Well. Bad news. Turns out none of my 3dfx cards work. 2 had physical damage. 3rd works for about 3-5 seconds before crapping out. So I just ordered a pair of what I believe to be the best voodoo2's off ebay.com. See graphic. This build is already basically Jurassic Park "Spared no expense" So oh well. Be back in a month when I get them.

110 MHz. interesting. I think mine only read 100 MHz.

I haven't seen these come with the heatsinks like this. Which manufacturer did this?

I have a dead Diamond Voodoo2 which doesn't work at all, and one STB Voodoo2 which works for 20-60 minutes, then stops working. Then next day works again. Probably a bad solder joint somewhere. Seems like there's a lot of dead Voodoo2's. Has anyone determined if there is any commonality to these failures?

Gainward Dragon. 110mhz memory does not guarantee that chips will work on such high frequency, even with sufficient cooling.

And AT Cases are not ventilated that well too...

The 3 cards I'm using I got one to work last night after cleaning and pin bending.
I originally got them from a big box of pci/agp cards that where not in static bags. That had been through two moves. Before I got them. So think worst conditions.
Yeah no reason the ram couldn't be down clocked. guess I'll run some old gpu tests on them when they come in and post up the results. They might turn out to be crap, and I got jipped. I'll wait and see. Hey stop killing my unfounded happiness mixed with buyers remorse dang it. 🤣. If I was working on facts and logic I wouldn't have started down this road. hahaha

Everything will workout in the end.

I would like to add that the 6200 PCI and especially the 6200A (NV44) PCI variants are - even if you get the "max" out of them - heavily limited by the PCI bus and only reach about the same performance as a GF4MX440. So the whole hassle of finding a driver that is not too new but still supports it greatly is not really worth it. As correctly stated before, it would be way easier for this system to tickle the "max" out of a GF4MX440. Even a MX4000 PCI (64bit at least) would be easier to "feed" here. Keep in mind that these late PCI GPUs really benefit from having a complete PCI bus for themselves, not having to share with any other devices, as otherwise the already present drawback from the 32bit 33Mhz PCI bus speed will be crippled even more. Most benchmarks out there for the 6200 PCI and GT 610 / 520 / 420 PCI don't take this into account, and it really makes a lot of difference.

On a related note..: This post was written on a daily driver system sporting a GT 610 PCI. Not "just because", but because it is the only option. Well... one of them.

Well so far I haven't found a 6200 that will work ( I have 3 different models) in this board so a 6200 really isn't on the table. Fx5200 is the newest found so far to function. Also not all games are as CPU heavy as others and can rely on gpu and gpu memory for fps over maxing out the bus. Also mx4000 won't post on the board, or at least the functioning one I have won't. This build is more about CPU and board compatibility than maxing out a bus.
When the build is completely done, I'll build a new post in the builds forum with a full list of parts I've tried and all the ones that worked.

Also I bought a new unopened fx5500 to try, came in last week. No post. So for primary I'm sticking with the FX5200 unless someone out there verifies another make and model faster/better/newer/older on the Q bios with the PCI Nx586 motherboard.

I would like to add that the 6200 PCI and especially the 6200A (NV44) PCI variants are - even if you get the "max" out of them - heavily limited by the PCI bus and only reach about the same performance as a GF4MX440.

That's not true: https://www.youtube.com/watch?v=Al_T6bSoclM

I don't see any comparison in that video to cards outside of the PCI bus. and when looking at the performance compared to a FX5500 (which is very close to a MX440), I see differences ranging from marginal to about 25%, depending on what game. So it's safe to say that although it is faster, depending on what game you play, most of the times no by much.

Your videocard is officially overkill when you have more vram than system ram.

BinaryDemon wrote:

Your videocard is officially overkill when you have more vram than system ram.

Lolz... yeah 196MB system ram maxed. GPU has 256MB. This is kind of a #YOLO moment. Because even if I tried to run a game that could use that 256MB the cpu would choke. hahahaha Don't get me wrong I did have that moment where I was like WTF am I doing. ¯\_(ツ)_/¯ But then I remembered my 1980's PlaySchool mailbox as a kid, if the shape fixs the hole. lolz. But I've had no issues with the PNY FX5200 so far. ***edited*** (my fx5200 is only dual vga)

Hardest part I've had so far is a fully functional mouse. I read elsewhere when you place a newer pci gpu in this board the on board serial stops working. Which is true. So I got a nice dual serial Siig isa card, but no matter the hardware address or the irq I had issues. In the device manager resource list there was always a "!" on the resource and irq. Tried everyone available by jumper then matched it in the device manager. It would detect it the first time, I would reboot, then it wouldn't work. I'd then delete the mouse out of the device manager and reboot. Then it would work until I rebooted again. It's a Serial Microsoft Wheel Mouse I bought off ebay brand new still sealed. So I figured okay try a new card. So I got a startech single serial port isa card. Better luck, kind of. You could reboot finally. But windows would still say no mouse on reboot but when you got in to windows up to the point explorer.exe was loaded it would start working. kind of strange. However I couldn't set the card to hardware dressing of comm 3 or 4 but 6 worked. Even though I could find no hardware using the address spacing. Ended up 3E0 and irq 9. But it still has a "!" not in the device manager list but in the resources lists for the comm port hardware address and irq.

By resource list I don't mean the device properties, I mean clicking properties on "computer" at the top of the device manager list. To see all resources in use.

Also there is no real way to disable the onboard comm ports. Even if you disable in the bios, even if you pull the comm 1/2 irq jumpers on the board they never disable completely and always end up using irq 3/4.

I doubt this will help anyone, but I figured I'd throw it out there. To see if anyone else had seen this. Spent a lot of time fighting this ghost with endless trial and error and maybe 50 reboots. Still not completely fixed and working but usable.

Well it's finally done. Though not completely happy with the NxBus through put at the moment, because of the inability to run bus sync mode to the cpu frequency. And the known mouse issue. But over all the system is functional, games work, SLi works, D3D works, Network works, Compact Flash is transferring at max speed of the onboard IDE, Sound from the AWE64 Gold works. No lockups or malfunctions in relations to Windows 98 itself or the motherboard.

I'm still working on over all CPU/Bus speed by replacing the onboard L2 Cache of the motherboard itself. Which has been done before, but I haven't see it done to a NexGen PCI motherboard as of yet, only the NexGen VLB version. If it has I haven't see photos of a cache modded NexGen PCI version.
You can follow the progress here and find info about this process and tips here;
http://www.cpu-world.com/forum/viewtopic.php?p=259471
This thread should have all related information, and links to previous works of hard working users that this process originaly came from.

I plan on making a new thread later showing all drivers, softwares, and hardware used. In hopes future people working on this can skip some of the hard work and move right to playing games and testing on future builds. I'll post on this thread when I do.

Just a small update;
Only have tested 2 graphics cards that are PCI that work, in Windows 95/98
Creative Labs CT6700 Riva TNT (async bus mode)
PNY Geforce FX5200 (sync or async bus mode)

The rest I've tested ether,
Don't POST
Beep on power on
Hard lock in Windows

(In Windows 98) Even with the two that I got to work, they still cause resource conflicts in Windows, making onboard serial ports not useable. No bios, onboard jumpers, or Windows settings would fix the issue. Serial mouse would never detect.
I had to add a add-in card for a serial comm port in a ISA slot. I used a Microsoft X03-69944 5volt serial mouse.
(In Windows 95) Manually adding Microsoft Serial Mouse works.

I pulled the voodoo 2's out and put just one in. So I could run a Ultra100 pci ide controller.

thehinac: I didn't run into such problems on my board. Try a Matrox G200 + Voodoo2 + Promise Ultra100.

Did you take a photo of your board with those SOJ SRAM sockets and 128kx8 chips installed?

feipoa wrote on 2021-07-24, 07:06:

thehinac: I didn't run into such problems on my board. Try a Matrox G200 + Voodoo2 + Promise Ultra100.

Did you take a photo of your board with those SOJ SRAM sockets and 128kx8 chips installed?

Thanks! I just got on ebay and bought the g200 I have 3 other Matrox cards but they didn't work. All different models.

Nah, never got around to finishing that on my PCI spare board. Still want too. It's been sitting on my desk at work for like a year. I'll get around to it eventually. I'm just lazy.

I recently purchased 10 SOJ-32 sockets with a plan of upgrading my only NexGen board to 1024K via sockets. I'm not sure when I'll get to it though. Clicking the "buy" button is a lot easier and faster than the actual work.

I plan to do such a stunt too, though without the sockets 🤣