Page 1 of 2

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

Posted: Wed Nov 14, 2007 8:04 pm
by DonPedro
In my grafic-card bench thread (which I hopefully will finish this winter) I ran into some trouble with my two gf-6xxx class cards (pny 6200 pci, sparkle 6600gt agp).

since the problems are of more general interest than it is related with bench-numbers I post my experience outside of my bench-thread, because nobody would expect such information buried over there.

geforce cards 6xxx and up need drivers of version 6x.xx and up in order to be properly installed.

with these drivers opengl-based games run perfect, but the problem with these drivers is that they do not run with directx(d3d)-based games!

the execption to this rule are the following games:

- half life
- unreal tournament
- deus ex
- expendable

remark: half life and unreal tournament are games that allow to choose between opengl-mode and directx (d3d)-mode. both modes work.

c&c red alert works too, but it is no 3d-game.
crimson skies run for a very short time (a minute or so), but then hangs.

the tool dxdiag runs the directx version 7 test with no problems, the following directx-8 test is aborted immediately and leaves the system in an unuseable state.

3dmark 99:
runs for about 2 seconds, then aborts (back to the desktop).

3dmark 00 and 01:
crash immediately, leave system unusable

some months later I digged further and made the astonishing discovery that this problem is not restricted to the geforce 6xxx and up models only. I have found out that ALL geforce card (starting from geforce 2mx and up) paired with driver version 6x.xx (and up) show the same anomaly!

the tests were done under win98se and xp for the 6xxx-models. I also tried several agp-driver version starting from 1.66 up to 2.xx.

the tests for the cards below series 6xxx were done on win98se only.

I also tried directx version 8 and 9, both show same effects.

motherboard used was asus p5a 1.03, ali-5 chipset, bios 1011.005

Posted: Wed Nov 14, 2007 9:47 pm
by AlleyViper
You can still edit the inf so that you can install older drivers with more recent cards. I guess some suport for advanced features will be lost, but you can try it nevertheless. In my experience 44.03 and 45.23 where the most stable all round with mvp3 chipsets. Later 60's would crash with d3d, and only early 50's would run every app, but sometimes they would be painfully slow as in 3dsmax.

Posted: Thu Nov 15, 2007 7:40 am
by DonPedro
manipulation of an older driver to get a gf-6-class gpu working with d3d-games makes no sense, because the old driver is a) not optimized to the new hardware (more pipelines, newer generations of vertex shaders, opengl extension etc etc) and b) does not make available to the user other htings like mpeg-decoding etc.

so in other words, what would be the point then to use a geforce 6 and later card? you get lower game speed and can not use all the other addtional features ....

you do certainly better with a geforce 59xx - card.

BUT MY QUESTION IS - and I forgot to mention it in my first posting - is anybody out there who is able to run a directx-d3d game on a super-7 system with a geforce-card (any) with driver version 6x.xx and beyond?

if so please post

game title
operating system
directx-version installed
agp-driver version
mainboard type and model
geforce model
nvidia driver version

Posted: Thu Nov 15, 2007 1:03 pm
by AlleyViper
It could be useful in the sense that almost no shader 3.0 stuff will be used as most video functions you can use in programs such as mpc are ps2.0 only (available from middle 4x. series and on). And by the same way, it would be rare to find a game playable on a k6 system that would use more advanced stuff than ps1.0, so having a limited driver still providing the raw clock and bandwidth advantage of newer cards would still be useful, but of course, in case it works.

The later card I've used was only a fx5200, so sorry I can't help you.

Posted: Thu Nov 15, 2007 4:56 pm
by DonPedro

ANY card does it!

if not I would like to hear about it.

Posted: Wed Nov 21, 2007 10:36 am
by dominicx
Hmmm, I'm a little bit confused. You concluded that no GeForce card can work on a K6-based system with 6x.00 or above drivers, but you presented in your 'Video card shootout' thread some results that you obtained for GeForce 2/2 MX using 71.84/82.16 drivers. How come?

Anyway, I suspect that it's the Alladin V's compatibility issue, or problems with the AGP drivers for it. Those cards (and/or drivers) could probably work on a MVP3 system, since (but it's just my opinion) it is more flexible if it comes to configuration (but it is also full of bugs). Do you have a possibility to check if GF 6600GT works on a MVP3 system? Maybe that's the key.

It seems that nVidia really neglects the issue of compatibility with older systems, so I will probably buy a Radeon 9800 Pro, which seems to be the highest AGP video card working on the K6 platform without any major problems. However, the power consumption of 9800 Pro is higher than 30W even when idle, so there might be a problem with stability. Although my Epox MVP3G5 has apparently one of the best AGP implementations among MVP3 boards, I'm not quite sure if the external power will recompensate the lacking Watts. :/

Do you consider trying any ATI cards?

Posted: Wed Nov 21, 2007 4:13 pm
by DonPedro

as I have stated in my very first post, the drivers above 6x.xx run perfectly open-gl-based games, but have big troubles with running d3d-based games.

so in my bench-table you will see that the bench numbers taken with 7x.xx. or 8x.xx drivers are bound to open-gl-games (with the exception of "expendable", - I have forgotten to mention that d3d-game in my first post, but will edit it accordingly)

so in case you have any geforce-card around, just try and see for yourself! and please report back, whatever the result. I really would like to know if anyone comes to the same conclusion.

edit: and yes, after I am through with nvidia-cards, I will also send some ati-cards through the bench-parcours:

- radeon 7500 pci
- radeon 9100 pci
- radeon 8500 agp

all of which feature 64mb videoram

finally I will test a

- kyro2 64mb agp

edit #2: yes, I do have a shuttle hot 597 mainboard with via-chipset. In my memory very far away I think I have tried that mobo too - with same results - but I am not sure anymore. mmmh, I will give it a try, but it will take some days before I can start with it.

Posted: Wed Nov 21, 2007 4:29 pm
by DonPedro

your sig says you are using a gf-2-gts on a via-based board. could you give your card a try with some 6x.xx up driver and test it with dxdiag / 3dmark01?

Posted: Wed Nov 21, 2007 7:35 pm
by AlleyViper
I've already did, tried on a gf2gts/fx5200/vanta lt PCI and not sure if I tested it too on a TNT1. It's a no go as said above.
I tried up to 71.89 which were the last ones to officially support tnt (via extra the supplied *.inf). 3m01 always crashes with a blue screen but glquake or q3 still runs. Both on this mvp3 shuttle, and in a similar baby-atx soltek mvp3 board. Tried both with a k6-iiip and this k6-2+acz, and I'm not completely sure if these tests were also done on w2k.
There were some early 5x.xx that wouldn't crash too, but after some gl timedemo runs they would start to stutter a lot so I'm always back to the 4x.xx series.

Posted: Thu Nov 22, 2007 10:55 am
by kalabok
what is the most simple test for that kind of d3d stability/support you are talking about?

Posted: Thu Nov 22, 2007 3:00 pm
by DonPedro

the most simple one is the one that comes with directx installed. it is the tool dxdiag and you can start it by going to start/run and keying in "dxdiag" (without quotes).

Posted: Mon Nov 26, 2007 1:06 pm
by DonPedro

this weekend I chose to run the tests again on my shuttle hot 597 board (mvp3-chipset) to get sure that my memory does not trick me.

I experienced the same problems as described for my p5a board. I did the tests with a gf3-ti200 card and the sparkle 6600gt.

the gf3-ti-200 works excellent with driver 14.10 (and will do up to below 6x.xx). then I tried driver 82.16 (latest for win98se) and driver 66.94. both failed in the way I described for my p5a.

the 6600gt paired with this mainboard performed even worse. I also tried 66.94 and 82.16, both driver versions were not even able to correctly install this card. the desktop was bound to 640x480 @ 16 colors. the reason I found out was that the card had ressource allocation problems. the device manager told me that a) the acpi-bios device uses ressources the card needs and b) after deactivation of the acpi-bios support in the bios the device manager showed a ressource conflict with some other system-device I don't remember now. I gave up to find a solution, maybe with more time and effort one will be lucky to run the 6600gt on this board.

so in other words, I advice you to not even think about buying a geforce 6xxx for using it with your shuttle hot 591 board. but this is not bad news at all: try to find instead a geforce fx 59xx - card. it is certainly cheaper and you will be able to run this card with your board with some certainty as long as you use driver version below 6x.xx AND can use d3d - games that use dx-8 or dx-9. also I think that the performance difference to the 6600gt model will be negligible. last but not least as far as I know this card also has the "new" cine-fx engine built into the gpu for better mpeg support (models fx 5200 and up).

dominiicx, I just realized that it is alleyviper who uses the shuttle hot 591 board, but I think that the problems will be the same with the mvp3g5 - board ....

Posted: Mon Nov 26, 2007 8:21 pm
by DonPedro

I wanted to narrow down the problems and so I went to through the whole procedure again, but this time on my asus p5s-vm board (sis530 chipset) with latest bios by jan steunebrink (bios is actually for the asus p5-99vm, but works on both boards).

I used a geforce-2-mx card for the pci-slot and run it with 44.03 driver. result: no problems at all. then changed driver to version 66.94. result: same problems as with the ali-5 and via - chipset board already reported! back again to driver version 52.16: everything is ok again.

summary so far: all 3 supersocket7-chipset boards (ali-5, via mvp3, sis530) show the same abnormal behaviour when driver version above 6x.xx is used.

one guy over at a german forum has told me that he runs an asus p5a 1.04 with winXPsp2, directx9c, geforce2-mx card SUCCESSFULLY using driver 66.93. he has no problems running 3dmark01 or the dxdiag-tool. I have asked him 2 or 3 times throughout the thread over there to publish the information regarding agp-driver used and also the registry entries for the agp-driver settings and nvidia-driver settings respectively. for a reason I don't know he does not deliver this information ..... .

Posted: Mon Nov 26, 2007 8:49 pm
by AlleyViper
Thanks for your effort in sorting chipsets out, it will surely prevent a lot of headaches. I wonder if that user had some misconfiguration as running in pci mode or with agp texturing off, who knows. I guess it sorts out being a simple agp driver problem specific to a chipset too.

Posted: Sat Dec 01, 2007 11:15 am
by dominicx
It looks like a wrong DirectX driver optimalization then. If it's impossible to run these card on neither MVP3, AladdinV, nor SIS 530 boards, it must be a problem with compatibilty between the driver and AMD K6-2/III(+).
Maybe nVidia doesn't keep the special 3DNow! optimalization for K6 systems anymore, and it decided to unify this optimalization along with 3DNow! that was implemented in Athlon CPUs - if that is the case, the driver crashes randomly in 3D environment. It's clear that 3DNow! implementation in K7 was different (much more efficient), not to mention the new set of Enhanced 3DNow! instructions. The new 5 DSP instructions in K6-2/III+ CPUs can be another factor that only affects the stability/optimalization, since its implementation differs from the original K7 design.

Anyway, it seems that Ati is much better in this field, but for how long - who knows?

Is anyone willing to test different versions of Catalysts? When I am able to play with the Radeon 9800 Pro I will share my observations, but for now I can only wait. ;)