Please say again, Which one is least problematic K8T800 chipset? Metal plate (older) or completely plastic?

Cheers,

I've been working on something similar myself. It's based on an oddball socket AM2 (of all things) Gigabyte GA-MF3 mainboard that uses the nForce 3 250 chipset witch has a proper AGP slot. Specs so far:

Athlon 64 LE-1660 2.8GHz single core
Gigabyte GA-MF3
2x256MB DDR2 667MHz stolen from some OEM system
Leadtek A400GT TDH Geforce 6800GT AGP
Creative Sound Blaster Live! CT 4830
200GB Samsung IDE HDD
Chieftec Dragon full tower case
3.5" + 5.2" FDD
Teac IDE CD-DVD RW combo drive
450W modded FSP 80+ PSU (cut off some sata connectors and soldered on more molex since the PSU only came with two)

Still working on getting it to run reliably, but the nForce 3 chipset is not really designed for win98 and drivers are a bit spotty. Initially I wanted to dual-boot XP (might still do it) so the plan was to run 2GB of dual channel DDR2 800, but the PC seems unstable with 800MHz DDR2 and it outright refuses to run 800MHz ram in dual channel regardless of who made the ram. Thinking the board always tries to run the tightest timings, I even tried a set of Corsair XMS2-1066 since the board has no BIOS options for setting ram speed (but oddly it has CPU overclocking options including voltage and even multiplier) and while it does post it hangs soon after. I was also planning to run a Gainward Geforce 7800GS AGP but I couldn't get the unofficial drivers to run correctly. Some games work fine like unreal tournament 99 and even quake, but others like Dungeon Keeper 2 flat-out refuse to start with the 7800. I was also thinking of using an Athlon X2 6400+ witch XP could take advantage of, but while perfectly stable in XP, I've had a myriad of issues with that CPU in 98. SATA is un-usable. Having the on board SATA controller enabled causes win98 to boot-loop so I had to go with IDE. Funny enough it works fine with a PCI sata2 controller, but can't boot off it.

It it's current configuration it seems to be stable enough, but it can't play everything my Thinkcentre can (p4 2666 / 6600GT AGP) so I'm going to keep working on it.

Socket3 wrote on 2023-05-23, 08:44:

Still working on getting it to run reliably, but the nForce 3 chipset is not really designed for win98 and drivers are a bit spotty.

I only have one nForce 3 Ultra motherboard (Gigabyte GA-K8NS Ultra-939), I'm dual booting Win98 / WinXP, and my experience with this build has been great. Haven't encountered any chipset-related issues in any of the two operating systems.

bloodem wrote on 2023-05-23, 09:38:

Socket3 wrote on 2023-05-23, 08:44:

Still working on getting it to run reliably, but the nForce 3 chipset is not really designed for win98 and drivers are a bit spotty.

I only have one nForce 3 Ultra motherboard (Gigabyte GA-K8NS Ultra-939), I'm dual booting Win98 / WinXP, and my experience with this build has been great. Haven't encountered any chipset-related issues in any of the two operating systems.

What driver combination are you using? I've had nothing but issues with both nForce 2 and nForce 3 chipsets under win98se: two Abit AN7 boards are completely unstable under 98 after installing forceware. CPUs tested are Athlon XP 3200+ both the KV4D and KV4C versions (2333mhz / fsb 333 and 2200MHz fsb 400) as well as an athlon 2600+ (2133Mhz fsb 133). 512MB of DDR400. The AN7's seem OK with ATi cards witch I find odd. Even my X850XT worked fine(ish). Disabling SATA in bios helped as well. I also have Gigabyte and an Asrock nforce 3 boards - similar results with both. The asrock board seems the most stable, as it does not lock up or display windows protection error after installing forceware, but some games cause lock-up after 30-60 minutes of gameplay.

Even worse with nforce 3 boards. I have several socket 754 and 939 boards from different manufacturers (DFI, MSI, Asus) and they all act the same - general instability or out-right refusing to boot after installing forceware.

All configurations mentioned above work flawlessly in XP SP3.

I've resigned myself to using VIA chipsets for my overkill '98 / XP builds - for socket A I went with the VIA KT880 (Asus A7V880) + 3200+ AXDA3200DKV4D (2333MHz) + Radeon X800 Pro AGP + 2x512MB DDR400 from OCZ. For 939 I went with an ECS KV2 extreme (VIA K8T800) - Athlon 64 3800+ single core, 2GB ram, Galaxy GF 6800 GS AGP unlocked to 16/6, 250GB Sata hdd - and for 754 I'm using an MSI K8MMV + Athlon 64 3400+ and an MSI FX5950 Ultra. Perfectly stable, all of them - no driver issues apart from the Asus A7V880 witch is only stable with the Via Hyperion PRO 5.23 chipset drivers. Anything else locks up the PC on boot when using an nvidia card and any version of forceware. They all dual boot XP SP3.

Socket3 wrote on 2023-05-23, 12:06:

What driver combination are you using? I've had nothing but issues with both nForce 2 and nForce 3 chipsets under win98se: two Abit AN7 boards are completely unstable under 98 after installing forceware. CPUs tested are Athlon XP 3200+ both the KV4D and KV4C versions (2333mhz / fsb 333 and 2200MHz fsb 400) as well as an athlon 2600+ (2133Mhz fsb 133). 512MB of DDR400. The AN7's seem OK with ATi cards witch I find odd. Even my X850XT worked fine(ish). Disabling SATA in bios helped as well. I also have Gigabyte and an Asrock nforce 3 boards - similar results with both. The asrock board seems the most stable, as it does not lock up or display windows protection error after installing forceware, but some games cause lock-up after 30-60 minutes of gameplay.

I've installed the nForce Win98 chipset driver from 2004, which is available on the Gigabyte website, on my motherboard's support page.
Are you sure you are not encountering memory size related issues? For me, the only patch that absolutely fixes any memory related instability is Rudolph R. Loew's PATCHMEM.
Fun fact: 1 GB of RAM in Win98 on a certain motherboard can behave very differently compared to 1 GB on another motherboard (everything else being equal), and this is where Rudolph Loew's patch will come in very handy. All other Win98 RAM size workarounds that I've tried have been hit & miss.

My full system specs:
CPU: AMD Athlon 64 4000+ (socket 939 – San Diego core)
MB: Gigabyte GA-K8NS Ultra-939
RAM: 1 GB DDR (using Rudolph R. Loew's PATCHMEM)
VGA: Leadtek nVIDIA GeForce 5900XT (driver 45.23 - forced installation)
SOUND: Creative Audigy 2 SB0240
HDD: Western Digital IDE/PATA 320 GB (limited to 120 GB with Seagate Tools)

And my notes from 2021:
"PROS: Awesome system, 100% stable, did not encounter a single issue during a week of testing in both Win98 and WinXP!"
"CONS: Since the MB has an nVIDIA chipset, no chance of getting sound working in DOS"

Socket3 wrote on 2023-05-23, 12:06:

I've resigned myself to using VIA chipsets for my overkill '98 / XP builds - for socket A I went with the VIA KT880 (Asus A7V880) + 3200+ AXDA3200DKV4D (2333MHz) + Radeon X800 Pro AGP + 2x512MB DDR400 from OCZ. For 939 I went with an ECS KV2 extreme (VIA K8T800) - Athlon 64 3800+ single core, 2GB ram, Galaxy GF 6800 GS AGP unlocked to 16/6, 250GB Sata hdd - and for 754 I'm using an MSI K8MMV + Athlon 64 3400+ and an MSI FX5950 Ultra. Perfectly stable, all of them - no driver issues apart from the Asus A7V880 witch is only stable with the Via Hyperion PRO 5.23 chipset drivers. Anything else locks up the PC on boot when using an nvidia card and any version of forceware. They all dual boot XP SP3.

I also have the A7V880 and have found it to be exceptionally good and stable for a DOS / Win98 / WinXP time machine. But, yeah, I've used it with the VIA driver version 5.23 in Win98 since day one (didn't even try other drivers with it, because I've concluded that these are the best drivers for newer VIA chipsets many years ago).

bloodem wrote on 2023-05-23, 13:36:

Socket3 wrote on 2023-05-23, 12:06:

What driver combination are you using? I've had nothing but issues with both nForce 2 and nForce 3 chipsets under win98se: two Abit AN7 boards are completely unstable under 98 after installing forceware. CPUs tested are Athlon XP 3200+ both the KV4D and KV4C versions (2333mhz / fsb 333 and 2200MHz fsb 400) as well as an athlon 2600+ (2133Mhz fsb 133). 512MB of DDR400. The AN7's seem OK with ATi cards witch I find odd. Even my X850XT worked fine(ish). Disabling SATA in bios helped as well. I also have Gigabyte and an Asrock nforce 3 boards - similar results with both. The asrock board seems the most stable, as it does not lock up or display windows protection error after installing forceware, but some games cause lock-up after 30-60 minutes of gameplay.

Are you sure you are not encountering memory size related issues?

100% sure. I kept the RAM down to 128 or 256MB on older systems and 512MB max on newer ones that are intended for win98 only, and on my dual boot systems I use Patchmem. I know at first sight it sounds like a memory issue but it's actually AGP configuration or the SATA controllers that cause my instability - sometimes the USB 2.0 on nforce boards or all of the above - but when it comes to all out refusing to boot I found the problem lies with the AN7's IRQ routing and ACPI table. I discovered that completely disabling ACPI and assigning resources manually - like disabling the com and LPT ports to free up IRQ 3,4 and 7 - then assigning those manually to the USB controller mitigates instability - but as soon as I make any hardware change, that all goes down the toilet - again.

On the nforce 3 it seems that the problem is AGP configuration. Just like on the VIA KT400A. For some dumb reason, ACPI assigns the same IRQ to one of the USB controllers and the video card, while another controller shares IRQ with the sound card 😠 - that can case my system to crash if there's a flash drive in one of the USB ports and I decide to fire up any app that uses 3D acceleration - or if I use an Aureal Vortex sound card. Creative cards up to the SB Live! don't seem to mind sharing an IRQ, and neither do Yamaha DS-XG - unless it's Legacy Audio device is sharing IRQ5 with anything else (that causes sound to work OK in windows games but if I fire up a dos game in win9x dosbox the system freezes). Disabling the USB controllers fixes the problem, same if I disable "assign IRQ for USB" in BIOS (but causes usb drives to stop working).

Unfortunately my nforce 3 boards refuse to use IRQ 3 and 4 despite COM ports being disabled in BIOS - same with the KT400A and some KT600 boards that do not have a "ACPI disable/enable" setting in BIOS (some do). On my DFI I have to turn AGP Fast Writes off and disable both agp master 1 ws read and write, which tanks performance with some video cards. Again, this issue is not present when using ATi cards (except for the X800 series these seem to freeze after a little while in direct 3d if I keep fast writes on)

bloodem wrote on 2023-05-23, 13:36:

I also have the A7V880 and have found it to be exceptionally good and stable for a DOS / Win98 / WinXP time machine.

I have 4 A7V880's... 2 are dead - haven't got around to doing any extensive testing on them - one works perfectly, and the other has 1 dead memory channel (2 slots closest to the CPU don't work - voltage to memory modules drops to 1.5v whenever any stick of ram is installed there). Decent board, good chipset - perhaps the best one VIA ever made - but the A7V880 is not my favorite KT880 board. I had a Gigabyte GA-7VT880 Pro that's been incredibly solid regardless of what I put in it. Unfortunately, I killed it a few years ago when taking it out of a case (one of the screws got stuck in a standoff and I irreparably damaged the board trying to remove it from the case it was in). Still hoping I'll come across another one in my local classifieds, but no luck so far. Back in the day when socket A was relevant I used to have an Abit KW7 - man that was a good board. Kept a barton 2500+ overclocked to 2200MHz for years on one of those babies. The Soyo SY-KT880 is not bad either. Not as fast as the Asus A7V880 or the Abit KW7, but really, really stable. I used one as a socket A test bench for years, then stashed it in 2008 or so and haven't seen it since....

What about forceware drivers? I noticed severe compatibility issues between forceware and early VIA drivers, particularly 4 in 1 series. Hyperion PRO drivers seem to work well enough with drivers 5x.xx and newer, but still have problems with older forceware releases.

bloodem wrote on 2023-05-23, 13:36:

But, yeah, I've used it with the VIA driver version 5.23 in Win98 since day one (didn't even try other drivers with it, because I've concluded that these are the best drivers for newer VIA chipsets many years ago).

Yeah, I guess there's something to that. I recently pulled out an Epox KT400A board - great on paper, it even has multiplier selection in BIOS - awesome for unlocked chips - but I could never get it to run stable under win98se. At one point I had it in my "to be repaired" box, thinking the capacitors failed. Had all of them replaced and the board toughly checked, to no avail - but out of boredom, one night I started messing with the drivers, and found a good combination - Hyperion 5.23 + Forceware 53.04. The configuration is Barton 2800+ / Epox EP-8K9A7I, 512MB of ram (single stick, Kingston CL 2.5 DDR400) and a Quadro FX 3000 (essentially an FX 5900 Ultra). I tried forceware 45.23 + all 4 in 1 drivers, it would freeze at boot. Then I tried 45.23 + hyperion 5.23, it would throw "windows protection error". Going to safe mode and uninstalling the Via 2.0/3.0 CPU to AGP driver would fix it, but the video card ran in PCI mode. FInally I found 53.04 on phill's website and it seems the PC is perfectly stable now.

[EDIT]

Re-reading all of that made me realize that Socket A platform can be a real pain in the @$$. Shame, since I have a real fondness for it, as it was my go-to setup un highschool and 1st year of collage so I'm pretty nostalgic about it - particularly nforce 2 boards. At this point, with limited time and resources, I've taken to building i865 socket 478 or 775 rigs when I need a reliable win98 machine, since they're usually very easy to get going and rather reliable... or 754/939 with a VIA chipset. At this point, the only socket A builds I haven't had issues with are either KT133 (as long as it has a FSB100 CPU in it - and I don't have any working KT133A boards unfortunately) or KT266A / KT333 motherboards. I've had numerous annoying issues with later chipsets, particularly nforce 2 and KT400 - only when it comes to running win9x tough, and KT880 boards are rare, and haven't been treated with "respect" by manufacturers. Asus took a piss when designing the A7V880. Sure it's fast, but they kind of have a way of dying for no particular reason (at least that was my experience). Soyo and Chaintech didn't bother making any optimizations and their boards are pretty slow... Gigabyte and Abit seem to be the only manufacturers who gave the KT880 any love. No Idea about MSI or Asrock, I've never owned a KT880 made by either of them.

VIA 0 bytes AGP memory bug confirmed:
ASRock P4V88 rev. 1.10 (VIA PT880).
BIOS versions: all.

I have created a thread collecting to all the info on the matter.