Does the Via PCI latency bug affect hybrids (AMD 761 + Via 686b)?

[bassoprofundo]

I haven't been able to get a clear answer on this. I don't have any of the SBLive! issues, but then again I also don't use the IDE controllers (SCSI all the way..) I was mainly concerned about the disk I/O rates and the slowdown mentioned in the big article at TecChannel. Can anyone clear this up for me?

 

[rickn]

I suppose it could. The southbridge is what controls the pci bus. The problem with the SB Live was really more a communication issue between the VIA northbridge and VIA southbridge. I've never had any SB Live or Audigy problems with my 8k7a

 

[Wildmana]

I was thinking of buying a KG7 some time back and asked the same question.

A peep replied, his reply was the 760 doesn't have the same buss arcitecture between north and south bridge that a full VIA solution has. So therefore you do not have the 686 bug. I don't know if thats right ro not. But the AMD760 sure seem stable........

 

[AA0]

the "bug" doesn't effect stability. It can also be fixed by optimizing your bios, don't need a patch.

 

[Athlon4all]

The bug doesn't effect most people (Only people using RAID or SCSI) but for those that performance is decreased by this bug, then yes, they will have performance losses on AMD 761+VIA 686B.

 

[Howard]

<< The bug doesn't effect most people (Only people using RAID or SCSI) >>


That's not what I've seen at all.

 

[Athlon4all]

<< That's not what I've seen at all. >>

Ok, bear with me. What have you seen? Here's why that statement is mostly true. This bug limits how much of the full 133MB/ps of the PCI bus is used, and the reality is that only SCSI and RAID use so much bandwidth that the bug starts to play a role. Everyone is "effected" by it, but only those that are using SCSI or RAID are using so much PCI bandwidth that the bug starts to effect performance. And plus for those, there are patches avialable.

 

[bassoprofundo]

Wildmana,

I'd heard the same thing, that it was really an issue with the communication between the Via Northbridge and Southbridge that was the problem for the bandwidth issue. I guess the only true way to know is to test it, but other factors can throw off results there too. I run an Ultra160 SCSI setup, so I would have concerns about it limiting my burst transfers. After a little digging on ViaHardware.com's forums, I found that in the description of his patch to resolve these issues, George Breese lists the MVP3, KT133/a, and KT266a as the chipsets affected by this problem. He doesn't mention the 761/686b combo. I'd like to just post and ask him, but it's probably been discussed to death somewhere over there, and I'm really in no emotional state to be flamed... Like I said, I haven't noticed any problems, so I guess I should just be content with that... Notice I said "should be"...

 

[Pabster]

Yes, the PCI latency issues affects all VIA south bridges, 686B is no exception. Which makes perfect sense, seeing as how ubiquitous 686B is.

I can speak from first-hand experience with 761/686B (8K7A) ... great board, AMD760 is stable as a rock. But 686B was nothing but headaches. I eventually decided to disable everything possible that the 686B controlled. I ended up throwing in a 29160 and a 10K Cheetah. Not surprisingly, transfers were pathetically slow. The same configuration dropped in to another (non-VIA) board scored nearly 50% better, and the performance difference was definitely noticeable.

 

[ST4RCUTTER]

Yes, the PCI latency issues affects all VIA south bridges, 686B is no exception. Which makes perfect sense, seeing as how ubiquitous 686B is.

You wanna know what's really ubiquitous? Posts by you, describing the 686B as ubiquitous. GH is littered with those... My god man find another adjective!

 

[bassoprofundo]

Pabster,

I've used both the FIC AD11 and the Epox 8k7a (using now), and it's been a while since I've put the SCSI setup on anything without any Via influence, so I haven't had anything to reference it against. They have been rock stable for me though. I know you're big on the SIS setups, and I'd like to switch over, but I don't know that I'm comfortable using the ECS board, which seems to be the only SIS-based Athlon board out and in mass production besides the Chaintech Micro-ATX (blech...) board. I know Leadtek makes one, but it's more expensive and harder to find, and they aren't really proven in the motherboard market yet. The ECS I sold you was supposed to go in a machine I was building for someone on the cheap. I probably should've just kept it and tried it out for a while. My only other alternative here (besides going back to Intel) is an NForce board, but those are still expensive as heck, too (and we don't need no stinking MX video...)

 

[Athlon4all]

<< is an NForce board, but those are still expensive as heck, too (and we don't need no stinking MX video...) >>

First of all, MSi K7N420 Pro is now at the same price as Dragon+, KR7A and A7V266-E, so I dunno what you're griping about, and if you still won't buy MX video, then in a week or 2, new nForce 415-D boards will be released which feature Twinbank but no MX video, and rumors have prices around $100!! That still too expensive?

 

[Howard]

My apologies Athlon4all, I thought you were responding to the topic question (the original VIA 686B +SB Live! bug, not the slowdown one).

 

[Viper22]

Ok so what patch and where do i get it for my system if it needs it? Im running an Adaptec 39160 with a Cheetah X15 Drive as well as a 10K Cheetah. I have an IWill KK266-R Board. Thanks. In also running WinXP Pro incase that matters.

Viper22

 

[bassoprofundo]

The patch I was talking about is the PCI latency patch found at www.viahardware.com.

On the subject of the NForce boards, nobody's griping herel. I'm just not interested in paying $140+ at the moment for a motherboard with more gadgets than I need. I don't need RAID controllers or onboard video or 5.1 sound. If I was building a system for someone, I'd recommend the NForce hands-down because the integrated goodies are decent and lower the cost of the system overall. For my needs, the onboard stuff just adds price. The other boards you mentioned are pricey, too ($120-150). I'm trying to keep it under $100, so if the 415-D boards are around that mark, I'm sold!

 

[Pabster]

bassoprofundo wrote:

"Pabster,

I've used both the FIC AD11 and the Epox 8k7a (using now), and it's been a while since I've put the SCSI setup on anything without any Via influence, so I haven't had anything to reference it against."

Understandable. If you've been stuck with VIA, you won't realize the performance difference -- especially with RAID/SCSI adapters which occupy the PCI bus. I can tell you I saw a full 50% increase in synthetic benchmarks, and a very noticeable performance difference ("real world") when I moved the same setup from my 8K7A to K7S5A.

"They have been rock stable for me though. I know you're big on the SIS setups, and I'd like to switch over, but I don't know that I'm comfortable using the ECS board, which seems to be the only SIS-based Athlon board out and in mass production besides the Chaintech Micro-ATX (blech...) board. I know Leadtek makes one, but it's more expensive and harder to find, and they aren't really proven in the motherboard market yet. The ECS I sold you was supposed to go in a machine I was building for someone on the cheap. I probably should've just kept it and tried it out for a while. My only other alternative here (besides going back to Intel) is an NForce board, but those are still expensive as heck, too (and we don't need no stinking MX video..."

I'm big on the SiS alternatives because they perform. And they don't milk you to death in the wallet. And you don't have to go hunting down the latest latency patch and 4-in-1 pack to get the job done. The board you sold me is rock solid, even running 150 FSB. The performance is absolutely stunning, notwithstanding the small price.

ST4RCUTTER:

'ubiquitous' is the only word I can usually assert without starting a flame shoot from all the VIA-lovers out here. If you prefer another adjective, I think 'sh!tty' fills the void nicely

 

[Outfits]

Posted this morning by angra @ andmb forums:

<< VIA has released a new patch to solve the PCI bus performance bug tecchannel.de found out a month ago. >>

You can read about it here

 

[bassoprofundo]

I'll have to check this latest patch out. Thanks!