chromatic47

So I was doing some registry pruning and once again stumbled upon this oddly titled key and value. Every time I see it it makes me go hmm. The full key is this: [HKEY_LOCAL_MACHINE\Software\Microsoft\J9<000000] "é "=hex:05,40,45,45 ...and it has been there since a clean install of Win98se. Google turns up nothing, so does anybody know, what gives with J9, and why would it ever be less than 000000? And further, what is up with that funky binary value? Somehow it seems a shame to delete this bit of cryptic until I understand its inscrutable purpose.

nm

I can confirm Opera 9.27 runs on Win95a (SP1 + DUN 1.4 + Winsock 2).

You were right -- I installed the v77.72 driver (with widescreen modes added to INF) and now have 1680x1050 DVI, full AGP functionality, no shutdown/restart problems, and Opera is flying along. It was not without some detective work. On this system, the trick to getting DVI to function smoothly was to first revert to VGA, clean out the v82.69 driver, and uninstall the monitor. Then re-install the monitor using analog connection while still in VGA. Next install the Nvidia v77.72 driver, and lastly again install the monitor using DVI connection. One difference from v82.69 is the functioning of the Monitor Scaling option under DVI. Monitor Scaling will not accept 1680x1050, resulting in either a blank screen or NVMODE.DLL errors. Other resolutions work fine with Monitor Scaling, and the remaining scaling options do support 1680x1050 (Display Adapter Scaling, Fixed Aspect Ratio, and Centered Output). I leave it set mostly on Centered Output so that the signal will not be going anywhere near the Nvidia scaling algorithms. If I do need to change to a lower resolution (and see it full screen) I switch to Monitor Scaling which produces image quality visibly superior to the driver's scaled output. (This is typical; hardware scaling is usually preferable to driver scaling, assuming it's a quality monitor.) Long story short, I am pleased. Thanks for your help.

I would test v77.72 immediately if it is certain to provide DVI 1680x1050. It was tricky even with v82.69 -- the monitor was juddering badly and only after much rebooting and random fiddling with the control panel did everything settle in to a great picture. Could be a finicky monitor. Unfortunately I have no idea how I fixed it and basically am hesitant to go through that again, unless it's a sure thing. That said, though --- I know I'm still gonna try it. Will report results.

Dave-H, thanks, I tried switching memory cache off and the crash persisted. But it did get me thinking about memory, because it turns out that Opera 9.27 runs fine under the native VGA driver. (Although, browsing the internet at 640x480 is kind of like looking at the world through a chink in a brick wall. ) An internet search "nvidia crash" led to gaming forums where similar crashes had occured. Apparently with later Nvidia cards, enabling AGP Fast Writes and/or Sideband Addressing can lead to instability "on some systems" (including xp and vista). So I disabled Fast Writes, and I have now been surfing without a hitch for an hour. Yeah! whereas before the tweak, I would have gone through about twenty crashes by now. So far so good .. Opera 9.27 is running great. EDIT: Hours later, still no crashes. Appears to be a fix. Side-note: Oddly the Nvidia AGP panel shows OS as not supporting Fast Writes ("No"), yet says Optimal is "Yes". Also it will not accept a change to either the Fast Write or SBA settings. The checkboxes can be cleared, but are checked again on re-open. I ended up using RivaTuner to force the setting at startup: Sideband Addressing=Yes, Fast Writes=No Info here: http://www.techspot.com/tweaks/geforce/

Symptom: Opera versions 8.54 and 9.27 crash the system after starting a browsing session. Sometimes immediately, sometimes after 5-10 minutes. The crash blanks the screen and locks up the computer. Only means of recovery is reboot. This crash always happens, with both versions; the only thing unpredictable is the length of time that elapses before it happens. Both 8.54 and 9.27 can be opened and will stay running as long as a session is not initiated. But once mail is accessed, or browsing begins, it's only a matter of time before the screen suddenly goes blank and the system freezes. Firefox 2.0.0.18 and OffByOne do not crash. And both versions of Opera run just fine on my old Win95a P233mmx (albeit 9.27 is rather slow, but 8.54 is a smooth performer). Below are some system specs, although I fail to see how the setup is a limiting factor, since FF and OB1 have no problems: - Win98se using 98lite Sleek v2 shell (Explorer.exe v4.00.950 + Comdlg32.dll v4.00.951+ Shell32.w95 v4.00.951; Win98se Shell32.dll v4.72.3812.600) - Original Win98se DUN - DSL (via Davicom LAN) - Soyo motherboard w/Intel 845 chipset, P4 1.6gHz overclocked to 2.0gHz, 512 meg RAM - nvidia FX5200 AGP video card, 82.69 driver - various processor intensive software performing flawlessly (e.g. Photoshop, Cool Edit Pro 2.1, software synthesizers etc) Has anyone seen this kind of crash before or have a clue what could be causing it? I'm stumped.

Thanks for alerting me to the 77.72 drivers. I may end up trying them and if so will report back with results for anyone with a similar question. But first I am going to try fiddling with the PCI Bus device enumeration options (Use Hardware or Use BIOS -- currently set to Use Hardware), and maybe the IRQ steering options too ... ya never know what may come from blind thrashing about. (Of course with prior backup of the registry, and a fresh Windows clone just to be safe!)

Thanks Multibooter. Same here, those selections all work. So I am thinking about two consistent behaviors of 82.69 driver on this system: 1) When restarting Windows without reboot (press Shift + select Restart) the system freezes at the "Windows is now restarting..." screen after considerable HD activity, apparently just prior to desktop display. 2) Additionally there is new behavior during a cold boot or warm reboot -- just prior to desktop display the monitor goes into "no signal" (standby) mode for about 3 seconds, during which HD activity continues, and after which the signal returns and the Windows desktop appears as usual. Given those two behaviors the problem appears to be somewhere in the difference between driver loading and/or device enumeration during bootup starting with BIOS initialization, and during restarting Windows after BIOS is already initialized. Please forgive if I am not using correct terminology. But I think the logic is correct. I wish there was a way to log every single step during both processes, but even the step-by-step option in the boot menu only goes so far and will not let me pinpoint the step at which the monitor blanks just before the desktop appears. Need more knowledge.

With v82.69 driver and FX 5200 card I have no shutdown problem. Reboot also works. But Restart Windows freezes up, as does typing WIN to restart Windows from DOS. Do the v77.72 drivers also permit these latter to function normally?

Well I finally settled on driver version 28.32, which provides both widescreen and the main control panel functions (color correction, D3D, OpenGL) but not video overlay or the NView dual monitors or multiple desktops -- neither of which matters for this machine. The tray icon also will not function but I would never use that anyhow, since I use Multires to change resolutions. The reason I still use Win95 on this machine is that the motherboard has ISA slots and one of my favorite sound modules -- the Roland RAP-10 -- is an ISA card. I could upgrade to Win98se but basically it would be more trouble than it's worth since the mobo has maximum memory capacity of 128 meg. So this box is essentially a dedicated MIDI controller, with Cakewalk Pro Audio 8 running just fine. I pipe the RAP-10 audio along with external synths through a mixer to another machine which has Win98se and P4 processor, 512 meg memory, and Cool Edit Pro. It's an efficient recording setup, since Win95 handles MIDI very well and is rock solid and fast. Thanks again for your input, I will try the user32 substitution from sheer curiosity.

Yeap all updates have been applied, along with DirectX 8.0a (the most recent compatible with Win95, as far as I know). Also, my mistake, I think of "b" and "c" as SP's but that is not correct terminology. I did discover one item of interest, Dependency Walker reveals that nvqtwk.dll (the systray icon dll of several early drivers) is linked to a missing export in user32.dll -- [ChangeDisplaySettingsExA]. Which probably explains why the tray icon just sits there with no resolution menu. At any rate, it looks like I won't have to actually install any drivers I test -- I can just extract the UI dll's and run them through Dependency Walker first to see if they are total friends with Win95a. btw thank you all for your replies. I don't know of any other forum to ask such a question and not have the solution presented as "Why the h*ll are you still using Windows 95a?!?" lol

luke, Win95 handles widescreen fine on this machine. I change resolutions often, going back and forth from 1024x768 to 1440x900 to 1680x1050 with ease. The only problems are with the user interface. glock_94, Thanks I am trying to save myself a lot of trial and error by this posting and get either a fix or a specific driver, but I may just have to keep auditioning. I will start with the high numbers this time. But I suspect that the UI while stated "Win95 compatible" is actually Win95 SP2/3 compatible, not SP1...

System: Windows 95a, nvidia GeForce2 MX 400 Is there a best nvidia driver version for Windows 95 SP1 and a GeForce2 card? I am trying early drivers on the theory that they would be more friendly to early Windows, but so far that has proven to be a bad theory. Several of the early drivers do work fine at the hardware level, giving sharp 1680x1050 and troublefree resolution changes, but the display properties extension and control panel applet are partly disfunctional. Clicking for instance on the nvidia Overlay tab in display properties causes the computer to lock up, but other tabs work okay. The system tray icon can be made to appear, but has no options menu. And depending on the driver version, the Win95 DirectX control panel applet produces invalid page faults in nvdd32.dll. I have tried even the earliest version on their FTP site (ftp://download.nvidia.com/Windows/) which is 12.41, and also 21.81 through 28.32. If any of these drivers should be working, then methinks there is a problem in the OS files... Any advice appreciated.

Directory Watcher - keeps track of changes to files in any number of specified directories. The list of changed files updates each time the program is run, and can be saved as a text file. I find it handy during installs, to see quickly which dll's they be messin' with , how much did the registry grow by, and what new files got added. LAST - FREE - Directory Watcher 1.2 --- http://homepages.pathfinder.gr/nsiolios/dirw10.zip And the download page: http://homepages.pathfinder.gr/nsiolios/dirw.html

Got an nvidia FX 5200 here too. It is a good card for 2D, I can't report on 3D/gaming capabilities because the main use of this machine is photo editing and graphics work. Bought it recently from ebay, "new in factory sealed box" (which turned out to be true in this case), mainly because it is passively cooled and can do DVI widescreen. My previous card the venerable TNT2 can display 16:10 also, but only analog. In combination with the 82.69 driver the FX 5200 has been well behaved, produces stunning color, and the DVI quality is superb. Plus it was $20.

EDIT: Deleted irrelevant suggestion about vcache. I have since learned that there are fundamental differences between Win95 vcache and Win98/ME. Per this article: http://www.aumha.org/win4/a/memmgmt.php

Excellent! Glad I could help, and glad it was the simplest solution.

There won't be any problems unless bmps are already associated with a program. To check it, look at [HKEY_LOCAL_MACHINE\Software\Classes\.bmp] and see if there is a Default string in the right side panel. If there isn't, then .bmp isn't associated with anything (and also why you can't find BMPfile) and you can safely Merge the code that CharlotteTheHarlot posted, or do it manually as I described. (Changing [HKEY_CLASSES_ROOT] will also change [HKEY_LOCAL_MACHINE\SOFTWARE\Classes] and vice versa.) If there is a default string, then bmps are already associated with a program (and also you won't find BMPfile). In that case you can still get the preview icons and keep the Open behavior but it takes a different approach, basically creating the bmpfile key and adding along with DefaultIcon key, a Shell\Open\Command key pointing to the program of choice. I can describe in more detail, it's still easy to do manually and can also be done with a .reg file Merge.

lol Plus thanks for showing how to do it with a .reg file .. it's good knowing how to do it manually, but for setting up a new machine with many icons it gets a little tedious.

For bmp, in the registry add a DefaultIcon key to the bmp filetype and set the Default string value to [%1] without the brackets. If DefaultIcon key already exists just change the Default value to [%1]. For ex. my bmp filetype is "bmpfile", and the reg key is: [HKEY_CLASSES_ROOT\bmpfile\DefaultIcon] If "bmpfile" isn't listed, look at the .bmp extension [HKEY_CLASSES_ROOT\.bmp]. If there is no Default value, but only double quotes "", change it to [bmpfile] no brackets. Then add a new key to [HKEY_CLASSES_ROOT], name it [bmpfile] and proceed as above. Same procedure with ico -- look at extension, locate filetype, add DefaultIcon key and [%1]. The %1 hack only works with bmp and ico. But for jpg, tif etc you can still assign different icons by adding or changing DefaultIcon key at their filetypes and point the Default string value to something like [C:\my_icons\rubber_ducky.ico] without brackets. Sometimes if extensions are already associated with say Irfanview or whatever graphics program, you would need to create a new filetype for them. For example my jpegs have a camera icon and I had to make a new filetype key, "jpgfile" and change the .jpg extension to point to new filetype "jpgfile" instead of old filetype "Irfanview" or they would still show the Irfanview default icon. Hope this helps .. it's simpler than it sounds.

Sounds like it is sending the same audio to the output jack that it sends to the internal speaker, which would normally be L+R mixed to mono. To test this -- if the balance control is set hard left or hard right and there is still output at the one audible external speaker (maybe a little less in volume), then the chipset is sending mono, and is sending it to only one side of the output jack. Theoretically you should be able to switch mono output to stereo via the Windows mixer (sndvol32.exe) depending on what controllers the OPL driver provides. If not, then the computer unfortunately wins, unless you can find a later driver with more config options. But if on the other hand the balance control panned hard left or right does kill the audio from the one working external speaker (or severely cuts the volume) then it is a different problem. There is a stereo signal present, but one channel isn't working. The problem then is most likely the output jack itself, possibly needing nothing more than a cleaning and/or slight bending of the contacts to get a good tight connection on the stereo plug. Those mini/micro jacks sometimes need attention after they get some miles on them. Also -- if you're using a size adapter at the output jack, of course make sure it's stereo-to-stereo .. I've goofed on that before. Also -- just in case, I would check the Libretto's BIOS for any audio config options. HTH

Well ... nutz. Guess I'll have to consider it penance for viewing enjoyment. Widescreen analog I could coax out of the 56.64 drivers by adding a 1680x1050 key to the registry display modes, but that trick won't work with DVI. (Actually I forced analog widescreen out of the 29.42 driver too, with the same hack, but that was on a TNT2 card.) Anyhow thanks for the bad news ..

Excellent. And, what you found about vflatd.vxd jogged my memory about certain drivers being bundled into vmm32.vxd instead of loading separately. There is a way to tell in Device Manager which are bundled, and if they are then placed separately into windows\system and-or windows\system\vmm32 it can cure problems with various devices, and also boost system efficiency and free up some resources and memory. Some say it helps, some say it doesn't .. but it obviously worked for you. Check this old thread from MSFN: http://www.msfn.org/board/98-fix-t12005.html (Especially the posts by Shotgun.) Now thanks to your solution I can make some possible improvements too. B)

Your new info brings up another possibility -- unless there is onboard video on the motherboard, or unless there is another video card plugged in, there should only be the single 0000 display key. No other is necessary. If that's the case (only the ATI adapter is available) then redundant keys can be deleted: 1) In regedit go to HKEY_LOCAL_MACHINE\Enum\PCI. Locate the VEN_ key for the video card and look at the string value for Driver (right-side panel under the BUS subkey). If it doesn't say "DISPLAY\0000" then change it to 0000. (Again, the only reason it would need to say 0001 is if the motherboard has onboard video or if there is a second video card occupying a slot.) 2) Go to HKLM\...Services\Class\Display and delete all keys but one complete key referencing the ATI card. It should be a complete key that contains the original minivdd entries. If the "untouched original values" key is number 0001, rename it to 0000. So after this action there is only one Display key, 0000, with all original values. Now Enum\PCI points to the driver in Display\0000, and Display\0000 points to the ATI driver with no redundant keys mixup possible. This should make all resolutions available again. If Windows tries to re-install the ATI driver on reboot -- let it -- but then repeat the above procedure, but this time keep only the highest numbered Display key (the one most recently installed), rename it to 0000 and rename the Enum driver entry to DISPLAY\0000 as before. Reboot again and all should be well with no install prompt. Hope this helps. If it doesn't, there may be a hardware failure because we have just told Windows how to get from card detection to proper driver in a straight line. EDIT: Nope not hardware failure, I just saw your results using higher VESA modes in DOS. If 800x600x8bit is working in DOS it should work in Windows too.