mockingbird wrote on 2023-10-20, 00:47:

Off the top of my head, Trickstyle would go bananas with newer Detonators... I took 8-bit paletted textures into consideration... It's overstated. Table fog OTOH I think is quite necessary.

You agreed with me on this point in the past.

Yup, and I still do. 😀 Currently, we know that Driver and Final Fantasy 7 and 8 are the most prominently affected games by the lack of paletted textures. Otherwise, it doesn't seem to be that big of a deal. I will say that Driver looks a lot better with those, but that's just one game, at least so far.

I do take back what I said however about Radeons and the very limited table fog support on 9x... In my opinion it's too limited to be satisfactory, and sometimes doesn't even work at all with the tweak. So I think newer Radeons and Windows 98 are a no go.

Yeah, it seems that every couple of months we find a previously undiscovered game that uses table fog. The list just keeps growing.

It's a shame that the Radeon fog tweak behaves in such an inconsistent manner under Win98. As it stands, the only way to get proper table fog support on Radeon cards is to dual boot with WinXP and use Catalyst 7.11 or higher.

Joseph_Joestar wrote on 2023-10-20, 03:07:

Yeah, it seems that every couple of months we find a previously undiscovered game that uses table fog. The list just keeps growing.

It's a shame that the Radeon fog tweak behaves in such an inconsistent manner under Win98. As it stands, the only way to get proper table fog support on Radeon cards is to dual boot with WinXP and use Catalyst 7.11 or higher.

Speaking of which, I do intend to post some results with the Detonator 77.72 drivers vis-à-vis its table fog support in the proper thread here... I have to dig up my Thief tower scene savegames though, I'm not sure where I put those heh. I had to go through the level and actually play it. I am taking 77.72 virtues for granted but haven't actually tested it yet, real-world.

mockingbird wrote on 2023-10-20, 03:18:

Speaking of which, I do intend to post some results with the Detonator 77.72 drivers vis-à-vis its table fog support in the proper thread here.

I have table fog working under WinXP on a GTX 970 using 344.11 drivers. I doubt you'll experience any problems with that feature on 77.72. You may get some unrelated game crashes and other issues with that driver version though.

I have to dig up my Thief tower scene savegames though, I'm not sure where I put those heh. I had to go through the level and actually play it.

I've attached them here. It's kinda funnny seeing those old savegames of mine pop up in the videos of various YouTubers. 😁

Nvidia 6k/7k series only lacks palette support. They have various other bug though. For example, can't do proper Gourad shading on color filled surfaces in Win9x environment and instead renders them with flat shading. Such surfaces were commonly used in early 3D games to safe space. Easily noticeable in PC Player Direct3D Benchmark on the logo.

The Serpent Rider wrote on 2023-10-20, 03:59:

Nvidia 6k/7k series only lacks palette support. They have various other bug though. For example, can't do proper Gourad shading on color filled surfaces in Win9x environment and instead renders them with flat shading. Such surfaces were commonly used in early 3D games to safe space. Easily noticeable in PC Player Direct3D Benchmark on the logo.

Interesting. Would the corrupted floor texture in Dino Crisis be related to that?

I did some limited testing with a GeForce 6200 a few years back, and I remember that game looking particularly bad on GeForce 6 series cards, regardless of drivers used.

Probably another issue. The floor has a texture. What I'm referencing is the surface that is filled with one simple color, which technically is not a texture, It's usually used on a primitive, i,e. part of a more complex model.
Original Lara Croft and Super Mario 64 models were famously made from primitive meshes, with some parts being color filled. Another example is the Shogo: Mobile Armor Division, protagonist hands that you see are primitives, which are just painted red. On GeForce 6800 that surface will have flat shading.

usermame wrote on 2023-10-19, 18:31:

I obtained a 6800GS (NV42) PCIe today and tested again. Indeed, the bridge chip on the 6800GT/Ultra (NV45) was the issue, but this card performs like a bat out of hell. With 77.72, I get 647.2 FPS, 100 fps more than the X800XT PCIe.

Compatibility-wise, this has to be among the best cards for Win98.

You say that because you only play OpenGL or DX8+ games. I also have an nvidia 6800 AGP (Asus V9999) and with DX7 or less just sucks. Try to play original Unreal with default DirectX engine. On my config the intro flyby nali castle runs at 1FPS or crash. I have problems with other games such missing menù, textures, and severe glitches. Furthermore, nvidia drivers that are too new break shutdown and standby as well as giving an error if you try to uninstall them.

I have now checked UNREAL GOLD under Windows 98 on G70 with ForceWare 77.72 and the official patch 226 final (There was no D3D support in the retail CD version yet and it is being added by the patch). So, DirectX 7 works flawlessly. OpenGL didn't work at first, but after adding ExtensionStringVersion, OpenGL also worked. Antialiasing AA\AFx16 makes the speed normal. Without this, the opening scene flies by faster than it did before, as I recall.

With NV4X, you can try earlier drivers as well. What is the undoubted advantage of the sixth series. Do not forget about nGlide, for the most problematic cases, when nothing else helps. In general, if we are talking about a PCI-E system, then in fact there are no other options. I do not take the PCX 5900 into account, due to its too low speed and not very good compatibility with later PCI-E motherboards, in conjunction with earlier drivers (Resource conflicts). Radeon PCI-E also do not take into account, because of the broken 2D. And the bottom line is the option with the most versatile, but one GPU, or the option with two video cards, where each will perform its own functions (PCI-E + PCI). You can try some very old PCI card. Or solve problems programmatically (Settings, wrappers, patches). Putting together another separate computer (and another and another) is the most obvious thing.

Unreal Gold always had Direct3D support. That being said, playing Unreal with originally shipped D3D API is sub-optimal, regardless of OS used.

DoZator wrote on 2023-10-20, 05:29:

Radeon PCI-E also do not take into account, because of the broken 2D.

Any more details on this?

I think the X800 cards don't support VBE 3.0 which might be relevant for some DOS gaming scenarios, but I don't remember anything else being problematic. Which issues with 2D content did you encounter?

Lack of VBE 3.0 is only problematic due to inability to switch refresh rates on-the-fly. But Radeon cards can be tweaked via BIOS. Then again, most GeForce 7 and some GeForce 6 have 60Hz VBE bug, so that argument is pointless.

The Serpent Rider wrote on 2023-10-20, 05:36:

Unreal Gold always had Direct3D support. That being said, playing Unreal with originally shipped D3D API is sub-optimal, regardless of OS used.

After installing from the original release CD, I only see these three options:

Yes, and pcgamingwiki claims the same thing in principle:

Direct3D 6, 7 Direct3D 6 was added with patch 218.[3] Direct3D 7 was introduced with patch 226, replacing Direct3D 6 support. To add in Direct3D 8 and 9 support, the unofficial OldUnreal 227 Patch is required. Direct3D 10 and 11 are downloadable mods.
OpenGL 1.2 Since patch 216.[3] OldUnreal 227 Patch adds OpenGL 2.0 support.
Glide 2.4 Glide (and PowerVR SGL) were only supported 3D acceleration options at launch.

And only after installing the patch do new APIs appear:

Joseph_Joestar wrote on 2023-10-20, 06:23:

DoZator wrote on 2023-10-20, 05:29:

Radeon PCI-E also do not take into account, because of the broken 2D.

Any more details on this?

I think the X800 cards don't support VBE 3.0 which might be relevant for some DOS gaming scenarios, but I don't remember anything else being problematic. Which issues with 2D content did you encounter?

We are talking about the low speed of 2D under Windows. This problem has already been discussed, repeatedly, starting from here (And further on the topic):

Re: PCIe devices on Windows 98 SE

And if for nVidia the problem can be completely eliminated by configuring MTRR LFB WC, returning bad metrics again in norm. Then, regarding ATI, it was only reported that positive changes did not follow after that. Pay attention to the "Stretching" parameter - for some reason it suffers the most. And TOM2D displays only the symptoms here, but the problem itself is somewhat deeper and is related to the speed of data exchange between the system memory and the adapter, which in some places without configuration is abnormally low on PCI-E cards. Which, however, does not appear in any way in D3D\OGL applications, but in GDI 2D under Windows 9x (It can be called somewhere else, of course) absolutely. Several people have already reported that these indicators could not be returned to normal on ATI PCI-E cards (for some reason, setting up MTRR did not affect this).

After installing from the original release CD

The Serpent Rider wrote on 2023-10-20, 05:36:

Unreal Gold always had Direct3D support.

DoZator wrote on 2023-10-20, 11:35:

We are talking about the low speed of 2D under Windows. This problem has already been discussed, repeatedly, starting from here (And further on the topic):

Re: PCIe devices on Windows 98 SE

Interesting findings. Thanks for the clarification!

I wonder if this performance penalty would be noticeable in games like Heroes of Might and Magic 3 or Commandos: Behind Enemy Lines.

Tom's GUI test was not designed to work in Win9x environment. So any result from there should be taken with a grain of salt. In fact, it has a minor (maybe) issue even with Win2k.

The Serpent Rider wrote on 2023-10-20, 12:46:

Tom's GUI test was not designed to work in Win9x environment. So any result from there should be taken with a grain of salt. In fact, it has a minor (maybe) issue even with Win2k.

Tom2D is just like a marker that allows you to visually check and make sure that there is a problem (You should look at the "Blitting" and "Stretching" parameters). If there is a problem, it runs for a long time, and the results are abnormally low. And so you can notice the problem with the naked eye in applications that use GDI (Especially everything related to blitting and stretching), for example, using the same KODAKPRV.EXE (Located in the WINDOWS folder), if they stretch the JPG photo to full screen and back around any angle. All these symptoms are clearly correlated with the indications of specialized instruments for measuring speed. So, for example, on my current system combined with a Quadro 4500 512MB (With NVCORE patched.VXD 77.72) by default, the results are as follows:

VideoMem Write (secs, bytes/sec): 19.3618 81235371 81235K 81M 0G
SysMem To VideoMem Copy (secs, bytes/sec): 18.5936 84591468 84591K 84M 0G

At the same time, "Blitting" and "Stretching" in Tom2D are 280 and 92, respectively. In KODAKPRV.EXE the image stretches with strong jerks.

After the MTRR correction, the results are completely different:

VideoMem Write (secs, bytes/sec): 1.0468 1502503786 1502503K 1502M 1G
SysMem To VideoMem Copy (secs, bytes/sec): 1.0481 1500749585 1500749K 1500M 1G

"Blitting" and "Stretching" in Tom2D are 2473 and 1034, respectively. in KODAKPRV.EXE stretches is much better.

Also, you can read about this problem and its various manifestations (workarounds) on the site at RayeR:
https://rayer.g6.cz/programm/programe.htm (Search for "MTRR" and beyond). But when I tried his method, I had to limit the memory to 3GB (Depends on the configuration apparently).

And in Unreal Gold (If you do not put patches), there may be some DirectX 6 (Which is replaced by the official patch with 7). I'll try to test this later, but for the purity of the experiment, I'll have to find this version (I only have a release copy, without patches, which doesn't have D3D yet).

But yes, it should work fine on the old hardware. However, even here I'm not sure if at least the GeForce FX series applies to this case?

So, I checked Unreal Gold and it turned out that this edition generally contains DirectX 7 out of the box, so you don't need to install any patches to enable D3D. And it works absolutely fine and nothing from the described usermame is confirmed. Everything works fine in Direct3D mode under Windows 98 with the ForceWare 77.72 driver on the G70 . The only thing to note is that initially (just now with vertical synchronization turned off) the game speed is too high. I didn't dig into the settings and didn't look for a way to normalize it at the engine level - I just turned on AA\AF x16 and the speed returned to normal, and the picture itself became better.

The Serpent Rider, are you sure this is Unreal Gold? I reread the original message of usermame and it clearly talks about "original Unreal", and this is a completely different edition, earlier . However, unfortunately, he did not specify which version of the patch he added Direct3D support for and with which driver version he tried. Because the original Unreal doesn't support Direct3D natively at all.

However, if you ignore Unreal, some features are already starting with the GeForce FX series (Maybe earlier, I didn't check) I've watched it myself. For example, if you install only DirectX 6 on Windows 98, then some games that support it run very slowly, as if without hardware acceleration. Does anyone know if this is fixed? Despite the fact that if you install DirectX 7 or higher, the games already work well. The problem will be on the 6\7 series guaranteed, if you do not install at least DirectX 7 or higher. Very similar to what usermame describes. I wonder, starting from which GeForce series this problem occurs for the first time?

DoZator wrote on 2023-10-20, 05:29:

I have now checked UNREAL GOLD under Windows 98 on G70 with ForceWare 77.72 and the official patch 226 final (There was no D3D support in the retail CD version yet and it is being added by the patch). So, DirectX 7 works flawlessly. OpenGL didn't work at first, but after adding ExtensionStringVersion, OpenGL also worked. Antialiasing AA\AFx16 makes the speed normal. Without this, the opening scene flies by faster than it did before, as I recall.

Thanks for the great tip on using MTRRLFBE to fix 2D performance (I assume just adding this to autoexec.bat is ok?).

How did you add "ExtensionStringVersion" for Unreal Gold and which program are you using in your screenshots to show game executable parameters please?

mockingbird wrote on 2023-10-22, 04:40:

Thanks for the great tip on using MTRRLFBE to fix 2D performance (I assume just adding this to autoexec.bat is ok?).

Yes, if you have such a problem, you can try adding this to the end of AUTOEXEC - "MTRRLFBE.EXE LFB WC" and MTRRLFBE.EXE to the root of the disk. See if the results change. It is advisable to limit the memory to 3GB or less for the duration of the experiment (Programmatically, of course).

All thanks to RayeR, for this wonderful tool that has already helped many users.

mockingbird wrote on 2023-10-22, 04:40:

which program are you using in your screenshots to show game executable parameters please?

These are the standard advanced settings windows in the original Unreal. They are called from the main settings menu. In GOLD, they redesigned the menu and hid these windows. But they are still called via the console (Preferences)

mockingbird wrote on 2023-10-22, 04:40:

How did you add "ExtensionStringVersion" for Unreal Gold

"ExtensionStringVersion" is spelled differently with different driver versions. It appeared in ForceWare 56.55 and was originally written differently (With the OGL_ prefix), then they, starting with one of the versions, abolished the postscript. In ForceWare 77.72, it now looks like this:

REGEDIT4

[HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\Class\Display\0003\NVIDIA\OpenGL]
"_UNREAL.EXE:ExtensionStringVersion"=dword:00000001

where "0003" is the number of the video card for which this parameter will apply (you may have a different value, "0000", "0001", "0002" and so on).

In general, I tested the usual "Unreal", as well as" Unreal Gold", both in OpenGL mode and in Direct3D. To add DirectX 7 support to the original Unreal, at least install the patch version 226a (UnrealPatch226a.exe) + d3ddrv.dll from version 226final. And in "Unreal Gold" 226final is fully integrated and DirectX 7 support is available out of the box. But at the same time, for some reason, some of the sounds were missing, and the other part didn't sound right. The same thing happened with the usual "Unreal", if you install the latest official patch" 226final " on it (But you can simply not install it, limiting yourself to the "226a" version). And in order to return the sound to normal, I had to replace "Unreal Gold "with " Galaxy.dll", taking it from the aforementioned patch "226a" from regular Unreal. In this case, the game sounds right again. Another problem concerns speed - installing AA\AF x16 completely removes this problem.

Yes, I think they updated the old sound engine in Unreal Gold and 226final:
http://www.hypercoop.tk/infobase/archive/unre … eleaseNotes.htm

Sound

Updated to the same sound code as in Unreal Tournament version 420.

The current version of the old engine is contained in patch "226a". So yes, it definitely makes sense to stick to the old version in this case.

But the earlier patch "225f" contains DirectX 6 and conditionally "works" (I must say, quite quickly), only the HUD and menu are not displayed.

Here, I did a little research on some possible bundles:

"Unreal", patch 225f (UnrealPatch225f.exe) + Galaxy.dll (226a), DirectX 6, ForceWare 77.72 with DirectX 7 (HardWare):

"Unreal", patch 225f (UnrealPatch225f.exe) + Galaxy.dll (226a), DirectX 6, ForceWare 77.72 with DirectX 7 (SoftWare):

"Unreal", patch 225f (UnrealPatch225f.exe) + Galaxy.dll (226a), OpenGL, ForceWare 77.72 with nvOpenGL.dll (71.84) with ExtensionStringVersion 1:

"Unreal", patch 226A (UnrealPatch226a.exe) + d3ddrv.dll (226final), DirectX 7, ForceWare 77.72 with DirectX 7 (HardWare):

"Unreal", patch 226A (UnrealPatch226a.exe) + d3ddrv.dll (226final), OpenGL, ForceWare 77.72 with nvOpenGL.dll (71.84) with ExtensionStringVersion 1:

"Unreal", patch 226final (UnrealPatch226Final.exe) + Galaxy.dll (226a), DirectX 7, ForceWare 77.72 with DirectX 7 (HardWare):

"Unreal", patch 226final (UnrealPatch226Final.exe) + Galaxy.dll (226a), OpenGL, ForceWare 77.72 with nvOpenGL.dll (71.84) with ExtensionStringVersion 1:

"Unreal Gold" + Galaxy.dll (226a), DirectX 7, ForceWare 77.72 with DirectX 7 (HardWare):

"Unreal Gold" + Galaxy.dll (226a) + OpenGlDrv.dll (2.1.0.7), OpenGL, ForceWare 77.72 with nvOpenGL.dll (71.84):

Working options, as you can see, are not few. If you have problems with one option, you can try another. In the latter case, for example with the unofficial OpenGL driver, "ExtensionStringVersion" is not even required. I found the most interesting bundle from the "225f" version of the game + galaxy.dll (from 226a). You can fully use OpenGL. Fixing D3D is already purely out of the realm of wishes, but you can try it (If you suddenly need it for some reason, when there is a working OpenGL), for example, by trying earlier nVidia drivers (Which is quite possible on NV4X/PCX series of video cards), as well as alternative unofficial D3D drivers for the game.

But in general, yes, with each game you can get tired of playing so carefully. Therefore, still, in the next PCI slot, some simple PCI video card from the very beginning of 2000, with support for DirectX 6 and old drivers, will not be superfluous (At least for testing purposes). But ideally, in general, it would be possible to write some wrapper for DirectX 1-6 => DirectX 7, so that it would make corrections on the fly. Because the example specifically with this game shows that in general, the initial support for DirectX 6 is fully accelerated by hardware, and the graphics look good. But such a small and annoying error as the lack of HUD and menu (With default DX6 drivers) - makes the use, if not completely impossible, then at least somewhat difficult. And, probably, it is possible to fix this (Which is what the developers did in d3ddrv.dll from 226final). But for other games, theoretically, there might not be such a fix, and it would be good to have a universal software tool for such old games under Windows 9x.

I've been seeing here that someone claimed that GF8 series supposedly were working under 9x but with no solid proof.
I'm willing to test that out with a nF4 machine for testing purposes.
(figured I'd go with something that is as much compatible as possible - I would've used a 915 based mobo but I'm out of them atm..)

In the (very likely in the slim to zero range) event that I get it working, what would be a set of specific tests to be run?
I'll be setting up the machine tomorrow. I've had some slight experiences with 98SE and SATA (over a 865G board, that is.), so I figured I'd give PCI-E a stab.