VOGONS


First post, by k24a1

User metadata
Rank Newbie
Rank
Newbie

Asking mainly out of curiosity. If I'm not mistaken the architecture of the R400 and R500 cards are somewhat similar... and the 6.2 drivers came out around 2006. Has anyone managed to get these running successfully on 9x? What I mean by R500 is cards like the X1000 series (e.g. X1950 XT).

Reply 1 of 25, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

I think they are only similar at the bottom end, as the top end uses 4x the pipes and shaders etc. So the X1300s are said to be able to be got working whereas the high end cards can't.... or maybe you can dumb them down to X1300 config.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 2 of 25, by marxveix

User metadata
Rank Oldbie
Rank
Oldbie

Only x1050 maybe, but its because its R4xx line of card, not R5xx, to be correct its RV370 that manufactured at 110nm. 😀

https://www.techpowerup.com/gpu-specs/radeon-x1050.c1812
https://www.trustedreviews.com/reviews/sapphire-x1050

Best ATi Rage3 drivers for 3DCIF / Direct3D / OpenGL / GLUT / DVD : ATi RagePro drivers and software
33+MiniGL / OpenGL Win 9x dll files for all ATi-Rage3 based cards : ATi RagePro OpenGL files

Reply 3 of 25, by k24a1

User metadata
Rank Newbie
Rank
Newbie

It is interesting how ATI had their 2000/XP 6.2 drivers work with the X13/6/8/900 series cards, but for some reason they decided not to implement it in the 6.2 9x version. Perhaps due to limitations with the software? I wonder if it could be possible to backport support for these cards, but seeing how the driver architecture is very different, I honestly don't think it will work.

Reply 4 of 25, by Dothan Burger

User metadata
Rank Member
Rank
Member

I always thought that ATi screwed up with R500. It seems like they should have just rejiggered the R420 for more pipes. Then they could have focused on their unified shader architecture. At the time it looked like ATi was going to dominate the next gen then G80 dropped.

Maybe then they would have been able to release 98 drivers.

Reply 5 of 25, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

By the time R500 dropped (October 2005), there was no incentive to write a new code for an OS architecture that would be completely dead in less than a year (July 2006). Here's your answer.

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

Reply 6 of 25, by NeoG_

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2026-07-16, 10:06:

By the time R500 dropped (October 2005), there was no incentive to write a new code for an OS architecture that would be completely dead in less than a year (July 2006). Here's your answer.

The spirit of the question is more like, has anyone found a way (through driver modifications etc), rather than if there were official drivers

98/DOS Rig: BabyAT AladdinV, K6-2+/550, V3 2000, 128MB PC100, 20GB HDD, 128GB SD2IDE, SB Live!, SB16-SCSI, PicoGUS, WP32 McCake, iNFRA CD, ZIP100
XP Rig: Lian Li PC-10 ATX, Gigabyte X38-DQ6, Core2Duo E6850, ATi HD5870, 2GB DDR2, 2TB HDD, X-Fi XtremeGamer

Reply 7 of 25, by Trashbytes

User metadata
Rank Oldbie
Rank
Oldbie

No unified shader cards work under 98/ME, its not a driver issue but a hardware one, namely the GPU hardware changed in a way that 98/ME could not support without major OS level rewrites by Microsoft, this was even more apparent with Xfire/SLI capable PCIe cards. (AGP never had the capability for multiple GPUs or even multiple AGP slots, there was some rumors AGP 4 would have supported SLI/Xfire but PCIe was adopted quickly by the industry and AGP got dropped like a hot rock)

The ATI cards from the X1000 series by rights should work since they are using normal shaders so my only guess here is ATI not bothering to backport a driver for 98/ME which I understand ..98/ME were both pretty much dead by 2005/6 and ATI was in a lot of financial trouble at that point so likely didn't have enough staff on hand to support more than XP. (ATI was bought out by AMD in 2006)

So while there are a few "modded" drivers that claim to work none of them I have tried actually do and many will utterly bork a 98 setup if installed.

The best you could get would be for a generic display adapter driver to work which isn't exactly useful.

Reply 8 of 25, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Someone probably could port open source Linux drivers, but that would be only for OpenGL.

Trashbytes wrote on 2026-07-16, 10:28:

AGP never had the capability for multiple GPUs or even multiple AGP slots

In fact, it had. That's why Rage MAXX exists.

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

Reply 9 of 25, by swaaye

User metadata
Rank l33t++
Rank
l33t++
Dothan Burger wrote on 2026-07-07, 21:20:

I always thought that ATi screwed up with R500. It seems like they should have just rejiggered the R420 for more pipes. Then they could have focused on their unified shader architecture. At the time it looked like ATi was going to dominate the next gen then G80 stopped.

Maybe then they would have been able to release 98 drivers.

R420 is just another form of R300. It was too limited to have around for a 2006 chip when games began to require Shader Model 3. Nvidia had already moved to Shader Model 3 in Geforce 6, albeit a slow implementation of the more advanced features. R520 is a SM3 chip with high performance SM3.

Originally R400 was possibly going to be a form of unified architecture. Xbox 360 has this original unified architecture highly tuned for 720p. From what I understand a high performance, more flexible PC version would have been too big on 130nm. This is why unified shader chips don't appear on PC until 80/90nm.

Reply 10 of 25, by onethirdxcubed

User metadata
Rank Member
Rank
Member

If it's possible to get OpenGL working from the Linux drivers then it's also possible to translate all Direct3D draw calls to OpenGL (albeit with high overhead).

SoftGPU uses Mesa and Wine code to translate Direct3D and GLIDE to OpenGL. Then the GL draw calls are passed to the host over a socket on Virtualbox/VMWare, or on hardware are rendered on the CPU with LLVMPipe (single threaded because Win98 cannot support multiple cores at all) and then drawn in a 2D VESA mode.

The overhead is so extreme on real hardware that SoftGPU currently can't even beat the VIA UniChrome. So creating another backend for SoftGPU that uses Linux's hardware GL implementation would help a lot. It might be possible to run the Linux display driver in a hypervisor like CSMWrap?

All the display mode setting on Win9x has to be done from 16-bit code but SoftGPU has already done that part.

Reply 11 of 25, by Dothan Burger

User metadata
Rank Member
Rank
Member
swaaye wrote on 2026-07-18, 15:36:

Originally R400 was possibly going to be a form of unified architecture. Xbox 360 has this original unified architecture highly tuned for 720p. From what I understand a high performance, more flexible PC version would have been too big on 130nm. This is why unified shader chips don't appear on PC until 80/90nm.

Forgot about the original R400, wish there was more info on that design. I'm looking at it more through purple colored vogons lenses, as the ultimate overkill 98 card.

Sad that we don't get much insider info from the Dave Orton era of ATi. A YouTube channel mentioned that Dave did not want to go on record, I found that interesting.

X1800 introduced the ring-bus that for reasons I can't explain, I never liked. And it wasn't replaced until the HD48X0 with a crossbar memory controller.

Reply 12 of 25, by rmay635703

User metadata
Rank l33t
Rank
l33t
The Serpent Rider wrote on 2026-07-16, 10:06:

By the time R500 dropped (October 2005), there was no incentive to write a new code for an OS architecture that would be completely dead in less than a year (July 2006). Here's your answer.

Yet Windows 9x active pc base was still 120 million machines by the end of 2006.

Dead zombie army still walking

Reply 13 of 25, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

For Windows 98, Radeon X1000 series doesn't provide a significant boost over X800 anyway. High end cards are still only 16 ROPs, 16 pixel pipes and 8 vertex. Pixel shader performance is irrelevant.

rmay635703 wrote on 2026-07-18, 21:55:

Yet Windows 9x active pc base was still 120 million machines by the end of 2006.

Dead zombie army still walking

In the context of drivers, that's also completely irrelevant, because it's a large userbase with mostly obsolete hardware.

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

Reply 14 of 25, by Dothan Burger

User metadata
Rank Member
Rank
Member
The Serpent Rider wrote on 2026-07-18, 22:05:

For Windows 98, Radeon X1000 series doesn't provide a significant boost over X800 anyway. High end cards are still only 16 ROPs, 16 pixel pipes and 8 vertex. Pixel shader performance is irrelevant.

The R580 is 48 pipes, so it's even more lopsided. but I think the bigger issues is it's still only 16 TMU's which for 98 seems a lot more important.

Reply 15 of 25, by swaaye

User metadata
Rank l33t++
Rank
l33t++
Dothan Burger wrote on 2026-07-18, 17:34:

Forgot about the original R400, wish there was more info on that design. I'm looking at it more through purple colored vogons lenses, as the ultimate overkill 98 card.

Sad that we don't get much insider info from the Dave Orton era of ATi. A YouTube channel mentioned that Dave did not want to go on record, I found that interesting.

X1800 introduced the ring-bus that for reasons I can't explain, I never liked. And it wasn't replaced until the HD48X0 with a crossbar memory controller.

I like the R500 cards for old games. I find their 16-bit color output fascinating. The dither is much different than previous Radeons and they are the last with 16bit dithering. It's cool to combine with DGVoodoo1 which allows 16 bit output. R500 also has top quality AA and AF among pre D3D10 cards and actually seemed to behave better than RV770 in my experiments once. Less mip map shimmer. The cards also support dual link DVI which was useful when I had a 2560x1440 monitor.

It would be great to hear insider tales but I can understand why nobody wants to go there. The ring bus was a curiosity for sure and also things like 16px / tx per clock for 3(4?) generations. Going dual chip instead of big die when AFR was becoming untenable. Etc.

Reply 16 of 25, by Dothan Burger

User metadata
Rank Member
Rank
Member
swaaye wrote on 2026-07-19, 15:20:

I like the R500 cards for old games. I find their 16-bit color output fascinating. The dither is much different than previous Radeons and they are the last with 16bit dithering. It's cool to combine with DGVoodoo1 which allows 16 bit output. R500 also has top quality AA and AF among pre D3D10 cards and actually seemed to behave better than RV770 in my experiments once. Less mip map shimmer. The cards also support dual link DVI which was useful when I had a 2560x1440 monitor.

They do intrigue me, and I have no the doubt the RV770 didn't look as good. My tests with an hd6670 vs 7950gt in HL2 with all AA and AF at max everything had the 7950 stomping it. Much clearer fence transparency AA and much better representation overall. The X1900 by all accounts has even better image quality.

swaaye wrote on 2026-07-19, 15:20:

The ring bus, refusal to build big halo chips as Nvidia successfully uses for mindshare, staying at 16px per clock for like 3 generations, making R600 a FP blending focused chip for 2007...

Agreed... though wouldn't it really be 4 generations at 16 ROPS? I know they updated RV770's but it's still 16.

Reply 17 of 25, by swaaye

User metadata
Rank l33t++
Rank
l33t++
Dothan Burger wrote on 2026-07-19, 17:03:

Agreed... though wouldn't it really be 4 generations at 16 ROPS? I know they updated RV770's but it's still 16.

RV770 is still 16 ROPs but they reworked things so they could finally dramatically step up texturing throughput. R420 through RV670 had the high end all 16/16 AFAIK.

Reply 18 of 25, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++
Dothan Burger wrote on 2026-07-19, 00:04:

The R580 is 48 pipes, so it's even more lopsided. but I think the bigger issues is it's still only 16 TMU's which for 98 seems a lot more important.

AFAIK R580 provides zero improvement over R520 in pre-shader era games, outside of marginally raised core/memory clock.

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

Reply 19 of 25, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie

X1000-series doesn't have a Win98 driver, and porting WinXP driver to work in Win98 is way more tricky than it is worth.
Also, there aren't that many SM3 games under Win98, so any tweaks done to R5xx to make it faster than R4xx are probably going to waste (except efficiency).

Note : If you are dead set on having X1000 series screenshot on Win98, you can just use AGP version of X1050 😉 (which is rebadged Radeon 9600)

The Serpent Rider wrote on 2026-07-19, 20:54:

AFAIK R580 provides zero improvement over R520 in pre-shader era games, outside of marginally raised core/memory clock.

Can confirm :
file.php?id=246027&mode=view
file.php?id=246028&mode=view
Unless you plan on using something way faster than Core 2 class CPU.