Attached are the NexGen POST screens for you to verify.

Thanks for attaching the nt4shell for nt351. I'm very nervious to test this out and think I'll save it for a day that I feel like gambling. It was a challenge to get my NexGen NT35 system running smoothly and I don't want to jeapardise this.

Part of the issue with NT4 running on my SXL2 is due to: Any way to run a DOS TSR program before loading Windows NT 4.0?

I removed the PS/2 mouse and put on a serial mouse and was able to boot to the NT4 desktop, but control panel would not open. I could use the keyboard, but all my start menu programs were missing. VGA mode booting would yield only a black screen.

I'm pretty sure I know why NT4 isn't running properly on the SXL: certain registers need to be set for proper cache invalidation and the program I use for this is in DOS. Also, the SXL2 is running in 1x mode without running that program. For this particular setup, I normally run cyrix.exe -f -r -cd -i1 in the autoexec.bat file. Without -cd, it runs in 1x mode. Without -r, the CPU is caching the 640K-1M ROM shadow area. Without -i1 the L1 cache is essentially disabled. So it is probably more beneficial to get a Win95 installation running well.

feipoa wrote:

Thanks for attaching the nt4shell for nt351. I'm very nervious to test this out and think I'll save it for a day that I feel like gambling. It was a challenge to get my NexGen NT35 system running smoothly and I don't want to jeapardise this.

You are welcome, it's pleasing to be able to help and thank you for the screen shots.

As I recall installing newshell2 was pretty painless and it was easy to revert. The only issue I recall what that after installing newshell2 SIV tried to use it's NT4 driver and as it was obvious why this was sorted within a couple of minutes. AFAIK whole of newshll2 is user mode .DLLs so it can't affect the kernel. The only issue I can think of is if the new .DLLs are compiled for i486, but suspect they are not. I was using my Dual Intel Pentium II (Klamath) system so i486 code would just run. On the NexGen all you would need to do is run IDON.com, but I expect you can't. On start-up SIV already does what IDON.com does so it would be almost trivial to put that code into SIVNT3.sys and set the service to boot start if there was an issue.

I have just uploaded SIV 5.41 Beta-25 which should report your CPU as NexGen Nx586-PF110, does it? I have figured out how to report the L2 Unified Cache information, is it correct?

Do most NT4 programs run fine on NT351?

Attached is the SIV start screen. My board does have 256 KB L2 cache, so I guess it is correct. I beleive the board only ever shipped with 256 KB SRAM, SMD - but there are solder pads for 1024K and I know one extremest from CPU-World managed to upgrade his RAM to 1024K. Do you think that would get reported correctly? The NexGen board is too rare for me to risk such a mod.

EDIT: I have accidently pressed the 'OK' button twice now and it exited the program, not intending to Exit. I'm curious why you choose to call the button 'OK' rather than 'EXIT'? Intuitively, I'm expecting 'OK' to go to another screen.

I performed a fresh installation of Win95c on my SXL2-66 system. Without any updates installed, e.g. the unofficial service pack, windows installer 2.0, etc., I receive an error. Note that IE4 web integration was not installed.

feipoa wrote:

I performed a fresh installation of Win95c on my SXL2-66 system.

Thank you for trying SIV32L, you need to Install the vcredist_60_x86 that is attached to the first post and the Tahoma font would be good.

red-ray wrote:

By default SIV uses the Tahoma font which may not be installed on old systems. If it's not present SIV should still work, but I have attached a copy you could install.
If MSVCRT.dll is missing you will need to install the attached vcredist_60_x86.

feipoa wrote:

Do most NT4 programs run fine on NT351?

My board does have 256 KB L2 cache, so I guess it is correct.

EDIT: I have accidently pressed the 'OK' button twice now and it exited the program, not intending to Exit. I'm curious why you choose to call the button 'OK' rather than 'EXIT'? Intuitively, I'm expecting 'OK' to go to another screen.

You will need to ask their authors and suspect having newshell2 installed will improve the situation. I know CPUZ + HWiNFO don't have NT 3.51 drivers so you would have to get Franck/Martin to add them. I plan to try AIDA64 once I have installed 3.51.

Thank you for checking this out and it should be assuming the information in COMMON.INC is correct. Who has the 1MB cache and/or can you ask them to try SIV please?

It's been this way since 2001, I can't remember why I chose [ OK ] and currently I am not inclined to change it. If you don't wish SIV to exit do -TRAY -NOHIDE, see Menu->Tools->Configure->SIV Qualifiers.

After installing vcredist_60_x86, I run SIV32L beta25 and receive a BSOD, the same one I receive if I try to run CPU-Z. I received no errors while installing Windows 95c and I'm going to try installing some programs to ensure the system is OK.

feipoa wrote:

After installing vcredist_60_x86, I run SIV32L beta25 and receive a BSOD, the same one I receive if I try to run CPU-Z. I received no errors while installing Windows 95c and I'm going to try installing some programs to ensure the system is OK.

VDD is a video driver. Check the device manager for video error, and eventually remove the device from device manager and reinstall it after a reboot.

I don't know if win95C includes scanreg, but if it does this can help to fix the issue. Restart in MS-DOS mode, then type :

1scanreg /fix

I'm using the Microsoft default CL-5434 drivers, so perhaps I'll install the Diamond drivers.

feipoa wrote:

After installing vcredist_60_x86, I run SIV32L beta25 and receive a BSOD, the same one I receive if I try to run CPU-Z.

Thank you for trying SIV on that system and as Frank mentioned it's the video driver that is causing the issue and maybe both SIV and CPUZ are doing something that upsets it.

1. After pondering this I am wondering:
2. Running from a command window what happens if you do SIV32L -SAVE -DBGINI | SIV_DBGOUT.log | more, this will generate the SIV save file without using the SIV GUI.
3. How far does SIV get before the VXD issue when using the GUI? Do SIV32L -DBGINI | SIV_DBGOUT.log | more and post SIV_DBGOUT.log so I can check.

Can't check this yet. I'm discovering that everytime the system crashes, either from this BSOD, or from a soft reboot that it doesn't like, Scandisk is finding lost clusters and my Windows system files are being destroyed. I'm not sure if it is due to the SCSI2SD device, or more likely, due to the SXL2-66 and cache coherency issues. The SXL2 is a really tricky CPU to get setup on these old 386 motherboards due to the cache. The SXL2-66 has some flush circuitry on it, while the SXL2-50 does not. The SXL2-50 needs to use the BARB software method to flush the cache (when using the Adaptec 1540 DMA SCSI controller, which I am using) and needs to have each 64 K of each 1 MB boundary set as non-cacheable. The SXL2-66, which contains some glue logic, I noticed on Win3.11 that I could use the faster flush invalidation method and did not need to disable caching on the 1 MB boundaries. Something with Windows 95 and system hanging is messing up my system files.

I'm going to try the more conservative SXL2-50 flush settings. If this fails, I'll need to use a non-DMA SCSI controller, the 1520.

feipoa wrote:

I'm using the Microsoft default CL-5434 drivers, so perhaps I'll install the Diamond drivers.

There is a CL 5434-HC-C chip in the ISA GPU I use for testing and all is OK on both W2K SP4 + NT 4.00 SP6a using the stock Microsoft drivers.

I plan to install 2003 Enterprise Server, NT 3.51 and W98 over the next few days.

Are you able to try Win95?

I'm currently installing the unofficial SP1 update for Win95 OSR2

feipoa wrote:

Are you able to try Win95?

No, the system booted Windows 95B V4.00 Build 1212 B on 2019-06-20 @ 13:44:58, but when I select that option it reports "Invalid system disk" and I can't see why, any ideas?

I am about the swap the 2GB SCSI disk for an 18GB one and will try to get W95 going after the swap. Last time I booted W95 it had:

1[PCI Bus] <- SIV32L - System Information Viewer V5.41 Beta-26 RDO::ray
2
3Bus-Numb-Fun  IRQ  Vendor-Dev-Sub_OEM-Rev Class (255)         Vendor and Device Description
4
5[0 - 00 - 0]       8086-1237-00000000-02 Host Bridge          Intel 82441FX 440FX Pentium(R) Pro Processor to PCI Bridge (Natoma Rev 2)
6[0 - 07 - 0]       8086-7000-00000000-01 ISA Bridge           Intel 82371SB PIIX3 ISA Bridge
7[0 - 07 - 1]       8086-7010-00000000-00 EIDE Controller      Intel 82371SB PIIX3 EIDE Controller
8[0 - 09 - 0]       3388-0021-00000000-12 PCI Bridge (0-1)     PLX PCI 6140/52/6254 HB1/1-SE/HB6 Universal PCI to PCI Bridge (non-transparent mode)
9[0 - 10 - 0]  11   9004-8178-00000000-00 SCSI Controller      Adaptec AHA-2940U/UW/2940D Ultra/Ultra Wide/Dual SCSI Host Adapter
10[0 - 11 - 0]  10   1011-0002-00000000-23 Ethernet Controller  Digital 21040 Based PCI Ethernet Adapter
11[1 - 00 - 0]  9    102B-0525-0D41102B-82 VGA Controller       Matrox Millennium G450 Dual Head PCI
12
13Total of 2 PCI buses and 7 PCI devices in 0.030 seconds.  Device names from D:\SIV\SIVPCI.txt   Date 2019-08-31 Version 235 (124,792)
14
15[  OK ]  [ Copy]   [Windows]  [Machine]  [USB Bus]  [Network]   [ SPD ]  [HID Bus]  [Volumes]  [SMB Bus]   [PCI Bus]  [PCI Dev]  [ALL Dev]

It also has NT 3.10, but as you can see SIV is not all it should be and CPUZ-95 won't start giving Unexpected DOS error.

Okay, I've got the P133 up and running, unfortunately, given how temperamental it is, I've ran into a nice BSOD (of course I didn't note the error *facepalm*) when trying to download SIV.

Trying again right as of now...

EDIT: Here we are! I don't have a Voodoo or Matrox Graphics Card, all I've got is a puny S3 Trio64 on a tricky i430VX mobo that has got more than a few tricks up its sleeves 😦

Vynix wrote:

a tricky i430VX mobo

Thank you for checking out SIV, I though I had added i430VX support, but I only half added it!

The Motherboard 512KB (is the L2 cache 512KB?) comes from the i430VX MCH, but the memory timing should be there. Watch this space...

red-ray wrote:

Vynix wrote:

a tricky i430VX mobo

Thank you for checking out SIV, I though I had added i430VX support, but I only half added it!

The Motherboard 512KB (is the L2 cache 512KB?) comes from the i430VX MCH, but the memory timing should be there. Watch this space...

For the L2 cache, it has indeed 512kB on-board, I have no idea how much my mobo (it's a Shuttle HOT-555A if that can help!) has, might need to check that again.

Vynix wrote:

For the L2 cache, it has indeed 512kB on-board

I forgot to ask the memory type, is it FPM or EDO?

I added missing 430VX support and a logo, press [ 67MHz EDO RAM ] to see it. How does the attached test SIV32L look and does CPUZ report the same memory timing information on it's [Memory] tab please?

Yes it's EDO modules that my rig have currently, I'll try this new build tomorrow hopefully (that if I can get it to boot again, it's quite the temperamental rig...)

red-ray wrote:

feipoa wrote:

I'm using the Microsoft default CL-5434 drivers, so perhaps I'll install the Diamond drivers.

There is a CL 5434-HC-C chip in the ISA GPU I use for testing and all is OK on both W2K SP4 + NT 4.00 SP6a using the stock Microsoft drivers.

I plan to install 2003 Enterprise Server, NT 3.51 and W98 over the next few days.

I can confirm that the Diamond driver did not help. And that BSOD causes errors on the hard drive. Perhaps CPU-Z and SIV cannot work with the VLSI Topcat chipset? This is/was a fairly uncommon pure 386 chipset. Kixs tested his SXL2-66 with a newer 386/486 hybrid chipset.