According to https://people.freebsd.org/~kato/cpuident.html all 486BL have MSRs.

debs3759 wrote on 2023-01-05, 13:11:

I don't have info for detecting 486BL or 486BL2, but 486BL3 has MSRs.

I suspect DIR0 needs to be used, SIV should have read this, but I can't be sure it did 'till I check Menu->Hardware->CPUID->CPU-0 in the same file.

red-ray wrote on 2019-08-22, 10:17:

As Windows 9X does not kernel mode support structured exception handlers (__try { … } _except() { … }) in VXDs SIV can't sensibly use a VXD to read MSRs so struggles with Pentium III and later CPUs on Windows 9X.

debs3759 wrote on 2023-01-05, 13:36:

According to https://people.freebsd.org/~kato/cpuident.html all 486BL have MSRs.

Given W9x does not support Kernel Mode SEHs and AFAIK the IBM 486BLs don't support CPUID so how does this help?

I wonder what happens if an RDMSR is done in User Mode and it's not supported, maybe the exception is other than GP, getting Invalid Opcode Exception (#UD) would be good.

After pondering I decided to see what happens with my Rise mP6 (Lynx) which does not have MSRs. On 98SE the RDMSR failed with 0xC000001D: Illegal Instruction, but on 2003 it failed with 0xC0000096: Privileged Instruction so I guess 2003 must have emulated the RDMSR, as a result SIV decided MSRs were supported which is not what I wanted 🙁

mastergamma12 wrote on 2022-12-31, 18:40:

Looks good now

Thank you for confirming this, I have just uplaoded SIV 5.68 Beta-05 so please use that.

debs3759 wrote on 2023-01-05, 13:11:

I don't have info for detecting 486BL or 486BL2, but 486BL3 has MSRs. BL and BL2 may, but I'd have to research or test that.

Mine is actually a BL3! It's just underclocked with a modified BIOS to set the multiplier to 2 instead of 3 so I could reliably get it to run with a 33MHz bus. Sorry for the confusion! So yes, the MSRs would be a good way to check for it in any environment that wasn't Win9x. I'm no developer so I wouldn't know -- my modified BIOS was hacked by a friend, I just burnt it to an EPROM and stuck it in the board. From what I understand about the instruction timing, if you've got a datasheet you're doing better than me, but the 486BL behaves much more like a 386 than a 486 in that regard, which makes sense given it's in the same package and uses the same FPU and such.

I'm not actually all that fussed about the incorrect video mode -- I'm aware it's a Windows bug, it also appears when using WMI to poll system info. I find it more amusing than anything, and while on anything faster, 98lite is faster than 95, it is simply not here so I won't be reinstalling it. I'd have more RAM in this, but with the peculiar way the 486BL handles cache lines since I'm given to understand it inherited that logic from the 486/386SLC with their 24-bit address bus, it can't cache anything above 16MB even though it will run with more.

With the 5.68 beta you linked me to, it saved the text file no problem, though the workaround you suggested also probably would have worked since the change looks to just be that it doesn't try to save in UTF-8 by default. It still does not correctly pick up on my video device, though I didn't really expect it to. Humorously, it somehow turns off my serial mouse while it's saving the sections. I have to reboot to get it to turn back on. Probably resetting the UART or something if I had to guess.

Checking the CPU frequency under the help menu using DIV just said it didn't support RDTSC. Checking with the MHz option gave a speed of 58.1MHz.

Looking at the text file, it mentions a Pro/100 ISA I'd removed to put a 3com card in a couple days ago since that Intel card wasn't actually using Intel hardware and drivers were thin on the ground especially for anything other than Windows.

I do actually have the ability to set up Windows NT 3.51 on a spare SD card, probably over the weekend. Would that provide more useful information?

Sedrosken wrote on 2023-01-05, 20:10:

Checking the CPU frequency under the help menu using DIV just said it didn't support RDTSC.

Thank you for using 5.68 Beta-04 to generate the save files. DIV is nothing to do with measuring the CPU frequency, it's there to compare the speed of 32-bit vs. 64-bit integer divides.

I suspect it's Menu->Hardware->H/W Detail->Serial that triggered the serial mouse lockup, is it? If so I will try and figure out why and adjust things.

SIV did not read DIR0 + DIR1 at all so I need to figure out why. I will add some trace code to a test SIV which I hope to post tomorrow.

red-ray wrote on 2023-01-05, 20:47:

Sedrosken wrote on 2023-01-05, 20:10:

Checking the CPU frequency under the help menu using DIV just said it didn't support RDTSC.

[...] I suspect it's Menu->Hardware->H/W Detail->Serial that triggered the serial mouse lockup, is it? If so I will try and figure out why and adjust things.

Yup, just tried it and sure enough that's exactly what does it. My mouse just croaks as soon as it pulls up that window. It's kind of funny, actually.

red-ray wrote on 2023-01-05, 20:47:

SIV did not read DIR0 + DIR1 at all so I need to figure out why. I will add some trace code to a test SIV which I hope to post tomorrow.

Looking forward to it. Like I said, I'll see about getting NT on a spare card to provide better information. Would 4.0 work better or would 3.51 be adequate? I only ask because 4.0 would be quite slow with this configuration.

Sedrosken wrote on 2023-01-06, 00:47:

Would 4.0 work better or would 3.51 be adequate?

Thank you for for confirming [Serial] triggered the issue, in the attached SIV32L 5.68 Hope-06 I hope I have fixed this, have I? If not you could do Menu->File->Save->Save Selected and deselect [serial] to circumvent the issue.

NT 4.00 has several advantages over 3.51 with the main one being is that it supports Kernel Mode SEHs meaning SIV can sensibly read MSRs. It also supports CP_UTF8, correctly handles SIVRES.dll and with SP5/SP6 you can have much larger disks. Back in the mid 90s I moved from NT 3.51 to NT4.00 on an i486 DX-50 ISA only system with 20MB and seem to recall speed wise they were much the same.

I have also added some trace code, please do SIV32L -DBGCPU and post the generated SIV_DBGOUT.log in addition to a new save file. With luck it will report IBM 486-BL and get the correct speed, but I expect I will need to make further adjustments.

I just hope SIV doesn't now report Cyrix CPUs as IBM 486-BLs, so a save file from a Cyrix model 4 CPU based system would be good.

red-ray wrote on 2023-01-06, 13:15:

Thank you for for confirming [Serial] triggered the issue, in the attached SIV32L 5.68 Hope-06 I hope I have fixed this, have I? If not you could do Menu->File->Save->Save Selected and deselect [serial] to circumvent the issue.

It's not that big of a deal honestly. I just reboot and it's fine. Under NT, I have no working mouse anyway, so it's moot. It's quite annoying, actually; I use a Genius NetScroll serial mouse, and while DOS/Windows 3/9x are okay with that with a driver -- usually the generic 4D Mouse Driver under 9x, the actual NetScroll software under Win3, and CuteMouse under DOS) -- none of those Windows drivers appear to be working for NT despite the fact that the software is supposed to support it. I'll need to buy an Intellimouse that's serial-compatible, probably. Navigating NT with a keyboard was a humongous pain, not just to run the software but to install it, the archival software needed to extract SIV, drivers for the display so it could show the whole window, and SIV itself. I didn't bother with SP6a as that'd take an eternity to both locate over my network and install.

[...] Back in the mid 90s I moved from NT 3.51 to NT4.00 on an i486 DX-50 ISA only system with 20MB and seem to recall speed wise they were much the same.

Yeah, it doesn't seem much slower than NT 3.51. Interestingly while NT is, on the whole, slower to boot and run than 95, disk performance is better and sound doesn't skip. Unfortunately no Trio64V+ VLB driver seems to exist for NT and the stock S3 driver fails to load -- that's better than on 3.51, where it loads but is incredibly unstable and corrupts the display at the drop of a hat. The stock VGA driver also likes to corrupt the display every now and again. The more I deal with it the more it seems the Trio64V+ was never really intended for VLB, despite the fact that it performs well in DOS and Win9x. I wonder if I might have a better time with a Cirrus Logic card of some sort -- It's not like I have so much CPU performance that I really need one of the best for the platform. Unfortunately I don't have any other VLB cards and I'm not willing to pay eBay prices for another more compatible one.

I just hope SIV doesn't now report Cyrix CPUs as IBM 486-BLs, so a save file from a Cyrix model 4 CPU based system would be good.

Don't happen to have one, that'll have to be someone else's purview.

Unfortunately as you'll see from the "screenshot" and the output file, it doesn't seem anything's changed. It's still picked up as an Intel DX-25/33 at ~58MHz. I couldn't actually pull anything from the framebuffer with PrntScr with the stock VGA driver, so you have to put up with my poor cameraman skills. I tried the -DBGCPU parameter, but it doesn't seem to make it do anything, it certainly doesn't put an output file anywhere.

Sedrosken wrote on 2023-01-06, 19:48:

I tried the -DBGCPU parameter, but it doesn't seem to make it do anything, it certainly doesn't put an output file anywhere.

Thank you for checking out 5.68 Hope=06. Please may I also have C:\WINNT\PROFILES\Administrator\DESKTOP\SIV32L\SIV_JST-CRAGSTONENT.dmi which is a dump of the BIOS and with it I may be able to better report the motherboard. What does the post screen report?

From a command window please do SIV32L -DBGCPU -DBGSDM > SIV_DBGOUT.log | more and see if that generates SIV_DBGOUT.log with the trace information.

I was only expecting you to test on W95C and have not added CPU detection code for NT 4.00, what does SIV report on W95C?

If you install the Tahoma Font I suspect the SIV panels will look better.

I had an Orchid Kelvin 64-ISA in my i486-DX50 system which worked well on NT 3.51 + 4.00, but they seem to cost an arm and a leg !

The attached SIV32L V5.68 Blue-06 should do better on NT 4.00.

BTW if you press [ Copy ] it will copy the panel to the clipboard and you can then paste it into such as paint.

red-ray wrote on 2023-01-06, 21:00:

Sedrosken wrote on 2023-01-06, 19:48:

I tried the -DBGCPU parameter, but it doesn't seem to make it do anything, it certainly doesn't put an output file anywhere.

Thank you for checking out 5.68 Hope=06. Please may I also have C:\WINNT\PROFILES\Administrator\DESKTOP\SIV32L\SIV_JST-CRAGSTONENT.dmi which is a dump of the BIOS and with it I may be able to better report the motherboard. What does the post screen report?

VOGONS doesn't support .dmi, hence the archive. POST screen reports "486BL CPU 66.5MHz".

From a command window please do SIV32L -DBGCPU -DBGSDM > SIV_DBGOUT.log | more and see if that generates SIV_DBGOUT.log with the trace information.

Done, both on NT and 95, though 95 is using the older Hope06 build and NT is now using the Blue06 build you linked with your last post. I'd have attached the output from 95 doing the same command line, but VOGONS only allows 5 attachments per post.

I was only expecting you to test on W95C and have not added CPU detection code for NT 4.00, what does SIV report on W95C? [...] The attached SIV32L V5.68 Blue-06 should do better on NT 4.00.

While I'm happy to report NT4 is now showing correctly, 95 is not. I could nitpick and say that the Blue Lightning is QFP, not PGA, but there's no way for the software to know that. Bus speed is actually 33MHz, but everything seems to misdetect that.

If you install the Tahoma Font I suspect the SIV panels will look better.

I'm sure they will, but I'm still navigating with a keyboard. I'll worry about how it looks later.

I had an Orchid Kelvin 64-ISA in my i486-DX50 system which worked well on NT 3.51 + 4.00, but they seem to cost an arm and a leg !

While I'm inclined to say that's just how it is now, it seems you can still get deals if you know the right people. My Trio64V+, despite being a fairly rare VLB card, only cost me some 40 bucks a year or so back.

In other news, I sifted through eBay listings for VLB GD-5426s, 28s, and 29s -- the results were not good. The best listing I found was around 60 dollars US after shipping for an "untested" 5426. I would have been tempted had I not known by now that "untested" on eBay means "definitely tested, doesn't work" these days.

BTW if you press [ Copy ] it will copy the panel to the clipboard and you can then paste it into such as paint.

When I do either that or press PrntScr, what I end up pasting into mspaint is pure white. I suspect it's a quirk of the stock VGA driver. Nothing else works on NT4 with my video card, you understand.

Sedrosken wrote on 2023-01-07, 03:58:

Done, both on NT and 95, though 95 is using the older Hope06 build and NT is now using the Blue06 build you linked with your last post. I'd have attached the output from 95 doing the same command line, but VOGONS only allows 5 attachments per post.

Thank you, things are now starting to make sense. I have attached SIV32L 5.68 Disk-06 which should fix the FSB speed + Disk SMART reporting on NT. The current state is:

1. I need the SIV_DBGOUT.log file from W95C to be able to figure out why 486-BL is not detected. In the NT4 one there is SIV32L CPU-0 EC C0000096, the W95C should have the same, but I suspect it's different.
2. The disk SMART reporting should have worked and does on my NT4.00 SP6a systems, I have tried to work around this, but suspect if you installed SP6a it would just work.
3. Looking at Menu->Help->SIV Lookup I can see the dbghelp + psapi DLLs are missing, please put the ones from Tahoma Font+DBGHELP+PSAPI DLLs for NT4.zip onto the system.
4. I suspect many of the SIV tooltips report COMCTL is too old, to address this you would need to install IE5.5 which installes a later COMCTL. I will also install the missing wininet.
5. I was hoping the BIOS POST screen would report the motherboard model and BIOS version, does it and if so what are they?
6. I would put all the attachments in one .ZIP file assuming it's <= to the 5MB size limit.
7. I can't see why a GPU is not reported on NT4.00 so please use the SIV32L -DBGCPU -DBGGPU -DBGSDM > SIV_DBGOUT.log | more command.

1. Ask and ye shall receive -- bear in mind this is from when I ran Hope-06, I haven't swapped back out to the 95 card but suffice to say it's not quite as simple as just swapping the card.

2. 3. 4. After spending 2 hours (yes, it really did take that long for everything) the NT install is "up to spec" as it were. I went a step further and installed IE6, actually. On 95, IE 5.5SP2 was already installed. I did not and would not opt for the "enhanced" desktop on either.
5. Microid Research MR-BIOS 1.65 modified to run the 486BL at 2x instead of 3x. Board is an Alaris Cougar with the OPTi 499 chipset.
7. Included also is the output from that, bundled with the new text output in winnt.zip -- but remember, stock 16-color VGA driver. It makes sense (at least to me) that Windows wouldn't really have much of an idea what's installed, especially on VLB which I'm given to understand was quite the hackjob of an interface.

Sedrosken wrote on 2023-01-07, 21:36:

Board is an Alaris Cougar with the OPTi 499 chipset.

Thank you for adding the DLLs , updating the NT 4.00 system to SP6a, checking out the latest test beta and posting the results. With SP6a it's two steps forward and one step back. The current state is:

1. On W95C SIV32L CPU-0 EC C0000005 was traced which is STATUS_ACCESS_VIOLATION. Getting this for an RDMSR instruction makes no sense at all, is different to what NT 4.00 signals and I suspect it's a bug in W95C. I will check what happens on a few of my W9x systems tomorrow and then try to then come up with a plan.
2. Until you install a Graphics driver SIV won't be able to report a GPU on NT 4.00.
3. I think the change from SP1 to SP6a caused HKLM\Hardware\Description\System\CentralProcessor\0 to be different, please may I have a registry export of HKLM\Hardware\Description so I can check what is there?
4. OK about the board being a Alaris Cougar, but does the system report this when it posts? I searched the .DMI file for Cougar and it's not there. The new test SIV should report IBM OPTi 82C499 and V1.65.
5. The disk SMART reporting is still not working and I suspect there must be an issue with the SMART_RCV_DRIVE_DATA for IDE_COMMAND_IDENTIFY so I added some more trace code.
6. Using the attached SIV32L 5.68 OPTi-06 please do SIV32L -DBGCPU -DBGSDM > SIV_DBGOUT.log | MORE and post new .TXT + .DMI + .LOG files.
7. I noted the reported speed is slightly lower so please post what Menu->Help->Conversions->CPU MHz reports, to do this press [ Copy ] then paste the text into notepad.

BTW I just found the OPTi 82C499 datasheet.

Sedrosken wrote on 2023-01-07, 21:36:

I haven't swapped back out to the 95 card but suffice to say it's not quite as simple as just swapping the card.

After pondering this it does not make much sense as all my W9x/WMe system are multi-boot with just one SD card/SSD. I usually put W9x on a FAT16 2GB C: then have NT4/W2K/WXP/2K3 on NTFS D:/E:/F:/H:. I noted you only have a 2GB SD card, is this the maximum supported disk size?

I was about to add Menu->Hardware->MSRs->MSRs IBM to dump the IBM BlueLightning MSRs then I noticed it was already there! I added it for SIV V5.41 released on 14-Sep-2019 so must have been looking at IBM 486BL support back then. I used this to check what happens on several W9x systems and they all report C0000005 (STATUS_ACCESS_VIOLATION) even for CPUs that support MSRs. Given this detecting an IBM 486BL on W9x is a catch-22 situation so I don't think I will be able to get SIV to correctly report your IBM 486BL on W9x and suspect no other programs will be able to as if they were run on a real Intel i486 doing a RDMSR in kernel mode would crash the system.

Sedrosken wrote on 2023-01-05, 10:40:

Wow, XP on a socket 8 rig, you must have an awful lot of patience. I found even 2000 to be nearly unusably slow on my PPro rig when I still had it, and it was a 1meg black-top chip running overclocked to 233MHz. I know you've got the PIIOD there, but even at 333MHz I'd think XP wouldn't be a fun time.

Windows XP seems to run surprisingly well on my Socket 8 system, provided it's set to performance mode. The high performance SCSI RAID 0 system drive probably helps quit a bit though. I should hopefully complete the build in the next month or so and will release a video showing its performance once it's complete! (It still needs the Vodoo5 5500 PCI card as well as 768MB of EDO RAM) I should hopefully have the Voodoo5 in my possession in the next few weeks!

My primary reason for running XP on this machine was to avoid the multitude of hardware conflicts (DMA address conflicts between the Matrox G450, the SCSI cards, the SoundBlaster AWE64 and the network card) I had when trying to get everything setup using Windows 98! Now that I successfully have Windows XP installed, the system will stay this way, as XP's far more intuitive than 98!

Sedrosken wrote on 2023-01-05, 20:10:

It still does not correctly pick up on my video device, though I didn't really expect it to.

I have just uploaded SIV 5.68 Beta-06 which should fix the reporting none PCI GPUs on Windows 95.

It worked when I tested it on W95B with my Kelvin-64/ISA which is the only none PCI GPU I have, given this I would be interested to know if it works in general.

Irinikus wrote on 2023-01-10, 10:06:

https://siliconimage.irixnet.org/index.php/Us … Albums/Irinikus

Did you know this is broken? It reports "Dang... Something went wrong! We tried really hard, but it's broken.Talk to your Gallery administrator for help fixing this!"

red-ray wrote on 2023-01-08, 15:34:

Sedrosken wrote on 2023-01-07, 21:36:

I haven't swapped back out to the 95 card but suffice to say it's not quite as simple as just swapping the card.

After pondering this it does not make much sense as all my W9x/WMe system are multi-boot with just one SD card/SSD. I usually put W9x on a FAT16 2GB C: then have NT4/W2K/WXP/2K3 on NTFS D:/E:/F:/H:. I noted you only have a 2GB SD card, is this the maximum supported disk size?

A quick response just to this instead of test results this time, I've been super busy between work and housesitting so I won't be getting around to playing with it more anytime soon.

It'll use whatever you've got, the limit is what your BIOS will handle. I've got a 128GB card in my PIII build. My main SD card for the 486 with Win95 on it is 16GB of which I'm using about 8. The 2GB card I have for NT is just the only card I have below 16GB. The BIOS can't handle any more than 8GB there, so I'm using an XT-IDE 386+ ROM in my NIC to give the onboard IDE LBA support. As far as Windows and everything else is concerned this turns the hard drive into an int13h device similar to a SCSI drive, but the bad news for Windows NT and OS/2 and such is that you would need a driver that doesn't exist to use it that way, so when I switch to the NT card I have to:

-boot from floppy and disable the boot ROM
-then swap the cards in the adapter
-reboot into CMOS setup and configure the 'hard disk' there -- I use a DOS utility on floppy to pull CHS values

And if I'm swapping back to the main card, I have to do all of that again only in reverse. I find it mildly interesting/puzzling how Windows 9x can load protected-mode drivers for the disk -- it's not in DOS compatibility mode -- but there isn't a generic driver that'll work with it for NT.

Sedrosken wrote on 2023-01-13, 19:46:

The BIOS can't handle any more than 8GB

Thank you for the detailed explanation, I suspected there were some BIOS limitations, but would never have guessed the hoops you needed to jump through. I can now understand why 95C sees a 16GB C: when NT4 only sees 2GB.

OK about being rather busy and I will wait 'till you have time to run some more tests. I don't have an i486 system and keep feeling I should get one, looking on eBay the ones I fancy seem overpriced though.

red-ray wrote on 2023-01-12, 18:54:

Irinikus wrote on 2023-01-10, 10:06:

https://siliconimage.irixnet.org/index.php/Us … Albums/Irinikus

Did you know this is broken? It reports "Dang... Something went wrong! We tried really hard, but it's broken.Talk to your Gallery administrator for help fixing this!"

Thanks for the heads up!

I've informed IRIX Network of the problem!

I have just uploaded SIV 5.69 Beta-03 which has greatly improved support for Windows NT 3.51. On one of my test systems using the native shell I see:

This is not ideal and I feel this is mainly down to the issues with the native shell as after I installed newshell2 many things started working.

If you have an NT 3.51 system I would be interested to know how SIV 5.69 Beta-03 does, especially if it's running on an i386 or i486 CPU as I don't have these CPUs so have not tested with them.