thanks to VDNKh for putting in the effort, when you've got another user replying with responses like the above, probably makes you wonder why you even bother.

Exactly my thoughts after putting the effort in to do the research, recommend a course of action to isolate it, find a probably good match datasheet, Linux kernel driver source (not easy to find) that deals with this exact thing, and find out the people affected don't know their ass from their elbow.

I mean really. "I have a via problem. Here's an amd register dump"

Wtf is the use of that?

Might as well send me a picture of a tree and call it the moon.

myne wrote on 2024-12-04, 14:16:

Exactly my thoughts after putting the effort in to do the research, recommend a course of action to isolate it, find a probably good match datasheet, and find out the people affected don't know their ass from their elbow.

I mean really. "I have a via problem. Here's an amd register dump"

Wtf is the use of that?

You're an f'in a$$hole! People like you don't belong in these communities.

Pfft.

I'm out. Deleting all the actually useful posts that were ignored.

Good riddance!

Good luck.

Sorry if I mess up, the comment was rude but noone has the same social skill or the same sensibility, either way the style doesn't make a statement actually false, what he claimed in a rude way may still be true.

Wow, what a specimen. Such an important topic and such a weirdo derailing it, 🤣.

Thank you VDNKh, for putting out the dumps. Hope, this'll get some traction.

its not about social skills, its about being condescending and telling someone their way of thinking is stupid which is the same as calling someone stupid, theres no excuse for that, especially when that someone is actively doing their best to figure out a problem for the benefit of the community.

if the other user had all the answers maybe he should have figured it out himself instead of spamming replies then berating someone for actually trying.

myne wrote on 2024-12-04, 14:20:

Pfft.

I'm out. Deleting all the actually useful posts that were ignored.

Jeez, okay.

I'm not sure where the break in communication is happening, but I dumped all the VIA device registers too. Like I already posted on this thread, all of the registers, except device 24 F2 and F3, are identical between 0112 and 0203. I don't see the point in double checking the registers against VIA's chipset datasheet when we know that this register configuration works on BIOS 0112. Nor do I see the point in posting all the identical registers here. I think it's a dead end and a waste of effort to examine them any further.

Nemo1985 wrote on 2024-12-04, 15:12:

Sorry if I mess up, the comment was rude but noone has the same social skill or the same sensibility, either way the style doesn't make a statement actually false, what he claimed in a rude way may still be true.

This hobby, and IT in general, does attract a lot of weirdos. I'm one of them after all.

I'm familiar with inspecting AGP registers for correctness from my work on this issue a while ago: Not Another Ultimate Windows 98 Build.

I read into the PT880 chipset's documentation back when I did that, while also comparing it to Intel's PCI documentation, and verified all the registers are set on the PT880 correctly. Since the relevant registers don't change on the K8v-MX at all between BIOS 0112 and 0203, we can rule out incorrect registers at this point. I know what I'm doing. I would love someone to prove me wrong though.

VDNKh wrote on 2024-12-04, 20:29:

This hobby, and IT in general, does attract a lot of weirdos. I'm one of them after all.

I'm familiar with inspecting AGP registers for correctness from my work on this issue a while ago: Not Another Ultimate Windows 98 Build.

I read into the PT880 chipset's documentation back when I did that, while also comparing it to Intel's PCI documentation, and verified all the registers are set on the PT880 correctly. Since the relevant registers don't change on the K8v-MX at all between BIOS 0112 and 0203, we can rule out incorrect registers at this point. I know what I'm doing. I would love someone to prove me wrong though.

I see, thank you for the clarification.

Posting the device register dumps if anyone wants to look at them. I found XP's a little interesting. I wanted to see an example of a working AGP aperture in BIOS 0203.

In XP, the "GART Table Base Address" in the AGP target device (D0F0, 1106 0204) is not programmed in the device register. Yet, the AGP aperture is functional. It ignores the set aperture size of 64 MB and uses 256 MB instead, according to 3DMark 2001 anyway. This happens in both version 0112 and 0203. I can't figure out why. Is VIA's XP GART driver doing something different? Is it XP overriding something? Is it Nvidia's display driver providing a soft-GART driver of sorts? XP SP1's included drivers are over written by all the drivers I installed. Either way, I couldn't gleam any useful information from this. Maybe someone else can. I wonder if it's possible to use RLoew's WDM extension for 98 with the XP VIA AGP driver?

Anyway, SoftICE here I come.

OS and drivers used, other info:

• Windows 98SE, vanilla, no patches applied
• Windows XP SP1, vanilla, no patches applied
• VIA Hyperion Pro v5.24a
• Nvidia Display Driver 45.23
• DirectX 9c December 2006
• BIOS settings are unchanged between all dumps. BIOS 0112 and 0203 do not reset any options, thankfully.

I had the same problem with msi k8t neo+sempron 3100+, radeon 9600. I tried many things and agp texturing was not available in dxdiag while 3d performance suffered. I tried many via 4in1 agp drivers, downgraded bios and what worked for me is to go to smartgart in ati settings and move the speed slider from 8x to 4x and after restart agp texturing went on, while 3d speed doubled in glquake from 210 fps to 450fps. I used bios 2.1, and via 4in1 524 from phils site. Radeon drivers 6.2. i then moved again speed from 4x to 8x and agp texturing still worked. Maybe this can help.

lr300 wrote on 2025-06-21, 17:47:

then moved again speed from 4x to 8x and agp texturing still worked.

You do this on every reboot? What memory value are you getting? (OP screenshot)

No,you do not need to do this every reboot. It stays 8x now.

Hi,
I had success with a K9MM-V board, latest BIOS, ATi R9100 AGP gpu, Fresh patched 98SE install using quick installer, added Via 5.24A from Phil's site, and Catalyst 5.9.
At first I couldn't get the VIA drivers to install, and it was reporting 0 AGP memory in 3DMark 2001SE, SMARTGART saying AGP disabled etc.
Managed to get the VIA drivers to install by deselecting the top option (Via INF updater) as suggested in a different thread on this site. AGP driver installed ok on reboot.
Now SMARTGART reports AGP 4x enabled, fast writes enabled, and 3DMark 2001SE shows total AGP memory as 55MB (I have AGP apperture set to 64MB in the BIOS currently).
Added about 2K to my 3DMark score. The CPU is an Athlon x2 64 4000, so likely to be GPU limited, not CPU limited. (This is my "Original Truform GPU + Fast CPU" WIP)

Happy to try other configs or provide more info if anyone would find it helpful. I have NVidia cards (FX5200/FX5900XT etc) and other ATi AGP cards (9700Pro, X800 PRO) I could try... AFTER I've backed up this working ATi config.

Tried this trick, unfortunately doesn't work on my setup.

Win98 SE EN
VIA 5in1 5.24
Pentium E6500K
ASRock 4CoreDual-VSTA
2GB DDR2
Radeon X850 Pro AGP

Switching AGP 4x or 8x from the control panel works... but AGP memory remains at 0.

You can also add the P3V133 to this list, as my system has the issue as well. Does anyone know of which BIOS version that’s best? I’ve tried all three of the common versions that are available (1001a, 1002, 1003-002) and it doesn’t seem to fix the issue.

I don't know if I've come across a one-size-fits-all solution, but I'll describe what worked in my case.

This weekend I changed the motherboard (I want chipset support for hw OPL3 soundcard and need DDMA support) at my retro W98SE rig and the choice fell to the MSI PM8M3-V H with the VIA PM800 chipset. After installing OS, drivers and software, I tested the graphics card in 3DM01 - while the previous nForce3 250 chipset had a score of around 18,000 points (Turion at 2GHz), suddenly I was only around 13,000 with the same card (with P4 @3GHz Prescott).
Of course, I know that something is going to account for the more powerful IMC in AMD CPU and the long pipeline in P4, but it still seemed too low. I remembered the VIA AGP texturing bug and of course - zero texturing memory in 3DM01 system info and the disabled, grayed and inaccessible option "AGP texture acceleration" in DirectX.

Disabling and enabling the two remaining items (Direct Draw and Direct 3D) didn't help, texture acceleration remained off and inaccessible. So I tried to find the root - reinstalled clean W98SE without SP and drivers, only DirectX 8.1 and found that AGP texture acceleration (and Direct Draw) was enabled and working. Of course, Direct 3D without an installed graphics card driver was disabled and couldn't be turned on. Next, I installed Forceware (7x.xx something for Geforce 6600GT) and then the Direct 3D option was turned on, but the "AGP texture acceleration" was disabled and inaccessible again.

Lastly I installed VIA Hyperion (v5.10a I think) drivers and after the reboot the "AGP texture acceleration" option finally became available and remained on. I tried the card again in 3DM01 and voilà! - the score has grown to just over 15,000 points, which is already the equivalent.
System info in 3DM01 reports 120MB AGP texture memory instead of zero (128MB set in BIOS, so it matches this). Everything remained functional even after the installation of the SP as well as the replacement of the GF 6600GT for the Radeon 9550 (after the installation of the driver of course).

So if this procedure helps anyone, I was happy to share - I've read many times on the Internet that I should always install the chipset drivers first after installing OS, and then the graphics card drivers and other components.

Ydee wrote on Yesterday, 15:33:

Lastly I installed VIA Hyperion (v5.10a I think) drivers

I'm very sorry, but VIA won't work at all without the 4-in-1 driver.
The drivers built into the OS distribution are practically inoperable.
For example, in the worst case, the network card might be completely missing—it's displayed in the BIOS, but not in the Windows Device Manager.
And this is a completely standard Realtek card on the PCI bus.
The 4-in-1 driver is always installed first, immediately after system installation, before any other DirectX drivers or updates.