Now this is a ugly situation concerning VIA motherboards

[Pabster]

AA0 wrote:

"AMD isn't big enough, they can't afford to make their own chipsets, it keeps them from making cpus, which make a lot more money. The 760 was an amazing chipset, really wish more boards would have used both AMD chips instead of the VIA southbridge. You can put any intel chipset against that."

AMD760 was the standard for a long, long time. It is an incredibly stable and high performing chipset, every bit as good as anything Intel has offered. What set back the crop of 760 boards was the ridiculous inclusion of the ubiquitous VIA 686B south bridge. 766 could have been substituted for a few dollars more and would have saved a lot of headaches. In fact, I'd probably still be using an 8K7A if only 686B were missing. Even disabling the on-board IDE, I was unable to resolve the performance deficiencies from 686B.

 

[Benedikt]

What about the SIS or ALI chipsets? Do they have the same PCI problems?
Tecchannel writes about the 760MPX and its two PCI buses; one is slow like VIA, the other is quite fast. Quite interesting too, that AMD obviously has the same problems (at least with one PCI bus).

 

[AZGamer]

<< In fact, I'd probably still be using an 8K7A if only 686B were missing. Even disabling the on-board IDE, I was unable to resolve the performance deficiencies from 686B. >>



There were no performance deficiencies with the 686B that were not caused by faulty drivers. After installing the latest drivers and 4 in 1s, performance was excellent.

 

[L00PY]

<< There were no performance deficiencies with the 686B that were not caused by faulty drivers. After installing the latest drivers and 4 in 1s, performance was excellent. >>



As usual, totally wrong. AMD themselves described the 686B as containing "two system-level incompatabilites". Their recommended solution for the common one most people face involves reconnecting the WSC# signal in a new configuration. This involves (among other things) connecting a previously unconnected resistor pad, adding a zero-ohm resisitor, and using a pull-down resistor. Hardly sounds like faulty drivers to me. After doing this, "data coherency issues are avoided". Then again, maybe a data coherency issue is part of excellent performance, for you and VIA at least.

But actually backing up the facts with actual references (like an AMD white paper is not your way at all AZGamer. I guess not everyone has the blind faith as you seem to hold in VIA, AMD included.

 

[AA0]

Should have bought the DFI board, weren't they the only ones to offer the whole AMD package?

 

[Wind]

Its been posted many times. The issue is not an issue in most cases. I agreed tht the PCI latency probs do occurred in the VIA chipsets (tested).

The patch do circumvent the prob. The most affected user will be those running RAID on their system.

There r 2 ways to go bout it
1. Install the patch.
2. Increase the PCI latency from the default (most mobo) of 32 to 64. It will solve the probs even w/o the patch.

For users not using RAID. Keep in mind tht u r not greatly affected & the performance decrease is not noticeable running daily application.

As for other PCI components...even w/ the bug...ur PCI components (other than RAID) hardly reach the available bandwith currently provided by the chipset.

Hope this clear the anxiety.

 

[Harvey]

Excuse a request from someone who doesn't visit this forum often enough, but anyone have a link to the patch?

TIA.

 

[AZGamer]

See the messages below for a link

 

[Harvey]

Two whole freaking minutes! What took you so long? :Q

Thx.

 

[Harvey]

I get the message, "You are using some sort of privacy-protecting software. Please disable it and try your download again."

It's probably my Zone Alarm, but I don't disable that for anyone, especially when going to an unknown site. I just did a search for VIA latency on Google.com. Here are some better links:

George's patch, ver. 1.4 (comment: -UPDATE- This updated version may cause overheating in some systems.)

George's patch, ver. 1.0 (comment says no overheating problems)

Several intersting downloads

<updated to fix links>

 

[AZGamer]

I know I'm going to regret this when I get my hosting bill (you guys know who you are), I'll host it on my site:

Linkified on my site

 

[AZGamer]

The latest one is 1.8 - the one on my site.

 

[WhiteWizard]

I've seen mainboards with VIA chipset with horribly bugs. For example: The EPOX brand, I already coment the "green line issue" present in the MSBackup of Win98SE, search it on their FAQ, also the 7KXA (Slot A) were the poorer IDE preformer of its class. I also seen this board corrupt the HD when trying to install Win XP with a Western Digital ATA 66 7200RPM.
The "green line issue" (or selfreset after loading desktop) were also pass along to the 8KTA (first socket A), however, others brands with same chipsets don't have this problems, so perhaps were a bunch of missbehavers VIA chipsets, or maibe is THOSE CHEAP AND UNTESTED MAINBOARDS YOU ARE BUYING. People make this as a rule:
avoid any mainboard that says in the ad: "First VIA xxxxx mainboard!!!!".

Alejandro

 

[WhiteWizard]

And finally eat this: My single IBM Desktar ATA 100 15Gb gives me 21000 points of Sandra 2000, and more than 35500 points with my actual RAID 0. That is poor performance for a 6 month machinne???
The RAID 0 is composed by two IBMs 15b, and were working flawsly for 6 moths with Win98SE, and now seems to work as good on Win XP PRO.
Using a A7V133, Athlon 1200 at 1333, 266Mhz bus.

 

[markrb]

Actually the newest version of the PCI Latency patch by George Breese is version .19 and can be
downloaded at www.viahardware.com.
According to Via hardware Via has admitted the problem. Here is a qoute from viahardware.com.
"VIA has acknowledged the bug, and their own fix can be found on our downloads page, as well as integrated into all 4in1 version 4.31 and newer."

Mark

 

[Grendel99]

Has anyone gotten the "Official" Patch? THe one that supposedly comes with the new VIA 4n1's? Any difference?