New nvidia Drivers don't work with VIA chipsets in WinXP

[GT98]

And this time its not Via's fault!

Link

Scott

[Kosh Naranek]

Just a small correction ... it's Both VIA and Intel chipset's which is having problems with the new 23.11 Drivers on XP !

[KeiFront]

Very bad QA.

[The PIT]

Must have an ex Via Qa person on the job

[KeiFront]

LOL

[The PIT]

On viahardware.com they reckon this has unearthed yet another Via chipset bug which there is a patch for.
Surprisingly my pal Roj hasn't blamed Creative for it and slagged Via off for once.

[Technoid]

I'm glad that I always wait a week or so before using reasently released drivers...

[The PIT]

Sounds like typical Via pass the buck like they did over the 686b corruption problem where they blamed motherboard maunufacters and Creative.

How did that go.

1) It happens on all boards including Intel. Mmm haven't seen it at work and we've got loads of computers.
2) Change the soundcard.
3) Changed it and now it only rarely crashes.
4) Or I haven't got a soundcard. Response deadly silence.
5) I've already got a Santa Cruz (Okay Via didn't recommend any cards but some sites did) and I've got data corruption. Deadly silence.

I come up with this scenirio for the 686b bug.

1) Slight problem of data corruption. Only occurs on rarely

Can be cured with Bios updates and driver updates.

2) Moderate data corruption. Machine trashed several times a month.

Bios updates/ Driver updates Soundcard swap or install alternative ide controller and disable on board ide.

3) Severe Corruption Machine can only run a few hours without trying to trash operating system. No soundcard fitted or screws up if even if soundcard is removed on next re-install.

Bios updates / Driver updates disable on board ide and use alternative controller slow memory timings down disable loads of bios settings or rma board and get another without the 686b crap.

[Kosh Naranek]

As I said ..... this IS not just a VIA or Intel bug .. it's an nVIDIA bug too ! Period !

There is a unofficial BETA patch floating around on the web, made by VIA which closes the RX55 register

Before VIA released this patch, many Motherboards manufactors upgraded their BIOSES to do the same thing on both VIA and Intel chipsets.

RX55 Register :
A memory timing device included in the main host controller on the GFX card !
The RX55 register is a memory write que timer !
It's main function is to prevent write data to be held in the memory queue too long !
When data in the queue times out it is given a higher write priority ! However this procedure fails when using an overloaded method as it's being done under Windows XP and that's why Microsoft recommends turning Write Combining off to solve this problem !
The NEW nVIDIA drivers uses much more memory than earlier drivers which is one reason why this problem is becoming so widespread ! and nVIDAI is in a competive situation with ATI so their drivers need to be fast --> pushing everything which can be pushed !

My point being ... VIA may or may not be a part of the problem but they are not the only ones to blame for this !

BTW
How many of you, who posted on this matter has actually experienced this problem ! ( VIA chipset and nVIDIA gfx card running on XP ) ?

[dZeus]

though it's funny to see VIA chipsets crap out when they're stressed out a LOT (like with Live! and now nVidia drivers).

Yet another reason not to get near VIA. Never had a problem-free experience with VIA chipsets (or at least as problem free as with Intel chipsets) and I won't go the VIA chipset route untill they prove to be just as compatible, stable and fast as their Intel equivalents.

[Kosh Naranek]

though it's funny to see VIA chipsets crap out when they're stressed out a LOT (like with Live! and now nVidia drivers).

You still don't get it, do you .. this is NOT just a VIA issue !

Funny thing is ... it's only on XP not 9x or 2K but only XP and even Microsoft gives a solution to the problem although without admitting anything !

http://support.microsoft.com/default.aspx?scid=kb;EN-US;q293078

And I will ask again : Has anyone posting in this thread experienced this problem ?

[Kosh Naranek]

Of course VIA chipsets sucks !

The above is being done on an EPOX8KHA+ with a new beta bios allowing a new memory divider !
And it's all being done with standard voltage etc.

Oh Yeah !

[The PIT]

Not a Via problem. Funny that they identified the problem, it's due to the mwq timer and this has to be disabled. Remember the 686b bug was also supposed to effect every chipset as well.

However some people maybe thinking they have this problem when it's related to another bit of hardware. I wonder how many have pulled everybit of hardware bar the video card and done a clean install to see if the problem reoccurs.

Also there seems to be confusion with some people with what chipset they're using. A few seem to think that since there running an Intel PIII there running on a Intel chipset. I've seen a few posts like that and if you look at the motherboard it's a Via chipset.

[dZeus]

Originally posted by Kosh Naranek
You still don't get it, do you .. this is NOT just a VIA issue !

I never said that in the first place?

Maybe I should have been more clear: It's funny to see VIA chipsets flake out when stressed out more easily than other better designed chipsets.

[Gurm]

So I get this shiny new GeFart3 Turd200 on sale at Best Buy's 5-hour extravaganza for $100. Great, right? Well, on any 2x.xx driver I get the damn nv4_disp.dll infinite loop.

Bleh.

So I go to the beta Detonator 23.10's and the problem goes away. I haven't tried the 23.11's yet, and I don't think I want to. VIA is making noise like you'll have to install the latest 4-in-1's to get this patch, and there's NO ****ING WAY those turds are going on my system ever again.

- Gurm

Edit: Downloaded the latest WPCREdit, and it looks like MSI has done their homework for once, as this patch is pre-loaded into my BIOS! Woohoo! Yay for intelligent motherboard manufacturers.

For those of you wondering, the fix for this is simple... fire up WPCREdit, make sure it knows about your motherboard (actually lists the device name, not just a hardware ID), and set register 55 to 0. NOTE THAT THIS DOES NOT WORK ON 266 CHIPSETS. On those chipsets, set register 95 bits 5, 6, and 7 to 0.

Remember to use WPCRSet to make the changes semi-permanent, and have fun!