VIA (accidently) admits to PCI bus problems

[Gstanfor]

http://www.tech-report.com/

Darth VIA unmasked!
by Scott "Damage" Wasson - 08:36 pm, June 5, 2002

COMPUTEX ? A little birdie told me about some very interesting events that took place here around or about Computex yesterday. Apparently, a VIA engineer was speaking directly to members of the press about the VIA PCI bus problem that has caused performance and compatibility troubles for some users, and he let slip the precise nature of the problem. It seems VIA's south bridge chips didn't include a PCI extension called bus parking that Intel implemented in its post-BX series of chipsets. Many PCI card makers simply assumed bus parking would be available to them, and when it wasn't, all heck broke loose?snap, crackle, pop on your SoundBlaster.
That's what I hear, anyway.

Driver patches seem to have resolved the worst of the problems for most folks, but PCI card makers have to play along in order to resolve such issues.

VIA claims its newer south bridge chips, especially the 8235 chip showing up in P4X333 and KT400 prototype boards, do not suffer from this problem. The new PCI bus, they claim, performs like a champ. We'll test that theory ourselves soon.

Remember, though?always two there are: a master and an apprentice.

So, there you go. VIA's own engineers admit there is a problem with theire PCI implementation.

What do all those who defend VIA and refuse to believe there is a PCI bug have to say now?

Greg

 

[kLezViruS]

Originally posted by: Gstanfor
http://www.tech-report.com/

Darth VIA unmasked!
by Scott "Damage" Wasson - 08:36 pm, June 5, 2002

COMPUTEX ? A little birdie told me about some very interesting events that took place here around or about Computex yesterday. Apparently, a VIA engineer was speaking directly to members of the press about the VIA PCI bus problem that has caused performance and compatibility troubles for some users, and he let slip the precise nature of the problem. It seems VIA's south bridge chips didn't include a PCI extension called bus parking that Intel implemented in its post-BX series of chipsets. Many PCI card makers simply assumed bus parking would be available to them, and when it wasn't, all heck broke loose?snap, crackle, pop on your SoundBlaster.
That's what I hear, anyway.

Driver patches seem to have resolved the worst of the problems for most folks, but PCI card makers have to play along in order to resolve such issues.

VIA claims its newer south bridge chips, especially the 8235 chip showing up in P4X333 and KT400 prototype boards, do not suffer from this problem. The new PCI bus, they claim, performs like a champ. We'll test that theory ourselves soon.

Remember, though?always two there are: a master and an apprentice.

So, there you go. VIA's own engineers admit there is a problem with theire PCI implementation.

What do all those who defend VIA and refuse to believe there is a PCI bug have to say now?

Greg

SHOCKING!!!

Thank god i ditched the AMD/Via and I'm VERY happy with Intel/SiS combo.. =)

 

[VFAA]

This is interesting news.

I run a ABIT KT7A with a VIA chipset and I haven't had any problems. I also have a Sound Blaster Audigy Gamer and it runs smooth.
I never wanted to go AMD but money forced me to. Now I can go Intel anyday but I'm quite happy with the AMD and VIA for now. Haven't had a problem yet.

 

[SHoddyCOmp]

Gar, are there any really good SiS AMD XP boards?

 

[GT1999]

Well, it's nice to know the root of the problem is now known.. very interesting to say the least. I'm glad to hear the problem is supposedly fixed now with later chipsets, too...

I wonder how the VIA engineer feels now that he spilled the beans

 

[SHoddyCOmp]

Probably wondering what job to look for next

 

[kgraeme]

It's some consolation to know that VIA at least knows they are to blame. I knew they were to blame, but that didn't really help.

 

[Adul]

like that was a big secrete, but now we know why at least.

 

[jaeger66]

Intel 850/E

Page 12. Number 5.

 

[Lord Evermore]

So, is "bus parking" a part of the PCI specification, or is it something that Intel created and tacked onto their implementation of it without having it integrated into a new PCI version (immediately or ever)? If it's part of the spec, then it's VIA's fault. But if it isn't, or wasn't at the time, then it's not VIA's fault that they produced a spec'd chipset while the cardmakers followed a non-spec system. I assume it was a part of the spec, but it doesn't make sense that VIA would completely leave out a part of the specification (especially after SO many different chipsets and they're only now adding it); a bad implementation of the spec I can understand, but not including all the base-functions of the spec seems odd.

 

[Gstanfor]

I wonder how the VIA engineer feels now that he spilled the beans

This is what gives the story it's distinctive reek IMO.

VIA has never admitted to the PCI bug, never helped end-users to fix it, instead they have denied it, covered it up, ignored it and prayed it would go away.

But you can't hide the truth forever.

What I would also like to know is how many reviewers knew this? and if they knew why did they not mention is product reviews?

Greg

 

[Lord Evermore]

jaeger66:

The link

I think you used the http link for an FTP site.

 

[jaeger66]

You're right, I just realized it after I posted!

 

[Linux23]

where is Pabster?

 

[BD231]

Originally posted by: kLezViruS

Thank god i ditched the AMD/Via and I'm VERY happy with Intel/SiS combo.. =)

 

[neuralfx]

So, is "bus parking" a part of the PCI specification, or is it something that Intel created and tacked onto their implementation of it without having it integrated into a new PCI version (immediately or ever)? If it's part of the spec, then it's VIA's fault. But if it isn't, or wasn't at the time, then it's not VIA's fault that they produced a spec'd chipset while the cardmakers followed a non-spec system. I assume it was a part of the spec, but it doesn't make sense that VIA would completely leave out a part of the specification (especially after SO many different chipsets and they're only now adding it); a bad implementation of the spec I can understand, but not including all the base-functions of the spec seems odd.

that's a very good point that i don't think anyone mentioned .. however you feel about via .. this is a valid point ..
-neural

 

[Diable]

Bus parking is PCI 2.2 standard. This quote is from the PCI 2.2 Compliance Checklist:

1.13. PCI Bus Master Parking

GENERAL FUNCTIONAL DESCRIPTION

Verify the ability of IUT to drive PCI Local Bus
to stable conditions if it is idle and GNT# is asserted.
Verify that the IUT can arbitrate and win or lose ownership of the bus.
Verify that the IUT can tolerate other transactions on the bus.

METHOD OF VERIFICATION

Program arbiter to park the master on the PCI Local Bus

CROSS-REFERENCE TO PCI LOCAL BUS SPECIFICATION 2.2
Page 74

FLOW

o Establish known Initialization State

o Bus Parking - Compliance Checklist Tests 1 - 4

- Program the arbiter to deassert GNT# to the IUT.

- IUT: idle state.

- Program arbiter to assert GNT#

- PC: Verify IUT drives AD[31::00] and C/BE[3::0]# to stable values within eight PCI clocks. PAR to follow one clock later.

- Program arbiter to deassert GNT#.

- PC: verify IUT Tri-states the above signals in one PCI clock

I've always heard that SB Live cards hog the PCI bus on boards using VIA chipsets and how they don't play nice with other bus hogging cards, I guess we know why now.

 

[moronNZ]

Gar, are there any really good SiS AMD XP boards

I think MSI 745 Ultra is very good...............It got good oc capability........It's the second most talked about SIS 745 AMD XP board. It's worth checking the discussion about this board on ocworkbench forum.

 

[RicardoMPX]

I know SiS' socket A chipset performs well....... I suggested my friend to buy MSI 745Ultra, and all the instability problem he had with Gigabyte's MB (based on VIA KT133A chipset) went away immediately. The board performs very well....and he is very happy about it.

 

[MrGrim]

Originally posted by: Lord Evermore
If it's part of the spec, then it's VIA's fault. But if it isn't, or wasn't at the time, then it's not VIA's fault that they produced a spec'd chipset while the cardmakers followed a non-spec system.

I don't care if it's part of the spec or not, I honestly don't. What I do care about is the chipset I'm buying to be PROPERLY TESTED. It's not like the problem only occurs with some rarely used hardware; take the Sound Blaster Live! for example, almost everybody has had one at some point. And to top it all they let Sound Blaster and the rest of the hardware makers take the blame even thought they knew it was their crappy implementation.

This is enough proof for me that VIA is not worth having in your machine.

 

[Gibzilla]

It may seem a bit harsh but i'm compelled to say this. I've been in PC repair business for IMO, a long time since 1996. VIA can't produce quality products to save their lives. Back in them days they used to manufactured keyboard controllers that actually bursted into flames. The VIA and SIS were running jokes in the back dungeons of repair departments along with SIIG. Maxtor didn't fare well either, they were dubbed bad sector maxtor aka suxtor. It seems maxtor has shaped up recently though.

Personally, I wouldn't touch VIA stuff with a 12' pole. Your milege may vary.

p.s if you knew how we manhandle your computer and laugh behind your back, you'd never bring it into our shops. But you didn't hear that from me , right?

 

[Insane3D]

"snap, crackle, pop on your SoundBlaster. "

Actually, the problem with the SB Live cards was more a result of a flaky ACPI header in the drivers than a direct conflict with Via hardware..


Anyway...continue on with the Via bashing...

 

[DaFinn]

OK, so VIA is bad? But, I have had a EPOX 8KHA+ board for 5 months now, and I am using SB LIVE, and Promise PCI ATA controller, and have not had any problems??? Absolutely 0 stability issues, sound works great without cracks, ATA controller is roughly 10% faster than the onboard IDE...

So, is there something wrong with my board since it works great?

 

[MrGrim]

Actually, the problem with the SB Live cards was more a result of a flaky ACPI header in the drivers than a direct conflict with Via hardware..

Anyway...continue on with the Via bashing...

Proof? Link? All I can find is the following.

Quoting John Gatt from the VIAARENA team:

The issue of the SBL problem was to do with data transfer

 

[Pabster]

Gibzilla wrote:

"Personally, I wouldn't touch VIA stuff with a 12' pole. Your milege may vary."

And I wouldn't touch it with YOUR 12' pole

It's a little late for them to admit anything. The issues surrounding VIA chipsets are well documented. And PCI bus problems are just the beginning. Save yourself a lot of headaches and avoid them like the plague.