I've completed a test fit and cable management of everything. I tried to keep things as unobstructed as possible to allow maximum airflow across the passive GPU heatsink.

For CPU cooling I replaced the 70mm fan that came with the heatsink with an 80mm Zalman fan. It's currently held on solely by friction with the existing screws. It takes some effort to try to remove it, so I'm hoping this will suffice as a mounting solution.

Shponglefan wrote on 2024-03-16, 15:12:

The 4200 Ti is highly compatible with DOS titles in my testing.

I have some vague recollections of GeForce 4 Ti cards causing minor issues with Build engine games like Duke3D. Something about V-Sync not working quite right which results in tearing when FASTVID is used? Been a while since I tested that, but I found this post which also references a similar issue. It may be platform specific though, not sure.

The Audigy 2 ZS will be used for Windows XP, 2000, and 98 SE. It's main purpose is EAX compatibility with support for up to EAX 4.0.

Under Win9x, when using VxD drivers, Audigy cards top out at EAX 3.0. From my testing, the only way to get EAX 4.0 compatibility is to use WDM drivers, which are of course suboptimal for Win9x.

Joseph_Joestar wrote on 2024-03-16, 15:37:

I have some vague recollections of GeForce 4 Ti cards causing minor issues with Build engine games like Duke3D. Something about V-Sync not working quite right which results in tearing when FASTVID is used? Been a while since I tested that, but I found this post which also references a similar issue. It may be platform specific though, not sure.

I've done a fair bit of testing with both Blood and Duke 3D. The only issues I've run into are flickering in resolutions at 1280x1024 or higher, and horizontal line artifacts when running under Windows 98.

In native DOS at 1024x768, I haven't run into any issues. Granted I am testing on an LCD monitor, I suppose issues could arise.

Under Win9x, when using VxD drivers, Audigy cards top out at EAX 3.0. From my testing, the only way to get EAX 4.0 compatibility is to use WDM drivers, which are of course suboptimal for Win9x.

I intend to use Windows XP for EAX 4.0 support.

Partition Planning

For handling the various operating systems, I've come up with the following partition scheme:

250 GB SSD (HD0)

• 5 MB - BootIt Bootloader
• 4 GB (NTFS) - Windows NT 4.0
• 40 GB (NTFS) - Windows 2000
• 120 GB (NTFS) - Windows XP

128 GB SSD (HD1)

• 1 GB (FAT 16) - DOS 6.22
• 1 GB (FAT 16) - Windows 3.11
• 2 GB (FAT 16) - DOS Games
• 2 GB (FAT 16) - DOS Games
• 10 GB (FAT 32) - Windows 95
• 36 GB (FAT 32) - Windows 98 SE
• 36 GB (FAT 32) - Windows 98 SE
• 8 GB (FAT 32) - Windows Files

This scheme is designed so that each OS will have its own partition to itself. OS partitions will be hidden from one another. BootIt does allow for more than four primary partitions, hence 8 partitions on the 128 GB drive.

Since Windows 2000 and XP setups only install to HD0, having the 250 GB drive as the primary drive for all NTFS installs seems the most straightforward.

The 128GB drive will be used for all DOS and Win9X installs. This will comply with Win9X limitations of a maximum drive size of 128GB without needing additional patches.

I realize it's probably redundant to have a separate partition and OS install for Windows 3.11, since it's technically just another DOS partition. But I figure this way I can keep each partition 'clean' when it comes to installing software or drivers for DOS versus Windows 3.11. Plus I can have BootIt boot directly into Windows 3.11 without needing an additional DOS menu.

The two Windows 98 SE partitions will be for A3D and EAX setups respectively.

The DOS Games partitions will be exposed to all operating systems to allow for ease of file transfers under various Windows installs (i.e. copying from USB sticks). The Windows Files partition will be exposed to all Windows OS to store common files (e.g. drivers) and facilitate file sharing between Windows installs.

Installed BootIt Bare Metal and set up all the partitions.

Got DOS installed and a bunch of games 'n things copied over.

In testing, I ran into an issue using CPUSPD:

Attempting to use the throttle capabilities of the program (including just displaying current throttle settings) resulted in the program crashing.

I had previously tested and used CPUSPD extensively with this motherboard + processor, so I know it was capable of working. But evidentially something had changed triggering this issue.

After a couple hours of troubleshooting I traced the issue to the RAM. Previously I was using 2 GB of RAM (2 x 1GB). However I had downgraded to only 512MB for the purpose of setting up the Win 9X installs due to limitations with those operating systems.

For whatever reason this also caused this problem with CPUSPD. To confirm this I replaced the RAM with the previous 2GB and it worked fine after that.

I even tested throttling some games like Dynablaster and confirmed everything was working.

Shponglefan wrote on 2024-03-17, 14:48:

Previously I was using 2 GB of RAM (2 x 1GB). However I had downgraded to only 512MB for the purpose of setting up the Win 9X installs due to limitations with those operating systems.

There's no need to physically remove the RAM for that. During the Win98 installation process, you can install R. Loew's patch after the first restart.

I made a mini guide which explains the process here. Steps 4, 5, 6 and 8 are applicable to your use case.

Joseph_Joestar wrote on 2024-03-17, 14:56:

There's no need to physically remove the RAM for that. During the Win98 installation process, you can install R. Loew's patch after the first restart.

I made a mini guide which explains the process here. Steps 4, 5, 6 and 8 are applicable to your use case.

True, this is certainly an option. Though I prefer to use 512MB to avoid having to babysit the installation process.

Attempted to install Windows NT 4.0. Unfortunately it spontaneously rebooted just after loading CD-ROM drivers. The final image before rebooting was this lovely series of stripes.

I recorded a video of what happened available here: Windows NT 4.0 install fail

I tried disabling hyperthreading and L1/L2 cache, but that didn't change anything. From what I've researched, it seems NT 4.0 has difficulties installing on later Pentium 4 systems.

I may have to try installing on a different platform and then migrating to this system.

Installed Windows 3.11 along with the Orpheus Win 3.1x drivers, CD Audio and the modified SVGA driver (to enable 256 colors).

I also installed and ran Civilization II which did work, at least initially.

Unfortunately I ran into an issue where I lost digital audio playback a couple of times. One of these instances happened after initially installing the CD audio driver.

This had a side effect of seemingly breaking the Orpheus II's Sound Blaster compatibility. Even booting into DOS and trying digital audio would now fail to play back correctly.

I could get everything working again by physically removing the card and re-installing it. But I'm not entirely sure what is going on and why it stopped working.

I've added some CPU throttling by setting a lower multiplier (x14 instead of x17), and disabling L1 cache before loading Win 3.11. This way I can at least see if it's a speed dependency issue.

On a potentially related note, I also lost audio playback with the Audigy 2 ZS under Windows XP.

This was particularly weird since it didn't report any issues with the hardware. Playback device and mixer volumes seemed fine. I even ran the Creative Labs diagnostics and it reported everything was fine. I could even try playing audio and it would act like audio was playing, just produce no sound.

Thinking I should re-install the drivers, I inserted the Creative Labs Audigy 2 ZS setup CD. Upon loading the splash screen, it suddenly started playing audio again. Quitting the setup program and audio was once again working in Windows XP.

I have no idea what this issue was.

This sound issue with the Orpheus II was bothering me, so I continued to try to diagnose it.

In turns out it wasn't Windows 3.11, but appears to a general DOS issue. I started testing it in pure DOS and deduced the following cycle:

1. Power on system
2. Boot the system to DOS and Sound Blaster digital audio wouldn't work (as per above screenshot of the Duke 3D setup).
3. Power off system
4. Physically remove Orpheus II
5. Power on system to BootIt menu
6. Power off system
7. Physically re-install Orpheus II
8. Boot the system into DOS and Sound Blaster digital audio would work properly
9. Power off system
10. Go back to step 1

In testing, everything was fine with soft and hard reboots. Sound Blaster audio would still work. But a complete power off would trigger the issue. And just removing and reinserting the Orpheus II wouldn't fix things. I would need to boot the system first without the Orpheus II, then power off, then re-install.

I had previously done extensive testing under DOS with the Orpheus II and had not run into this issue. But some things had changed since then.

To further try to diagnose the issue, I started reverting to a setup similar to my previous DOS testing. This included:

• Removing all PCI cards
• Reverting DOS driver from 0.58b to 0.57b
• Reverting GUS IRQ to 7 (and reserving in BIOS) instead of IRQ 3
• Modifying EMM386.EXE setting to remove the parameters X=A000-BFFF and I=E000-EFFF

What is really strange is changing those EMM386 parameters seems to be what has fixed the issue. I was wondering if maybe something was using that block of memory, but according to Microsoft Diagnostics, that range is listed as available:

While I tested each change one at a time, maybe it's just a coincidence? Or maybe it was a combination of things?

I've since done a bunch of testing in both DOS and Windows 3.11, and digital audio appears to be fine. Re-installed the Audigy 2 ZS and things are still working. Haven't re-installed the Voodoo2 or MX300 yet, though.

I also installed and tested SimCity 2000 under Windows 3.11. It also worked, although the MIDI music pitch seemed a bit off. It's also surprisingly sluggish despite running on a Pentium 4 @ 3.4 GHz. I'll have to test the Win95 version at some point and see how it compares.

Shponglefan wrote on 2024-03-14, 21:28:

The main operating systems for practical use will be DOS 6.22, Windows 98, and Windows XP.

If you plan on using your ProArt monitor with this rig, I made a mini guide which allows for its drivers to be properly installed under WinXP and Win9x, thanks to a small INF file modification.

Joseph_Joestar wrote on 2024-03-20, 08:31:

Shponglefan wrote on 2024-03-14, 21:28:

The main operating systems for practical use will be DOS 6.22, Windows 98, and Windows XP.

If you plan on using your ProArt monitor with this rig, I made a mini guide which allows for its drivers to be properly installed under WinXP and Win9x, thanks to a small INF file modification.

This rig I plan to use with a CRT given it's partially intended for DOS gaming.

That said, this will be useful to try with my main XP rig with which I do use a ProArt monitor. The nVidia drivers I have installed seem to think I'm using a TV set. 😅

Partition Planning - Part II

I've been rethinking my partition scheme after watching the LGR video on Windows Me. I've never used Windows Me on one of my systems, so this might be a good opportunity to try it out. I have no idea how practical it will end up being, especially in light of doing dual Windows 98 SE installs. But it could be fun to mess around with.

I haven't installed Windows 98 SE yet, so revising partitions will be trivial.

My revised scheme is as follows:

250 GB SSD (HD0)

• 5 MB - BootIt Bootloader
• 4 GB (NTFS) - Windows NT 4.0
• 40 GB (NTFS) - Windows 2000
• 120 GB (NTFS) - Windows XP

128 GB SSD (HD1)

• 1 GB (FAT 16) - DOS 6.22
• 1 GB (FAT 16) - Windows 3.11
• 2 GB (FAT 16) - DOS Games
• 2 GB (FAT 16) - DOS Games
• 10 GB (FAT 32) - Windows 95
• 20 GB (FAT 32) - Windows 98 SE
• 20 GB (FAT 32) - Windows 98 SE
• 20 GB (FAT 32) - Windows Me
• 8 GB (FAT 32) - Windows Files

20 GB per Windows 98 install should still be plenty for my intended use. I have no idea how much I'll end up using Windows Me, but if it proves stable and useful, I want to give myself enough space to take advantage of it as well.

I've also toyed with the idea of setting up a partition for Windows Vista, another OS I've never used. But after reading this thread about Vista performance on a Pentium 4, I might save that OS for a beefier build.

As I continue to test DOS sound card settings, I've re-installed all PCI sound cards and set the Orpheus II back to the settings I was using prior to the audio issues.

This is as follows:

SBPro: A220 I5 D1
GUS: A240 I3 D7

This enabled me to once again set IRQ 7 in the BIOS to PnP instead of reserving for legacy ISA. Currently I'm only reserving IRQ 5 and DMA 1 and 7 for the Orpheus II.

In further testing, DOS and Windows 3.11 audio seems to be working fine. This further confirms my suspicion that it was those EMM386.EXE parameters that somehow was triggering the Orpheus II Sound Blaster Pro digital audio to fail.

I'm still using the 0.57b drivers and will probably continue to do so.

I've also acquired a couple Pentium 4 651 (Cedar Mill) processors. The motherboard included a Pentium 4 650 (Prescott 2M) which I haven't bother to change.

I'll probably going to swap it with one these Cedar Mill processors since they should run cooler and consume less power.

I've also ordered some 631 (3.0 GHz) and 641 (3.2 GHz) processors. I'm not sure which speed I'll ultimately go with. Since this system is mostly geared towards DOS and Windows 9X gaming, with some lower-end XP usage, top performance isn't a huge concern. I wonder if a slightly slower speed might be better from a throttling perspective. Although at ~3 GHz, I'm not sure if a couple hundred MHz difference would end up mattering.

Shponglefan wrote on 2024-03-21, 16:57:

I've also acquired a couple Pentium 4 651 (Cedar Mill) processors. The motherboard included a Pentium 4 650 (Prescott 2M) which I haven't bother to change.

I'll probably going to swap it with one these Cedar Mill processors since they should run cooler and consume less power.

Yeah those are the 65W D0-stepping models. You can also undervolt them to make them even cooler, at least my 651 undervolts like a champ.

Sombrero wrote on 2024-03-21, 19:13:

Yeah those are the 65W D0-stepping models. You can also undervolt them to make them even cooler, at least my 651 undervolts like a champ.

That's good to know, although this particular motherboard / BIOS doesn't allow for changing CPU voltage. It's actually quite limited in terms of CPU options.

But it does have ISA slots. 😁

Swapped out the prior Pentium 4 650 for the Pentium 4 651.

In removing the heatsink I discovered a mess of thermal paste underneath. There was even little bits in the socket itself. I didn't attempt to clean the socket itself as it didn't seem to be causing issues. And removing the entire motherboard wasn't something I was planning to do.

The heatsink for this motherboard is surprisingly heavy (weight is 679g). It takes up the same physical space as a stock Intel heatsink, but comes in at over 3 times the weight.

It appears to be solid copper and I hope all that copper should suffice in keeping the CPU temps manageable.

After swapping in the 651 was greeted with a successful boot.

Just saw that Karl's Retro Channel (KITR) posted a multi-OS build using the same DFI G7S620-N motherboard: The ULTIMATE retro build for DOS / Windows 3 / Windows 98 / Windows XP

He went for a different component set up with dual ISA sound cards, Voodoo2 SLI and an ATI X850 XT Platinum. No PCI sound card though.

It's interesting to see the different options a motherboard like this can provide.

That's a sexy cooler with all that copper!