xrayer

I got report that current patched driver now works with 256MB VRAM card so anyone interested to try with 512MB?

Hi, I know this is an old thread and Xompie was probably abandoned but I found it quite recently and finally had time to play with it because I see some advantages in patching individual apps instead messing whole system. So I installed Xompie (the last version from Github) on my up to date XP SP3 + Posready updates with fresh install of MSVC++ 2013 & 2015 redist. Problem is that any program I patched didn't work then (those supposed to work like Adobe DNG converter that was mentioned by author). LDD finds all dependent DLLs after patching in my windows\system32 but the patched program just terminate with message window: application was terminated due to press CTRL+C. I tried install back all xompie alpha versions from 0.6 to 0.2 (on windows from backup without previous xompie) that I found in the thread. I read it all througly but nobody reported such problem here. Any idea what can go wrong? I even tried to patch a small w32 console Hello World program compiled by mingw that normally runs on XP but after patching it didn't work too (again CTRL+C message), strange... Could be problem that I use localized (CZ) Windows? I also tried Dibya Extended Kernel digged somewhere in outer space, I was able to run Adobe DNG converter 9.6 but it caused crashing of all my browsers - Seamonkey, Palemoon, IceAPE. Then I finally found One-Core-API - I wonder it's not discused much more here, I just found one 2-page thread where someone asks how to install. It seems this is far more recent and largest kernel extension project. I installed it and happily found that it really can run some newer apps but I have the same problem with crashing browsers. The bad thing is that the API extension is enable by default for all apps or I don't know how to disable it by defult. There's not much docs. I like that in W98 I can set KernelEx disabled by default and I enable it only for indiviadual apps. Then I discovered that the problem is caused by WineD3D component. If I uninstall it then browsers run fine. Probably they are trying utilize DX10/11 that they detects but it's not well implemented yet. When I just rename d3d11.dll to something elese, Palemoon will start and run but Seamonkey start with blank window content that doesn't redraw. The various compatability setting of the exe that I tried had any effect on it, again this is so bad to have such features enabled by default. I have to say that I'm running One-Core-API-Binaries-1.1 release that was the only I found compiled and it's from 2018 but sources on github seems to be more recent - anybody rebuild it or have a link on newer binaries?

The previous version of patched driver caused crash (memory protection BSOD). This was probably due to Dword at 144h that I left untouched but it seems it may be code size so I had to adjust it. Now it should go a bit step further, file reuploaded: http://rayer.g6.cz/1tmp/7772mod.vxd

No, 77.72 will not work with 7900/7600, AFAIK only 7800 works (G70 core, others are G71/73 - too new).

BTW Windows XP can be installed with uniprocessor kernel so then you better compare singlecore performance in w98 vs XP. Also you can just swap your CPU to an older singlecore one to see if SMP cause so high gain in XP - I doubt so. Newer videobios would be good to try but aslo I doubt it help, it usually can fix/break some VESA features... I flashed a lot of VGAs but never it had significant performance difference (except bioses with different clock settings). In the distant past I had a big headaches with AGP on non-intel chipset boards. They were cheaper and I had not much money as a student so I had to suffer all the hell of installing VIA4in1 drivers, nvidia drivers, DX, OS reinstalls, bios flashing... After that I finally changed MB to intel ZX440 (later BX440), just installed intel INFs and all problems has gone. So I rather sticked with intel chipsets for intel CPUs for all my next systems. I totaly ignored Pentium 4 as I think it was the worst CPU intel ever made and I just jumped in when C2D Conroe was released. First I got Asus MB with intel 945 with PCIE and it runs very well with Win98SE and 7600GS, later 7900GT. Same for next MB Gigabyte with intel G31. So if you have a chance to get intel chipset MB for good price go and try it...

As you know, R.Loew made nvidia videomem size patch for >256MB cards, that modifies NVCORE.VXD of 82.69 drivers. A friend of mine asked me if the patch can be backported to 77.72 drivers because he claims the newer drivers renders artefacts in some games on his GF7800. I never had GF7800, only GF7600 and GF7900 and had no problems with 82.69 drivers (except shutdown problem). So I tried to look what the patch changed on NVCORE.VXD ver 82.69. I found 3 code hooks by JMP/CALL that referenced a piece of new code added in a caveat near EOF. I tried to find similar code in ver 77.72, patched it, copied new code and modified the offsets to match. I'm not sure that I found the right matching code and how much the 77.72 differs but some parts looked pretty similar. I only don't know what to do with 3 changed Bytes at offset 144-146h. I thought it may be a checkusm in LE header but it's further and my tool says the checksum is 0 even in original file. I hope Windows would not complain and load the driver anyway. I cannot test it and he too coz his GF7800 had died. So if someone interested plese test it on GF7800/512MB and let me know how it works. Patched driver is temporarly permanently uploaded here http://windows98.xf.cz/index.htm#NV7722PATCHED Install original 77.72 driver and then replace NVCORE.VXD by this patched one from the package.

Cool, is there plan also for extender of userspace API (some kernelex ressurection for NT with e.g. a help of WINE codebase)? If I remember I saw some attempts but nothing alive...

Hi, is there some newer unofficial build of MPC-HC player than 1.7.13 (2017) or LAVfilters newer than 0.70.2? I found an (seems to be good enough) alternative PotPlayer: http://potplayer.daum.net/ It still supports XP in 2020 and seems to be often updated. I also reccomend to install Strongene Lentoid HEVC Decoder http://www.strongene.com/en/hevc/hevcProduct.jsp for those with old VGAs without HW H265 decoder. This codec is ~2x faster on CPU than LAVfilters/FFDshow! It should support also OpenCL decoding on GPU but probably only on ATI/AMD cards didn't work on my GTX 670 as GPU load was seen only ~8% If someone knows about working OpenCL/CUDA hybrid H265 decoder that works on nvidias let us know.

Hi, did someone try to compile some recent 7zip for Win9x? I tested that official ver 18.06 is working with KernelEx but only the commandline tool 7z.exe but not the GUI. GUI version 16.04 can run with KEX but only some parts are working, eg. extract files from archive but cannot create new archive.

On P31/ICH7 chipset the IDE legacy mode works just fine, I didn't use AHCI there. I installed the INF files after Win98, no slipstream. In device manager I had proper ide controller working in DMA mode. I installed nUSB 3.3c The only device I had disabled in BIOS was onboard intel HD audio. I had some issuses with SB Audigy 2 and sound crackling (all OS I had installed) so I replaced it by Audigy 1 and it works fine. Used Creative driver 3.0.0.18 (according to INF file) I was lucky to have real floppy. it's sometimes usefull so didn't need to hunt phantom :) The shutdown probloem with latest nvidia driver was never resolved probalby. I used a hack: instead regular shutdown I use a batch file that set a flag (some file on disk) and reboot windows instead, then autoexec.bat checked if flag was set and then clear the flag (delete file) and call FDAPM to shutdown machine. BTW as he has only 512MB RAM no need to himem.sys hacks. But I reccomend to install R.Loew RAM patch and install more RAM.

Aha, I just found that Borland C supported something called a precompiled headers. This should generate some extra file from *.h files that is then combined with *.c files via specific BCC command line option. It should help to speed up compilation. Maybe he used some historic slow PC for building DOS tools? If you find some free time please upload packed compiler, maybe he had placed headers and compiling batch/makefiles there. I don't hurry with it...

Many thanks for sharing with us. It may look he used some Borland C, I'll try to rebuild. I wonder he used a lot of inline asm code and there's no any include statement, even for <stdio.h>. Maybe he didn't even use stdlib. If I cannot rebuild it will be still usable to study the code how DMA IDE commans calls works. UPDATE: Well, the code looks a bit cryptic but it compiles with BC++ 4.52, just to ignore tons of warnings about missing prototypes. This can be fixed easily creating a header file...

Thanks for permission of distributing the patches. In a case that even source codes become freely available I would be interested in extending the TRIM tool support also on FAT16 FS. I started to write my own TRIM years ago with partial success but after discarding the first block (I checked that it was really blanked) my IDE controller left in some error state with busy LED on and I didn't manage to solve this problem. Nobody can help me with this because not much people knowledges goes so deep and want to share it. RL made it working...

That's a pitty he disappeared with his deep knowledges of this old systems. I remember we discused various HW problems and he helped me with RTL8111B under Win98 via his free SPLIT8MB.EXE tool. R.I.P.

I finally found the problem, I had corrupted the file WINDOWS\INF\UNREGMP2.EXE I just replaced this one file in my current XP and installation of drivers was OK, no more explorer crash...

I have found that my problem is not triggered only by nvidia installer but also others. When I tried to use DDU I had to install .NET before and I got the same explorer hang/crash problem after .NET installed. So it seems some more generic problem not nvidia related, it was just coincidence that I first saw this problem after installing nvidia driver. I have ~1/2 year old backup of my WXP installation that doesn't have this problem. So I'm trying to isolate which component became broken in my current setup. I still have nothing in focus but seems it's not registry. I could import my entire user profile from current setup to old backup windows and it works but after copying some other components it breaks. It's quite time consuming. If I remember I didn't installed much new apps neither a driver since that time. I only installed some XP embedded updates...

I moved a little bit further. Diffing register files I found that for some reason it was completely deleted the content of registry key [HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Explorer\Discardable\PostSetup\ShellNew] When I restored this key from backup before driver installation explorer.exe stopped crashing when right click on desktop but it crashes a bit later when I move cursor to "new" entry in context menu. It seems the content of key above gets deleted again on next crash so it's not a cause but consequence that have further impact on system. When I try to import complete HKCU backup .reg file I got warning that some keys was unable to import because are used by system and explorer still crashes then. Only when I replace offline the NTUSER.DAT file it fix. I need some offline registry editor that allows me replace multiple keys from backup .reg file to track down which key is corrupted.

Hi, I upgraded VGA in my PC from 7900GT to GTX670 (WinXP SP3 with current XPE/POS Ready updates). Even when I cleaned out old drivers with Driver Sweep I'm having troubles with newly installed nvidia drivers. I tried different versions from 305.x to latest XP compatible ver. 368.81 and I installed only the driver (no update, physx, etc.) The problem is when I clicked with right mouse button on the desktop (or press windows context menu key) the explorer.exe crashes with typical error window with available stackdump to send a report. When I click OK the desktop disapears for a second and explorer process restarts itself. When I go to control panels I can start nvidia control panel normally and everything else seems to be working. When I uninstall nvidia drivers via control panels add/remove it still keeps crashing when click on desktop. I repeated instalation process more times and found that the problem appears just before the driver installation is finished (at ~90% of progressbar). Next I tracked the problem must be in windows registry. When I (offline) restored my previous NTUSER.DAT file the problem disappeared but of course I have not working driver then. I also tried Nirsoft shexview to disable all nvidia shell extensions but didn't help. Any idea? I really want to avoid reinstalling.

Hi, after some longer time I tried to update KernelEx 4.5.2015.3 by Drugwash with latest version 4.5.2016.16 from here (1st post) and found that some apps, e.g. TCPview 2.54 stopped working - now showing MSVCRT error. I isolated that this issue happen after KEXBASES.DLL file was updated so I keep older version and it seems to work without problems. Did I missed something else that needs to be updated from other package than files in KernelEx2016.16.7z archive?

Your version of the story is quite misguiding. You probably see the history via AMD fan angle of view. I, as an intel user, would see it a bit differently. I'm not blind to adore intel beyond the limits. I appreciate AMD's success and inovations they made in the past. The biggest one was integrating the MCH into the CPU that dramatically improved memory bandwidth. Intel was some way suspicios about that but finally they adopted in new core Nehalem generation... But I wanted to clarify and complete the information that after successfull Pentium III intel was developing new microarchitecture called Netburst - Pentium 4 generation that trurned out to be not much successfull. They was forced to gain power with high frequency that caused overheating issues. In that ages AMD ruled, that's true. But the card had flipped when intel decided, finally, to trash Netburst and continue with P3 evolution (P3 in fact never ended but was still developed by some small Israel team who created Pentium M and then Core solo and core duo - this was aimed to notebooks - intel Centrino not desktops/server) and they came with new Core 2 duo that was, IMHO, the most successful intel CPU that brought very significant performance gain with limited power consumption. It's not true they made 2 small weak cores. They are full featured cores that can share large L2 cache. Every benchmark or real test show you, that even on single core the C2D performed better than P4 at much higher frequency. That was due to improved microarchitecture and cache not because of 65nm technology (also P4 was manufactured by 65nm process but it didn't help them much). So intel quickly overrun AMD and then continued with also very successful core i7 Nehalem and Sandybridge. AMD response was making 6 or 8 cores CPUs but they have currently worse perf. per core and still in 2016 there are lot of singlethread apps that matters. Simple because some algorithms cannot be divided and paralelized. BTW I'm not happy that intel has no a real competitor now, their evolution slow down a lot (new core generation can brings say 0-5% gain in common apps) and prices stays high... BTW this was a reason why I decided to upgrade to 4 years old Sandybridge 2600K instead of latest Skylake CPU, it gives me similar performance but for ~1/3 of price...

Yes, I already patched PE header version more times and sometimes it works. But it's not enough to run e.g. newer version of Adobe Lightroom that needs newer API AHA, VAIO seems to look promissing, I'll try...

1. I use Windows XP SP3 with recent XPE fixes and PAE >4GB patch on my upgraded machine core i7-2600k / 4GB RAM as primary OS. I also use XP at work because we need at least one PC with older OS to be able to use some specific HW which doesn't have drivers for Vista and newer systems... 2. I don't have serious problems, I only found that XP is not capable tu utilize hyperthreading even worse it makes some apps run slower. I experienced some apps (usually singlethread but WinRAR too) run some tens % slower than with HT disabled. Compared to Linux where I got about 20% faster compilation with HT enabled and use make -j8. 3. I never used IE8 on XP as I never upgraded IE6 from basic install. I was always IE hater from ancient times of Win 3.1 where I installed Netscape Navigator and later go with Mozilla till now...

Hm, still nothing? I can see there's kernelex for Win2k - is it capable to run some Vista-only app? Or just brings XP-only apps to W2k? I would wonder if W2k gets better support than XP...

BTW I use the PAE patch for about 1 year and dind't find more problems than I described here before. Just only one a bit uncomfortable behavior that may be linked with this patch - when I plug in USB some flash disk that I use 1st time then Found new hardware wizard appear and I need to point the usbstor.sys driver and install it again (of course it's already installed and usbstor.inf is present in INF directory). When I plug this flashdisk again it's recognized automatically and no more Wizard (it probably store it's s/n to registry). On my notebook with less RAM where I didn't applied PAE patch I never see wizard for any flash device and it works automatically. I suspect the problem may be caused by using usbstor.sys from Win2k3 server (I changed INF too) and driver signing. I tried to set This Computer|Hardware|Driver Signing to Ignore but any change. I also tried other tips found for this issue but still appeard wizard for new flash. Do you, other users who replaced usbstor.sys, observed same problem? (you need to try with some new flash disk not one you already use).

I wouldn't wonder if MS have such intentions pushed to manufacturers. But I can imagine that DOS is used in embedded and industry more than Win9x so there may have more demand for DOS drivers. BTW DMP Vortex86EX SoC also have DOS packet driver as they support FreeDOS as development platform. Also it's more easy to write a packet driver than some VXD/WDM hell for Win9x so I understand that manufacurers don't want waste time for a few remaning Win9x users. Fortunately Win9x can cooperate with realmode DOS drivers so problem solved And yes, I read an article with quotes of some MS-man about how strong they are pushing Win10 against all existing OSes, including Win7 and 8.x. They are doing all they can to force users to update, even talking ambiguously about older OS quality and security. And the time is playing for them. After a few years the support for Win7/8 will end and it fall into the same trap of HW incompatability like Win9x/2k/XP is now. They simple want to get all users into new cage of SaaS (System as a Service) to gain more control of user system and data. But who would wonder in nowdays f*kin world, they just learned from Google and Apple If I wouldn't need to use a lot of specific SW and HW linked to MS platform at work I switch to Linux but now it's not alternative for me. And virtualization is also not 100% working solution...