ( and in this case, standard pc, then there is no problem, until of course i switch to acpi after os install ). i suspect the reason could be the way i slipstreamed blackwingcat's ahci / sata drivers in my own custom iso MIGHT have not been the "perfect" or "clean" method, but then again, it happens with blackwingcat's iso version as well, so it doesn't appear to be an issue with not slipstreaming correctly. Maybe i need to specify adapter in setup manually if using AHCI / slipstreamed iso package?

dont know if the image will work and if it does, for how long, but i uploaded below. the error message on 2nd attempt was exactly the same. https://ibb.co/VpHXL30d

Ok, i did some testing. I did fresh install using "HFSLIP120825_2K_20241225_13h08m" with standard pc option with ahci enabled in bios. after os install, i made sure all stuff installed, shutdown the pc safely, then on reboot, switched to "Advanced Configuration and Power Interface (ACPI) PC" from "standard pc" in device manger, shut down pc safely, then on reboot caused bluescreen, and i also attempted fresh install with same iso and ide mode enabled in bios from beforehand, same bluescreen upon reboot after choosing "Advanced Configuration and Power Interface (ACPI) PC". I did not touch any system files. although the error apparently relates to storage controller according to chatgpt ( upon 1st attempt ), when the error happened again on 2nd method attempt, chatgpt then dialed it back to acpi. I wanted to upload a picture of the bluescreen but couldn't. If anyone knows how to upload within the max limit for msfn attachment, let me know. I've tried all sorts of methods but cant get it to work, either copy pasting from paint / google chrome copy image doesn't work or uploading directly the png file of 76 KB doesn't work either.

When you say for example switching the hal in the device manager might not work properly if someone messed with the driver.cab, does that mean that i'd need to match exactly what i installed with in order to change from standard pc to acpi in device manager? Because it doesn't seem i'd need to do that, in other words, as long as the system has all the updated files from beforehand and none of the older / "incompatible" ones, then that would be sufficient to go from standard pc to acpi ( success is a different story ), which is what i was eluding to in my earlier post. The way i'd replace acpi or related files is simply searching the system and replacing it anywhere it's located with the updated ones, such as in the case of just before installing "Windows2000-UURollup-v11-d20141130-x86-ENU" if not using "HFSLIP120825_2K_20241225_13h08m" or "YSI_Win2kPro_R3.3_DEV-TEST-processr", but rather from regular 2k pro sp4 iso with update rollup 1 from beforehand like you mentioned. I will attempt again with at least one of these options with installation via standard pc and try attempting change to acpi, and will see what happens.

Thank you for the info. When i was speaking with chatgpt earlier, it had also referenced 3 of 5 of the new files you added ( halacpi.dll (ACPI uni-processor), halaacpi.dll (ACPI multiprocessor), halmacpi.dll (APIC multiprocessor). It said that these files may have an implication to why i can't get past the "setup is starting windows 2000" part, because it has do with "a pre-OS ACPI/IOAPIC initialization deadlock in BIOS", in other words it cannot properly translate / understand and put together these things, so it stalls/breaks at this part, and that the other files like "ntkrnlmp.exe" aren't involved at this point yet, ( those come later in the setup/install phase ). I know i also mentioned i wouldn't install with standard pc and then enable acpi after os install, but at this point, i dont think ill get any hope installing with acpi, but at the same time, i dont see any sort of particular method would work after os install to get acpi to install correctly, because it still comes down to a careful combination of files, and if anything in the combination is not exact and / or any individual file isn't finely patched/tweaked for windows 2000, then it will just lead to a unbootable system and no way of restoring with safe mode last known configuration either, at least i tried this a few days ago and i couldn't get the last known configuration to work ( would give a black screen with blinking cursor endlessly ). So that being said, i am still leaving this as one of the last resort options again. Currently, bwc's version probably is the quickest to do this method, because his system doesn't need a long update process, almost everything is already installed from beforehand, i just replace the acpi / hal related files, then go to device manager to change to acpi presumably and see if it works better with these files?

ok, here's what i got for comparison of different iso packages / reference points: the "HFSLIP120825_2K_20241225_13h08m" iso package: acpi.sys - version 5.00.2195.6920 mup.sys - version 5.00.2195.7006 ntkrnlmp.exe - version 5.00.2195.7392 ntkrnlpa.exe - version 5.00.2195.7392 ntkrpamp.exe - version 5.00.2195.7392 ntoskrnl.exe - version 5.00.2195.7392 fastfat.sys - version 5.00.2195.7075 ntfs.sys - version 5.00.2195.7050 udfs.sys - version 5.1.2600.5512 dmio.sys - version 2195.7058.297.3 ntldr - date modified 2007-02-17 - 303 KB official windows update file list mentioned by "Start Me Up": acpi.sys - version 5.00.2195.6920 mup.sys - version 5.00.2195.7442 ntkrnlmp.exe - version 5.00.2195.7473 ntkrnlpa.exe - version 5.00.2195.7473 ntkrpamp.exe - version 5.00.2195.7473 ntoskrnl.exe - version 5.00.2195.7473 fastfat.sys - version 5.00.2195.7075 ntfs.sys - version 5.00.2195.7050 udfs.sys - version 5.00.2195.7032 dmio.sys - version 2195.7058.297.3 ntldr - date 2004-12-28 for the "YSI_Win2kPro_R3.3_DEV-TEST-processr" iso package below. acpi.sys in i386/driver.cab is same file version but different product version as old.acpi.sys that is contained in i386, and that file product version is "5.1.2600.7777.8", where as the earlier 2 files read "5.1.2600.1106" for product version. i'm not sure what the installer logic is on this iso, but i'm assuming the "old.acpi.sys" one isn't the default ACPI that installs, but is it even possible to choose which one to install in this case? acpi.sys - 5.1.2600.7777 for file version, product version varies because of multiple variants in the package. mup.sys - version 5.00.2195.7006 ntkrnlmp.exe - version 5.00.2195.7394 ntkrnlpa.exe - version 5.00.2195.7394 ntkrpamp.exe - version 5.00.2195.7394 ntoskrnl.exe - version 5.00.2195.7394 fastfat.sys - version 5.00.2195.7075 ntfs.sys - version 5.00.2195.7050 udfs.sys - version 5.00.2195.7032 dmio.sys - version 2195.7058.297.3 ntldr - date 2004-12-28

Thank you for info, i likely won't be going the path of standard pc to ACPI enable after os install, however...., the information you posted on the files is giving some interesting findings because i checked at least "mup.sys" and "ntkrnlmp.exe" from your list, and they are newer than the versions included in "HFSLIP120825_2K_20241225_13h08m", which doesn't really make sense. Why would BWC's package use older ones, considering your list is apparently based off "official" windows updates? Of course, what makes things more complicated is that BWC's package is also integrating extended kernel which includes files not from your list, like halacpi.dll, etc, though i actually took that reference from his older "Windows2000-UURollup-v11-d20141130-x86-ENU", which isn't even the one his more updated iso includes. I never used his newer extended kernels after OS install because they didn't work for my system's environment, so i only installed up to ""Windows2000-UURollup-v11-d20141130-x86-ENU". I also checked "Windows2000-UURollup-v11-d20141130-x86-ENU", and at least "ntoskrnl.exe" is a older version than the one from your list, so also interesting.

I'm having a very hard time trying to install windows 2000 on modern hardware, specifically on a sandy bridge system at the moment ( hp probook 6460b with 4 GB memory installed and latest bios ). I have tried so many different approaches / methods, from disabling serial / parallel ports in bios, disabling other devices like fingerprint device, trying in ide mode vs ahci, limiting to 1 core, disabling hyperthreading, etc, but i just cant get any sort of method to work with default acpi install, or even manually choosing any of the acpi related entries from the F5 menu. Because ACPI is needed for many functions like gpu driver ( intel hd 3000, etc ), installing in standard pc is therefore not a choice. All systems have same issue, they get stuck at the "setup is starting windows 2000" part. The specific media i tried was a relatively minimal modifed 2K SP4 iso with ahci / sata drivers preloaded + acpi.sys 6921 integrated as referenced in some guides, though i didn't delete the "usbhub20.sys" file in the installation contents. Other iso i tried is the one from here "https://w2k.skver.space/isos/YSI_Win2kPro_R3.3_DEV-TEST-processr.iso", and "https://archive.org/details/windows2000-hfslip-english-finished" with filename "HFSLIP120825_2K_20241225_13h08m", all end up with same hang.

It complained on a missing export relating to unicode when trying to start, i forgot the name, i think it was called something like "ChangeDisplaySettingsExW". but from what chatgpt is telling me, there are more dependency issues with this particular version of gom player if trying to run on windows 95, so it may be harder to get to work properly than say xnview 2.13.

I have not been able to find any media players that work on windows 95 that support 10 bit h.264 or some of the more modern codecs, with h.265 likely being too far in scope of what can be modded to get working, as even 98SE cannot use such codec unless you install those kernelex stub related updates which is already difficult / confusing to do and introduces other problems, so i was hoping that someone can get at least 10 bit h.264 to work, that would be reasonable, but i also specifically would like gom player 2.1.43.5119 to work on windows 95, as it does support 10 bit h.264 and with some modern skins can look really nice with playlist function. If not that, perhaps modify one of the newer ffdshow's that support the codec and use it alongside haali splitter that is already win95 compatible + a compatible media player of choice. as for alternative media players to use with external codecs / system codecs, there are not a whole lot that support windows 95 and have a more modern playlist ui function, the only one i can think of is zoom player, though exact steps to do set it up with codecs is unclear. im also not sure on last / best version of zoom player to use, as i have had issues getting the program to run on windows 95 at any basic level, ( i've tested as old as 5.x and as new as 10.x ), but the program would either not load properly or would throw all sorts of errors, but there's a chance it could be a combination of the testing environment not being set up right, like improper gpu driver / missing those external codecs, etc. there is also tcpmp mod, but that one unfortunately only supports the older h.264 codecs and so it's really only a fallback option at the moment for limited video playback.

i did not paste screenshot as the forum only allows a certain file size, i dont know how to workaround this, that's why i mentioned it in my earlier post, but i also edited that older post so the link should include the file for a few days at least. as for the other responses, we are getting some good progress here, it shows that even on 98SE, one needs kernelex 4.5.2 for xnview 2.13 to work correctly, at least for webp in testing, for windows 95, shows 2 missing depedencies in your screenshot.

apparently webp can be lossy or lossless in the same file, so that part isn't too important for here, but i did upload a sample webp file, as well as a picture of what i mean by the directory part where you can see the thumbnail. if you can, please use the webp file from the easyupload link that i attached and view it in the xnview directory, as well as opening it / selecting it within the program just to know that both work. if the link doesn't work, let me know. i didnt know how to upload the picture here, that's why i included it in the easyupload part. https://limewire.com/d/GVipQ#9miaxwhRQ6

ok i was just hoping someone can test and prove working via screenshot of a sample webp file with your patched version, since that was one of the main reasons i wanted to get the newer xnview version to work. i do not have a windows 95 system to test currently and probably wont for a little while.

no it's ok, you don't need to make a video, i now understand the instructions, but i should still ask, what is the "xnview.exe" file for? it's a bit large in unzipped form, at around 4.9 MB, with package zipped at 1.6 MB, but im still trying to keep file sizes as small as possible. Also, a few more questions - could you show a screenshot of testing a sample webp file in xnview 2.13's directory thumbnail view and / or the file opened to show it's working in the program? Also, is that xnview.exe file you supplied derivative of the full version of xnview 2.13? Just was wondering if it matters if it is or not, since i will be installing the full version.

Maybe im misunderstanding from the screenshot, but is that shell32.dll 4.00.1111 the version your windows 95 pc was using at the time, because isn't latest official version 4.72.3612.1700 for windows 95?

ABCDEFG, can you provide clarity on what your attached file is from your post with screenshot of xnview 2.13 running? I downloaded it and inside appears to be a helper file and a xnview application. What do i do with these?

Also may be relevant, "legacy extender" which helps with compatibility problems with certain applications made with newer versions of visual c++. The thing is, i haven't been able to find a full / installable package, as the one from the website broadly titled as "Legacy Extender for Visual Studio 2005/2008" seems to be only a developer package. The initially searching was about running newer versions of dosbox on windows 95, and when i encountered this line "VS2005 compiled DOSBox *Windows 95 works with LegacyExtender. Fixes InterlockedCompareExchange" from a vogons.org thread, it made me want to search what "legacyextender" was, as the interlockedcompareexchange was specifically the missing dependency from xnview 2.13 / 1.99.x on windows 95.