Nice find! Looks like a Rage XL chip?

This page lists DOS compatibility on an AGP equivalent for reference:

https://gona.mactar.hu/DOS_TESTS/

Have a read first:

Did anyone try these modern ATi Rage XL PCI?

ATI Rage XL 8MB on 486 PCI

Pierre32 wrote on 2021-09-07, 06:02:

Have a read first:

Did anyone try these modern ATi Rage XL PCI?

ATI Rage XL 8MB on 486 PCI

That's interesting. Won't even post on socket 7 systems.

The ones I ordered are still in the mail but I'll check on my Socket 7 when it arrives. I have a range of cards with notes on them about compatibility that makes little sense. Sometimes cards which do not work in my Slot 1 system work fine in a Socket 7 and so on.

I would caution against making any sweeping claims against all motherboards because of this. I continue to have will-not-post issues with even more modern cards like the GeForce 8800 GT. I even had an issue with two seemingly identical GeForce FX 5500s would post different results. These issues occur across PCI, AGP, and PCI-e.

Yeah it's not all socket 7 systems. It wouldn't post in one of mine, but did work in a SS7 system. Possibly a PCI voltage thing, a topic discussed in the 486 thread by people who know more than me.

I'll be interested to hear your results. More data is good!

I received three of these today and they are very strange. They have a QC sticker for 2020 and appear to be recently made but the GPU die is visibly beat up on all of them. I wonder what kind of a lot these are from that the GPU package looks like this. It doesn't make sense that they would salvage GPU dies from old cards just to manufacture the same card again new.

I have been having intermittent compatibility issues with PCI cards and this one appears to fall into that category. These cards do NOT work in my ASUS P2B Slot 1 with Intel 440BX chipset. Age isn't a factor as any card that fails to work in this way consistently works in my much older Intel AN30TX Socket 7 with Intel 430TX. Newer boards generally have better compatibility but sometimes older ones do too.

I don't know what the issue is that causes this but these are still useful cards for retro computers with only PCI slots for video with the caveat that they simply will not work in some systems.

These "new" cards might be using crappy DRAM if this is any indication AliExpress AGP cards

If I had to make a guess, it would be that the chips are coming off noughties P4 class Xeon server boards. Therefore I would be trying drivers intended for those if no others were working.

Einherje wrote on 2021-09-07, 06:15:

Pierre32 wrote on 2021-09-07, 06:02:

Have a read first:

Did anyone try these modern ATi Rage XL PCI?

ATI Rage XL 8MB on 486 PCI

That's interesting. Won't even post on socket 7 systems.

Confirming that this card works in my Socket 7 system but there is some kind of compatibility problem with some boards.

darry wrote on 2021-09-22, 03:51:

These "new" cards might be using crappy DRAM if this is any indication AliExpress AGP cards

I think the idea of a cheap PCI card with basic OpenGL support and 8MB of ram, slow or not, is very attractive. I think the same argument has been made about all of the GeForce FX 5500s that are being manufactured in China as well; Phil made a good video about this. Depending on what you choose to pair it with and the use case of the system, these cards will be more than sufficient and are fit for purpose.

Do you think PCI voltage might be the difference between your motherboards?

https://www.ni.com/en-au/support/documentatio … patibility.html

Interesting. I just got an SIS 6236 16MB AGP which is based on the S3 Virge. I was intrigued enough that I just ordered a Rage Pro 128 16MB AGP upon reading this thread. I love benchmarking and comparing video cards and I just took a deep dive back into vintage hardware. With modern GPUs, it is really easy to look up and compare benchmark results. But it can be much trickier with vintage hardware. I would like to get a database going, either here on Vogons or elsewhere. a Passmark for potatoes. I would probably give something like an HD3850 a score of 20 000 because it represented the effective end of this era we are discussing, just as the RTX3000 series probably will be seen in the future as representing the end of the era we are currently in. When I benchmarked an RTX 3080 for the first time, I ROTFLed. It just blows everything else out of the water.
There are some cards from the early '00s listed on Passmark. But with scores of like 2 and 3 and 5, it really doesn't help to compare vintage hardware. I nearly crapped myself when I looked up the HD3850 AGP I bought today and saw it actually got a "score" of over 400! About 1/20th of a 1650Ti but still. Impressive for 2007.

Pierre32 wrote on 2021-09-22, 04:39:

Do you think PCI voltage might be the difference between your motherboards?

The cards are all keyed for both 3.3v and 5v. I don't know of a way to test whether the motherboard is failing to deliver the requested voltage.

HangarAte2nds! wrote on 2021-09-22, 04:51:

I just got an SIS 6236 16MB AGP which is based on the S3 Virge.

It is not and neither of those is known to able to control that much memory.

I tested this today and it feels faster than a PCI TNT2 M64 in OpenGL but slower in DirectX. There were also some graphical glitches in Unreal but Unreal Tournament ran fine, albeit very slowly.

3DMark99 ran without issue but was also slower than a TNT2 M64.

Overall I'm very happy with it. It's about as fast as I was expecting it to be. Buy this for the basic OpenGL features. DirectX is passable and is probably fine for older titles.

I used driver version 4.13.1.2655 which is dated April 2002. I believe this is the latest driver available.

These are, to the best of my knowledge, 32-bit memory bus cards that will basically suck ass at everything except basic Dos modes..

I think you're underestimating the usefulness of an inexpensive but capable PCI card with basic 3D acceleration. I think these are a great value for the features they deliver.

I imagined that the chip itself might have been not even "new" from some unused chip remaining packages, but the important thing should be that beside some incompatibilities it's really sure that the card was built as supposed to and not becoming a problem for the old mainboards it might work on when compared to the original Rage series PCB quality.

I tested this on DOS and it's very fast. You need to enable the LFB option with FastVid for Quake at 640x480 but otherwise it's great.

GLQuake has some kind of flickering issue, likely driver related. You would think that this would have been fixed towards the end of driver development. ATI's minigl driver for Quake fixes the flickering but is missing some blending modes. Fitzquake works great with only minor issues in the menus.

This is not a great chip to begin with so it will take some time to figure out what kind of issues it may have with certain games. I don't think anything that will pose a problem will run fast enough to be playable anyways.

I don't know what tests to perform for problems which may exist in DOS but the few tests that I ran feel about as good as a basic S3 ViRGE. And this card has a lot more features than that one in Windows.

HangarAte2nds! wrote on 2021-09-22, 04:51:

Interesting. I just got an SIS 6236 16MB AGP which is based on the S3 Virge.

Can you take a photo of the card, please? Meanwhile, I live under the assumption that S3 Virge and SiS 6326 are two different graphics chips from two different independent manufacturers,