Problems with using GeForce drivers version 6x.xx and up

A place to discuss all aspects of games and gaming.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

mmh,

I think the problem is not directx per se, because many (if not all) opengl-games need directx too and they all work very well. directx is a "wrapper" for graphics (d3d, opengl, glide), sound, netplay, input-devices and so on.

let me summarize/conclude:

- the problem is not related to any specific chipset, but these chipsets have in common an agp-interface version 1.0
- it is not related with a specific version of agp-driver
- the problem is not related to any specific version of directx installed, it shows up with dx7, dx8(a), dx81, dx9c
- the problem showed up only so far with games that need dx7 or up AND don't use opengl but d3d; found exception to that rule: crimson skies (dx5, d3d): runs a minute then freezes
- tests so far have shown that games in 2d don't show this behaviour
- cards effected: any geforce when driver 6x.xx up is used

so I would say nvidia changed something in their d3d-code with drivers 6x.xx up which, when applied on an agp 1.0 board is responsible for that mess.

my believe that this problem is related with agp 1.0 boards is fostered by someone who had quite similar problems on an asus cubx-l board with intel bx-chipset (agp 1.0). (actually this guy and his success-story made me buy a gf6600gt for the asus p5a). he has a very long thread discussing his problems and at the end found a solution. unfortunately for you non-german speakers here his thread is in german, but I can tell you that he solved the problem as follows (I hope I translate correctly):

system win xp

goto device manager / system devices / cpu-to-agp bridge: this driver is to be exchanged manually, one should use the driver for "pci standard-pci-to-pci bridge".

thats it. I was not in the position to do this with the xp-installation on my p5a because there was no such thing as a cpu-to-agp bridge .....

for those who want to read the original german thread, go here
dominicx
Newbie K6'er
Posts: 10
Joined: Sun Oct 07, 2007 7:47 am

Post by dominicx »

DirectX does not contain OpenGL. It has D3D, but also many other components that are indeed needed by games that run in OpenGL mode (OpenGL only for graphics, sound and input still depends on DirectX). That is why you see DirectX along with the OpenGL games. The drivers for graphic cards contain both D3D (DirectX) and OpenGL modules, which are indepent. The change in the D3D module often is not present in OpenGL module, which is apparently much more backwards compatible, thanks to open, better (?) developed structure. You could see this in many cases, especially if it comes to S3 graphic cards, which never had good D3D drivers, but OpenGL worked OK (remember Savage 2000? OpenGL worked fine, but T&L was never enabled in D3D. They called it a bug in the harware, but I personally believe that D3D is harder for implementation, for both hardware and driver layers).

What I meant in my previous post is (as you concluded) that nVidia changed the code of D3D module, probably to minimize the size and to increase efficiency, letting go the K6-3DNow! optimalization and leaving only universal (or not even changed) K7-3DNow! implementation. This might be the reason of unstability.

Do you have any basis to believe that it's strictly AGP 1.0-related problem? Are there no GeForce cards running 6x.xx drivers on Intel BX, ZX, Via Apollo 133 (AGP 2x) or AMD Irongate 750 chipsets? The thing is, that in the German thread you posted, he only changed the driver from AGP-CPU bridge to PCI-PCI bridge, which resulted in having only PCI mode enabled. He showed that performance is not that much worse, but you basically loose all the profits of AGP. In fact, you already tried cards that were physically PCI, and they didn't work with these drivers. So I don't really think it is only AGP-mode related.

What you could do, is to put a PCI card, and change the driver in Windows, then try new drivers. Maybe it could affect the graphic card driver somehow?

This option of course exists, at least for MVP3, but sometimes is called differently (AGP-PCI bridge, etc., in fact I'm not even sure if it has AGP in the name). It depends on the implementation of AGP, I remember that for SIS 735 (north- and southbridge in one chip, MuTIOL architecture) you couldn't be even sure if you had AGP according to the device manager. ;)

It's easy to identify the .dll files for D3D, what if you install latest drivers (6x.xx up) and in DOS replace them with those from the last stable version? Maybe that would work? You loose some functions and optimalizations, but maybe it's possible to use this combination?
User avatar
AlleyViper
Junior K6'er
Posts: 45
Joined: Thu Dec 15, 2005 2:37 pm

Post by AlleyViper »

DonPedro wrote: goto device manager / system devices / cpu-to-agp bridge: this driver is to be exchanged manually, one should use the driver for "pci standard-pci-to-pci bridge".
That's simply running the card pci mode via the agp slot. Might be usefull if the problem is due to agp texturing or not enough current on the slot, making it work with lower bandwidth. In your p5a it should be the only driver with a reference to "agp", despite any other name it must be possible to try it on that board too.
The problem is that I'm almost sure that when I had a pci vanta lt which should have nothing to do with the agp bridge, it would bsod too in 3dmark with drivers 71.89. Still, if it helps, in the next days I'll try that driver swap on my mvp3.
Shuttle HOT-591P 512KB L3 REV.2.0, bios 591PS025 @100/66/33 1:1
K6-III+ 500 ANZ 256KB L2 @600MHz/2.0V + CM DP5-6I31C-A1, TT 6x2.5cm/7krpm/38cfm fan @7V
2X256MB pc133 CL3 @100MHz/2-2-2-5
HP GeForce2 GTS 32MB AGP
SB AWE64 Value ISA, SiS900 NIC, Trust USB 2.0 5-Port PCI
Samsung Spinpoint 80GB, Quantum Fireball 6.4GB, Yamaha CRW3200E
Mini AT case (ATX modded) with:
8cm TT Std Fan @7V
Generic 300w ATX
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

@alleyviper
dominicx wrote:DirectX does not contain OpenGL. It has D3D, but also many other components that are indeed needed by games that run in OpenGL mode (OpenGL only for graphics, sound and input still depends on DirectX). That is why you see DirectX along with the OpenGL games.
thank you, that is exactly what I meant to say.

... letting go the K6-3DNow! optimalization and leaving only universal (or not even changed) K7-3DNow! implementation. This might be the reason of unstability.
I don't think that it has anything to do with 3dnow! support. I know only of opengl-based games that are written that way to support 3dnow!-capable cpus. since they work fine with drivers higher than 6x.xx why should nvidia write their opengl-driver (6x.xx up) correctly but fail with d3d-drivers of the same driver version?
Do you have any basis to believe that it's strictly AGP 1.0-related problem?
no, I don't say it is strictly agp 1.0 related. I just summarize that I know only of agp 1.0 systems showing this kind of problems.

Code: Select all

Are there no GeForce cards running 6x.xx drivers on Intel BX, ZX, Via Apollo 133 (AGP 2x) or AMD Irongate 750 chipsets?


well, there I just happen to know of the problems this guy had with installing a gf6600gt on his bx-chipset board. have not heard of anybody else trying to install a card like this on an agp 1.0 board.
So I don't really think it is only AGP-mode related.
again, just saying that all boards used so far for the testing with these drivers all have an agp 1.0 slot only. yes, the troubles still show when a pci-card is used, but it is still used in an agp 1.0 board.

What you could do, is to put a PCI card, and change the driver in Windows, then try new drivers. Maybe it could affect the graphic card driver somehow?
sounds interesting, but please elaborate a bit more, I don't know exactly what you mean.

It's easy to identify the .dll files for D3D, what if you install latest drivers (6x.xx up) and in DOS replace them with those from the last stable version? Maybe that would work?
thats a nice idea, I will give it a try! :)
how do I identify all related d3d-dll-files?



@alleyviper
In your p5a it should be the only driver with a reference to "agp", despite any other name it must be possible to try it on that board too.
yes, but I really tried everything I possibly and reasonably could. could be that my memory failed me and it was the other way around: that I found that agp-entry but I was not able to change the driver for a pci-to-pci bridge because it simply did not show up to be selected .... where and how to get such a pci-to-pci bridge? important note here: under xp the device manager already had that standard-pci-to-pci bridge loaded (but separately from the agp-thing as far as I remember) and using win98se I was left in the cold because where to load this pci-to-pci-bridge from? there is no such thing to download (where from?) and the system does not show it by itself.
goodneit
Newbie K6'er
Posts: 6
Joined: Sat Jan 20, 2007 9:29 am

Post by goodneit »

Buyer beware: If you're looking to purchase and install an nVidia graphics card in an attempt to breathe new life into an older Socket 7-based system, you will likely run into problems with the supplied drivers, particularly if you play DirectX 3-D games.

My experiences were the exactly the same as those reported on here by DonPedro.

My setup is a PCI inplementation of the GeForce FX-5200 graphics card installed in a VIA Apollo MVP3 chipset motherboard (no AGP slot) running a K6-3 processor.

In addition to receiving errors when attempting to play DirectX games, I noticed that Microsoft's DxDiag utility aborted when it got to the "spinning cube" portion of the test designed to test Direct 3-D in DX8 mode. Attempting to rerun the same test yielded a message stating that Direct 3-D functionality wasn't available (the Test Direct3D button was greyed out).

I tried installing nVidia driver versions 61.76 and 81.98 under both Windows 98 and ME (81.98 is reported to be the last driver version which works with these operating systems). I also did a fresh install of Windows XP and loaded (current) nVidia driver version 175.19. Each test iteration resulted in the same DirectX 3-D anomaly. I then reloaded each operating system (98, ME, XP), installed an older driver version (56.65), and re-ran the Direct X tests. To my surprise, driver version 56.65 worked flawlessly for each of the three operating systems.

So the bottom line is that if you do install an nVidia graphics card in one of these older systems, you may want to install an older driver version. I found v56.65 on www.driversguide.com.
Post Reply