Via Chipset's Flaw

[Bozo]

There seems to be a problem when using VIA chipsets relating to Hard Dsik performance.

Full Story (http://213.40.196.64/content/3/23502.html)

[NorthernYankee]

Another reason to get the XP333

--NY

[mdzcpa]

I've been working on this problem for a while now. I noticed that my ATTO benchmarks on my KR7A-RAID with a RAID 0 configuration, although much faster than a single harddrive, were still not where they should be.

My orginal thread discussing poor RAID 0 performance and my theory that it was the VIA chipset was posted on the Icronic Forums a few weeks ago. The link to that thread is here (http://www.icronticforums.com/showthread.php?threadid=17913) . You can also reference my ATTO scores on that thread with and wihout the "unofficial" VIA Latency Patch, and the ATTO scores for my KG7-RAID which did not seem effected by the PCI problem.

Although this issue certainly is not good for VIA chipsets, it is also not as severe as the TecChannel article seems to indicate. From an everyday, seat-of -the-pants point of view, the problem is not even apparent. My ATA 100 RAID 0 HD benches are still twice as fast as a single ATA HD benchmark. And without the use of ATTO, I could not even prove if there was an issue at all. I did notice a few dropped frames during video edit as compared to the KG7, but it was so immaterial I didn't realy pay any attention to it.

Don't get me wrong. From a competitive point of view, the VIA chipsets need to keep up with it's competition in mass storage read/write capabilities. And this will be a real issue for those who must have the fastest mass storage speeds. But to the average desktop user, or even PC enthusiast, this issue is rather moot. It is widely known that RAID 0 hardly makes a difference in day to day desktop PC use anyway.

I hope VIA works it out though. I like competition and choices in the market, and VIA has put a lot of pressure on Intel, SIS, and Acer Labs to improve thier products for a better price.

[Bozo]

I must admit i didn't really look at the article in too much depth, I thought that the kind of people who are trying to change voltages to overclock their pc and other people in these forums who certainly seem to dig around with the inner workings of their pc more than me, may find this of interest..A bit over my head to be honest with you;)

[Bindibadgi]

So is that just with the onboard IDE/RAID channels or if u use a PCI controller card? or both? :eek:

[mdzcpa]

Bozo,

The link was good info and your post definitely worthwhile! I hope my response did not sound like an "old news" comment. I was just trying to lend my own personal experience to the thread in hopes of bringing a greater understanding to the topic. And I wanted to lend a little "day to day" perspective to the problem.

Quite frankly, I'm glad the "issue" is making it to the forefront of PC enthusiast news. I was sort of poo-poo'd by a few forums when I first brought this issue up a few weeks back. I was basically told that my problem was in my RAID cluster size (and that apparently i did not know what i was doing setting uo the RAID)...I did not beleive that for a minute.

Now that this has come to light more, I feel a bit vindicated.

Bindibadgi,

Both.

Whether you use an on-board solution, or an add-on card, the data from these devices must travel through the PCI bus (just like regular IDE channels). Due to the latency characteristics of the VIA PCI Bus, large volumes of data are slowed by Bus interuptions.

To be honest, I beleive this PCI Bus latency "issue" to be more of a chipset characteristic versus a bug. It's just a limitation of the chipset. There are no stability issues. And, it will effect only those with RAID or SCSI set ups who are looking for the absolute fasted speeds. Again, day to day desktop users will not even notice. I suppose with ATA 133 drives (versus my ATA 100 drives) the losses would be more material.

All in all, it is definitely something that should be addressed by VIA because other chipsets do have better peak mass storage read/write speeds. VIA should not bottleneck the system compared to other chipsets without expecting some bad press over it.

By the same token, I felt the articles at TecChannel "felt" a little anti-VIA. They made the issue seem pretty severe inlight of the low number of users who will likely really see diminished day to day performance. VIA has been smacking Intel around a little with recent legal victories and their P4 DDR chipset, and I wouldn't be surprised if someone at Intel jumped all over this chance. Just a thought.

[Bozo]

Originally posted by mdzcpa
[B]Bozo,

The link was good info and your post definitely worthwhile! I hope my response did not sound like an "old news" comment. I was just trying to lend my own personal experience to the thread in hopes of bringing a greater understanding to the topic. And I wanted to lend a little "day to day" perspective to the problem.

[B]

didn't sound like an old news comment at all. just pointing out I would let you guys who get down to the 'nitty gritty' side of things check it out.. then if there was major problems i would sit up and take notice:)

All I want is my pc to boot up properly and to work as i expect. I am not bothered by a little under-performance but i know others want the maximum possible.

[The Therion]

also check this out ---> <a href="http://www.tecchannel.de/hardware/817/1.html">benchmarks of various motherboards against VIA KT266A</a>

[The Therion]

and here's a patch for it --> PCI Latency Patch (http://www.tecchannel.de/hardware/817/7.html)

[GITster]

I'm quite happy with mine :D

[GITster]

This is off my 'system 2':

[minty_altoid]

21k from a single 75GXP is pretty sweet....

[mdzcpa]

Yep. That is a good bench for a single drive.

The VIA PCI Bus "issue" will only effect RAID or SCSI users as only these devices are capable of producing enough data stream to bottleneck the bus.

This is why I say very few users will even be effected by this issue. Most desktop users do not use RAID or SCSI. And if they do (like I do), RAID and SCSI make very little difference in day to day performance anyway, so the "issue" isn't even apparent. With RAID or SCSI loading the OS and a few games is just a tiny bit faster, but websurfing, downloading, email, gameplay, etc, is all uneffected..

One of the few things a typical desktop user may do which RAID or SCSI does improve is video edit. Because I do some video edit (just for fun), I like the RAID set up I have. But even doing this, i could hardly tell the PCI Bus "issue" was there. The RAID is still very quick (much faster than a single drive). I just noticed that it looked like a was dropping a few frames during edit as compared to the KG7 RAID I had been using just a few weeks earlier. So I looked into the matter, and learned of this VIA limitation. Even so, the RAID on the KR7A does just fine for me anyway.

I suppose the conclusion of the original article may be valid. "Professional users" requiring the absolute fastest RAID and SCSI has to offer may wnt to avoid the KT266A. But, IMOHO, for everyone else (99.9%) the "issue" will not efffect them att all.

BTW, I just benched my new ATA 133 HD on both my KG7 and KR7A and they were about the same. So single ATA 133 users have nothing to be concerned about. Single drives, ATA 133 or not, just do not produce the data stream required to bottleneck the VIA bus.

I'm sure VIA has known all about this. They probably figured that if id did not effect the vast majority of users it would not be a porblem. But now that the Internet Press is onto the issue, they have no choice but to respond.

[The Therion]

mike,
will this problem cause frames to drop when video-capturing? (with ok cpu speed)

[AFroDYme]

i could because
video streaming lays a continues load on every part of the system

and if your pci-bus can't handle the stream form your capture card to the hd's (it all has to pass over the pci-bus )


and we all know that even compressed video takes some serious bandwith