VOGONS


First post, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

I finally got around to setting up Windows XP on my trusty i5 2500K MSI P67A system as an old PCI-E GPU test bench and the first card I tested on it (Quadro FX 4500) worked flawlessly.

Next, I have three Quadro FX 1300 cards, which are basically underclocked Geforce FX 5950 Ultras (NV38GL) with much much slower memory. I would consider these "first generation" PCI-Express GPUs, so this is about as far back as this machine will need to go. These use a bridge chip for the PCI-E interface... and it seems likely that's the source of the issues.

They all boot up to the point of loading Windows XP, but after that I have been unable to get them to work right. On the first one I was able to get to the point of trying to install drivers (it needed older ones than what I had installed for the FX 4500) and as soon as it started actually installing devices the system dropped to a black screen with a blinking white cursor. After that it would reboot to the same black screen with white cursor, even when choosing "VGA Mode". I even had the system blue screen in safe mode one time with one of these cards installed.

After uninstalling the other drivers completely with a different card installed, as soon as I try to boot with an FX 1300 it will reach the Windows loading screen and after a few seconds it will drop to a black screen with no blinking cursor, but with the monitor still receiving signal. I can drop the FX 4500 back in and it immediately works.

The FX 1300 cards themselves are pristine and based on the way they were packaged I would assume they are either brand new or were used lightly 20 years ago, pulled out of workstations and immediately put into ESD bags for storage. Absolutely zero cosmetic wear or damage, no broken SMD components, etc.

So, what gives? The motherboard is from ~2011, and the cards are from ~2003-2004. So, there are some years between them but this is the first time I've ever heard of anything like this. Also, there are no PCI-E 1.1\2.0 settings in the BIOS. I have also run a thorough memory test and run through 3D benchmarks with the FX 4500 so it isn't a system stability issue.

Next I will try running DDU to remove all traces of GPU drivers and then try installing a much older driver. Seems odd that if it was a driver issue that it'd still cause issues in safemode or with the drivers uninstalled... but hey, it's Windows XP. Stuff happens.

If anyone has any experience with PCI-E cards of this generation (with bridge chips), your input would be appreciated.

Now for some blitting from the back buffer.

Reply 1 of 15, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Whelp, after fighting with stupid DDU for 45 minutes I was finally able to wipe all traces of nvidia drivers from the system, then I downloaded a Quadro-specific 91.85 driver package from nvidia, installed it, aaaannnd.... black screen with blinking white cursor right in the middle of the driver install.

Meh.

Installing Forceware 61.77 dies during the install process with a crazy flickering screen. Forceware 93.81 crashes with a white blinking cursor, and all newer versions I have tried do the same.

This is really annoying, because on paper these cards would be a really awesome addition to a newer PC with PCI-E slots for running really old games. I noticed the memory chips on my cards are rated for 350Mhz too, which is quite a bit higher than the 275Mhz that the card apparently runs them at stock.

Also, a helpful tip for anyone who wants to run Display Driver Uninstaller on Windows XP:
Make sure to get DDU version 18.0.0.2 from here. For whatever reason, 18.0.0.4 (the last version to support Windows XP) would crash when I tried to start the cleaning and I found only one other post of someone experiencing this online, and they were on Windows 10. Thankfully, they mentioned that 18.0.0.2 worked fine. Sure enough, it worked great for me.

Additionally, you'll need .Net Framework 3.0 or later to get the program to even load. I downloaded this (surprisingly huge) offline installer, and that worked for my fresh XP install.

EDIT: Tried the card my other PCI-E x16 slot and it made no difference.

Now for some blitting from the back buffer.

Reply 2 of 15, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

That's HSI bridge issue. I had problems with NV45 chips even on X58 boards, although they worked with abysmal performance. I don't recall any issues on AMD boards though.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 3 of 15, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie

Use latest driver (169.96 for Quadro), and see if that helps.
No point in trying to install older driver, if card is giving you issues.

Reply 4 of 15, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t
The Serpent Rider wrote on 2025-03-02, 20:06:

That's HSI bridge issue. I had problems with NV45 chips even on X58 boards, although they worked with abysmal performance. I don't recall any issues on AMD boards though.

Ah, so this is a known thing then.

I've had trouble tracking down any information about cards that use an AGP to PCI-Express bridge for PCI-E cards. I wasn't aware that some Nvidia 6000 series did this until you mentioned it. I see that NV45 is just NV40 (6800 GT, 6800 Ultra, etc.) with a bridge integrated into the die, so any of those cards on PCI-E actually utilize a bridge, even if there isn't a separate chip.

I'm curious if I'd have similar problems trying to run one of those cards on this system.

It's really lame that they were so picky so soon after release. I can probably run a non-bridged X300 card from around the same time period in my X570 board with no issues, but because of the stupid bridge chips we have cards from 2003-2004 not working in boards from 2010.

Any idea where the cutoff is? I wonder if an Intel P35 or P45 board would still have these issues.

agent_x007 wrote on 2025-03-02, 23:01:

Use latest driver (169.96 for Quadro), and see if that helps.
No point in trying to install older driver, if card is giving you issues.

I originally ran a 175.xx driver that supported the Quadro FX 1300 and it had the same problems. Considering the card was from the era of Forceware 5x.xx, it seemed unlikely that newer drivers would fix bugs with such an old card.

Last edited by Ozzuneoj on 2025-03-03, 03:45. Edited 1 time in total.

Now for some blitting from the back buffer.

Reply 5 of 15, by fosterwj03

User metadata
Rank Oldbie
Rank
Oldbie

Yes, this is a known issue. I got my FX 1300 working on a P45 board, but it's performance left me wanting. I have lots of much better PCIE cards which leaves the FX 1300 an interesting novalty, but nothing more.

Reply 6 of 15, by fosterwj03

User metadata
Rank Oldbie
Rank
Oldbie

Oh, I should have mentioned that I have a GeForce 6800 GS running on my Sandy Bridge system (Q67 board with an i7-2600k). It works just fine with NT 4.0.

On the other hand, I haven't gotten this same card to work correctly on any newer Intel system with NT 4. It will boot into NT 4 on my Ivy Bridge system, but crash after displaying the desktop for a moment. I only use this card with NT 4, so I don't know if newer OS's have the same issue.

Reply 7 of 15, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t
fosterwj03 wrote on 2025-03-03, 03:35:

Yes, this is a known issue. I got my FX 1300 working on a P45 board, but it's performance left me wanting. I have lots of much better PCIE cards which leaves the FX 1300 an interesting novalty, but nothing more.

Good to know this isn't an isolated thing at least. Wish I hadn't spent hours fighting with them, 🤣.

I assume the situation is similar with all of the Geforce PCX 5xxx series cards as well.

It's very odd that that I haven't come across this before. It seems like people would be tossing these cards into overpowered XP systems to maintain good compatibility with games that don't work as well on 6-series cards and up (or Radeons)... but I just haven't seen reports of people trying and failing with this. Even with the PCI-E 6800 GT, Ultra and Quadro FX 4400 cards, I've never heard of these issues before.

People will talk about how Windows 9x doesn't complain about PCI-E cards, but presumably you'd have to be using one of these 6-series or FX series cards that have a PCI-E bridge to use a PCI Nvidia card in Win9x... but maybe there are no issues at all running these cards on older platforms that work in Win9x. I wonder what the cutoff is?

I would like to see the FX 1300 compared to a comparable FX series AGP card on a few different platforms to see where the performance starts to take a dive.

Now for some blitting from the back buffer.

Reply 8 of 15, by Dothan Burger

User metadata
Rank Member
Rank
Member

Yeah I've had the same results with an FX1300 and a PCX 5750, performance just isn't right in 98. I think I'm going to retry playing with those cards.

Reply 9 of 15, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie

Note about performance loss on 512MB+ RAM things : LINK

Reply 10 of 15, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++
Dothan Burger wrote on 2025-03-03, 04:15:

Yeah I've had the same results with an FX1300 and a PCX 5750, performance just isn't right in 98. I think I'm going to retry playing with those cards.

VESA testing shows extremely low RAM throughput on Z390 with a PCX 5900. Have not tested with the 5300 or 5750 yet.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 11 of 15, by Dothan Burger

User metadata
Rank Member
Rank
Member
cyclone3d wrote on 2025-03-04, 21:08:
Dothan Burger wrote on 2025-03-03, 04:15:

Yeah I've had the same results with an FX1300 and a PCX 5750, performance just isn't right in 98. I think I'm going to retry playing with those cards.

VESA testing shows extremely low RAM throughput on Z390 with a PCX 5900. Have not tested with the 5300 or 5750 yet.

How did you test that and what driver are you using? Maybe I can compare with the FX 1300 on an Intel 915 chipset.

Reply 12 of 15, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Get vesatest from this post:
Highest single threaded performance for Freedos baremetal with SVGA programming

Get dosbench suite from Phil's Computer Lab:
https://www.philscomputerlab.com/dos-benchmark-pack.html

Run the dosbench suite and run MTRRLFBE to enable write combining.

Then run vesatest with different resolutions and color depths to get the RAM throughput.

You can run PCPBENCH (included in Phil's pack) with the modes command to see the different VESA modes available.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 13 of 15, by Kruton 9000

User metadata
Rank Newbie
Rank
Newbie

I want to say right away that if someone thinks that rare GeForce PCX 5 series cards or their Quadro analogues are the best of both worlds (good compatibility with old software and new hardware with quite high performance), then this is not!
1. They do not support the most compatible old drivers, as the regular GeForce FX does.
2. AGP system with a Core 2 processor can squeeze everything out of these cards, you should not expect acceleration from faster PCI-E systems.
3. Since these cards still operate on the AGP bus internally, the PCI-E interface itself will not add performance even theoretically.
4. As mentioned above and discussed on this forum even earlier, cards with AGP/PCI-E bridge chip have big performance issues in Windows 98, and perhaps not only in it.
5. You also mentioned problems with newer systems, which I had never even heard of before.

Yes, GeForce PCX 5 still supports palletized textures, which 6 series does not, but without compatibility with old drivers and with performance problems, they are not as good as they could be in theory. GeForce 6 series is at least cheaper and faster. ATI cards don't have such problems at all.

Reply 14 of 15, by Dothan Burger

User metadata
Rank Member
Rank
Member
Kruton 9000 wrote on 2025-03-13, 17:44:

Yes, GeForce PCX 5 still supports palletized textures, which 6 series does not, but without compatibility with old drivers and with performance problems, they are not as good as they could be in theory. GeForce 6 series is at least cheaper and faster. ATI cards don't have such problems at all.

I don't know of any ATi card that's bridged from AGP to PCI-E, are there any?

Had no problem running the FX 1300 / PCX 5750 (besides being slow) on a P55 chipset after the OS install but it would lock up during the installation with the either one, It would get to like 23 minutes left and just stop. I've had luck running these cards in 98 and XP with 915/945/P55 chipsets so far.

Someone earlier mentioned that AMD chipset didn't have any issues with these cards, can anyone confirm that?

Reply 15 of 15, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

ATi never bothered with such solutions, although their Rialto PCIe-to-AGP chip probably works both ways, just never implemented.

I must be some kind of standard: the anonymous gangbanger of the 21st century.