KT333A sooner than expected !

[senior guy]

Looky here. Yeah, I realize that neither the KT333 or KT333A can be fully exploited with the Palomino, nevertheless there are still some useful enhancements (esp. in the KT333A) that may help some of us now - like integrated UCB 2.0 and the faster V-Link, not to mention that the chipset ensures a future-proof mobo!

 

[IntelConvert]

Apparently, VIA's mission is just to 'pump-out' one new chipset after another - faster than anyone else! Too bad they don't put some of their developement efforts into perfecting existing chips - like fixing the PCI hardware problem - which btw, they've acknowledged.

 

[subhuman]

Seriously they REALLY need to fix the PCI problem, and get rid of the bugs. At least they confirm it now, it was frusterating when I was trying to get help solving it and everybody said "it is FINE" when it wasn't...

 

[AA0]

Why do they REALLY need to fix the PCI problem? What problem are you referring to? Do you even know what it is? Its a small performance thing, it isn't a massive bug that makes you run into problems, so there isn't any problem for you to solve, thats why you got no help.

When you look at VIA compared to other vendors they are not bad at all. For IDE performance, they are better than both nvidia and sis, but behind ALi with the last review I've seen. PCI performance they are ahead of nvidia but behind sis, not sure of ALi.

Their chipsets aren't bug filled, get off the bashing bandwagon.

 

[NicColt]

>like fixing the PCI hardware problem - which btw, they've acknowledged.

It appears that some people just like to post for posting sakes and blow hot air.

Even with the so called flaw and according to Anand himself who wrote this article Anandtech's OWN review of AMD's Athlon XP 2000+ vs Intel's 0.13-micron Northwood by Anand which was made on a KT266A based ASUS A7V266-E obviously before the patch, with a 1.67GHz (2000+) cpu compared to an ABIT TH7-II RAID (Intel 850) and a P4 2.2GHz which is 500+ Mhz faster and with the Content Creation Performance (which is real world performance) it could only achieve a TIE

>Seriously they REALLY need to fix the PCI problem

They have fixed it. And according to the Tec channel "now with the Increase in burst transfer rate is now 20 to 30MB/s with all the cards it is tested"

If this is true, then that means that this would now make the KT266A even faster than an P4 i850 rig that's 500+ Mhz faster.

I say again, unless you need to copy or move 20 or 40 Gig files over a raid array repleatedly on a daily basis you won't even notice the problem anyway. Now that Via has an official fix for it and Even if in the real world you don't need it, it will now beat ANY chipset on an equal setting for Content Creation Performance.

>Their chipsets aren't bug filled, get off the bashing bandwagon

For that I have to agree. For all the FUD packers out there, let this die.

>I realize that neither the KT333 or KT333A can be fully exploited with the Palomino

Not even with the upcoming Thoroughbred.

>nevertheless there are still some useful enhancements

I agree with that, and thing is that Via had plans to release the KT333A 'weeks' later not months, so I take it that Mobo manufacturer's were pushing for a release. either way it will just make the KT333A future prood like you said.

 

[Athlon4all]

Interesting info. NicColt I agree totally. There are cases that this "Bug" does actually effect performance, but the cases where it does are rare, and that is good to hear that they've supposedly fixed. I truthfully, am skeptical about what Mike Magee said:

<< The KT-333A, on the other hand, is expected to show significant performance increases, >>

I really doubt that VIA is gonna pull another trick out of their bag like they did with KT266A. Plus without a 166fsb CPU, it'll yield next to no improvement from DDR333 so as far as I'm concerned, KT333A will be made only by it's features, not it's performance.

 

[Killrose]

But I just up-graded to the KT266a. Damn them.

 

[IntelConvert]

Nic~ No bashing intended, just the facts. We all read and draw whatever conclusions we choose to believe...

The problem - as first reported here, with a subsequent followup here.

VIA's recent release of a new patch to improve the burst rate with RAID controllers is here, which is being received with mixed reports as to its benefits as you can read for yourself here.

As for Anand's Athlon XP vs. Pentium 4 comparison (appropriately pointed out by Nic, above), IMHO, it just goes to show how much better a processor the XP really is over the P4 (although the Northwood core does close the gap quite a bit). I can only imagine how much of a distance the XP would have placed between it and the P4 if it had the benefit of a chipset that was as capable of controlling the PCI bus as Intel's i850.

 

[NicColt]

><< The KT-333A, on the other hand, is expected to show significant performance increases, >>
>Plus without a 166fsb CPU, it'll yield next to no improvement from DDR333 so as far as I'm concerned, KT333A will be made only by it's features, not it's performance.

Very well said.

I think when he said "expected to show significant performance increases" it's only if you push the FSB and memory over spec to 166 and DDR333 and don't forget that there still isn't a JEDEC standard for DDR333 memory.

>But I just up-graded to the KT266a. Damn them

don't feel bad at all. if you run a KT266a and a KT333a board at 'spec' performance improvements will be very minimal. The KT333a boards will just be more easier to push the memory and cpu over specs like the iWill333 does now anyway. The KT333A board will not show any real improvements until the Hammer and a DDR333 JEDEC memory standard comes out which is a bit away.

Athlon4all hit it on the head, you will buy this board on features such as USB2.0 and ATA/133 ect. not on performance.

 

[NicColt]

>The problem - as first reported here, with a subsequent followup here.

I know all that and I don't dispute this, but I'll explain it this way.

If someone told me that while driving my Ford Mustang GT at 170Mph it's only really doing 150Mhp, is this reason enough to say that my Mustang is defective and that the entire Ford line is defective and that I won't ever by a Mustang again ? Most people will never ever drive at 170Mph but yet we have FUD Mongers going around saying that the entire maker is defective. When you put things in the entire picture, how often do you drive at 170Mph. and second it only happens in Mustangs, would that prevent you from buying a Ford Taurus.

You have these people city driving in chevettes spreading Fear Uncertainty and Doubt (FUD) over nothing. Let it die.

 

[senior guy]

Getting back to the point of my originating post for this topic - Even if KT333 (or A) doesn't result in any performance gain over KT266A, it brings useful features like integrated UCB 2.0 not to mention a more future-proof mobo than we now have with KT266A. So bring it on!

 

[WetWilly]

I expect the the KT333 is nothing but a remarked KT266A. I've been PMing with someone here on the forums for a while on PCI dividers, and my theory is the KT333 is why KT266A boards don't have explicit PCI dividers of 1/5. You could virtually duplicate the Asus A7V333 by taking the A7V266-E, adding the external VIA USB 2.0 chip used by the A7V333, swapping the VT8233A southbridge for the VT8233 southbridge, and flashing a new BIOS with the 1/5 PCI divider that's already supported by the A7V266-E's clock generator. You don't need a "new" chipset at all. There is a layout change in that the DDR sockets appear to be closer to the CPU socket on both the Asus and Gigabyte KT333 boards. Then again, KT266A boards have been known to hit 166MHz and over without this change. Also kind of interesting that the German reviewers of the A7V333 couldn't run the 166MHz FSB stably.

As for the VIA/PCI issue:

1) Just because some users haven't experienced it doesn't mean the bug doesn't exist.
2) You can't rely on IDE-dependent benchmarks using the KT266A chipset to prove/disprove the existence of the bug. The KT266A's onboard controller uses V-Link and doesn't go through the southbridge's PCI bus controller. It also means the VIA patch does nothing for the KT266A's onboard IDE performance. If you use an external IDE/SCAI/RAID controller, you're using the southbridge's PCI controller and the bug shows back up.
3) The bug isn't completely fixed. VIA's patch (officially called the RAID Performance Patch) is for specific controllers. It doesn't always help if other PCI devices (like high-end sound cards or video capture devices) are having the traffic/burst issues.

BTW, I own and still use a VIA board so I've got a license to be critical of them

 

[senior guy]

<< You could virtually duplicate the Asus A7V333 by taking the A7V266-E, adding the external VIA USB 2.0 chip used by the A7V333, swapping the VT8233A southbridge for the VT8233 southbridge, and flashing a new BIOS with the 1/5 PCI divider that's already supported by the A7V266-E's clock generator. >>

While iInteresting WW, me thinks it's a wee bit easier (and safer) to just buy a KT333 (or better yet, a KT333A) board.

 

[WetWilly]

me thinks it's a wee bit easier (and safer) to just buy a KT333 (or better yet, a KT333A) board

Dang, senior guy, wouldn't you rather roll your own?

 

[Jace]

I'm glad to see these chipsets coming out now too, and I also hope that Via has done some major revisiting on their PCI implementation.

Sorry to take the thread back off topic for a moment, but some people in this forum need to wake up:

The PCI issue that has come up is very real, while it may not affect a lot of users, gamers, office users and hackers, there are MANY people who have invested in systems that SHOULD have been able to work as advertised. For example video editing, many users are finding that because the defects in the Via PCI bus their videos are not editing well, I happen to be one of those users. I may not push the PCI bus with disk I/O like the majority of users are complaining about, but I need to send plenty of data to RAM/CPU. There are even more issues, how about a gigabit ethernet card which could push around 125MBytes/sec, your Raid array would be very ineffective in combination with this configuration? I'm glad that I decided against building a server around a Via motherboard a little while back.

Nicolt, this isn't just a KT266/A issue, this is also a KT133/A issue. It has even been observed on MVP3 and MVP4 systems. Realize that a lot of people who have encountered this problem are really aggravated with a company like Via who decides that they are only going to issue a patch that searches for certain Raid controllers and patches those controllers only, rather than fixing the entire bus, which I realize they might not be in a position to do, due to the design. That doesn't alter the issue. Yes, if someone sells me a product that is supposed to be able to send 133MBytes/sec over a bus, it had darn well better be able to. Call it Fud if you like, it's a nice, easy excuse not to address a serious isse.

As far as whether I'm just spreading FUD, but here are the facts. Every system I build is AMD/Via based and will be AMD/Via based because they do have good price/performance, UNTIL the next time I need to build a power workstation to handle video editing, disk i/o, or network i/o that encroaches on the reduced amount of PCI throughput that a Via based system can provide. Unless I find out that Via has fixed these problems, they simply are not a competitor in that arena. Another reason to be concerned about this issue is what it does to Amd's image. AMD has a wonderful product that has it's image tarnished by any botch ups that Via has since they are very dependent on one another.

This is not an issue that we need to just let "die" because it's ok to by ripped off a little bit.

 

[IntelConvert]

senior guy: ...FWIW, ocworkbench.com has a piece today stating that the Inquirer's report (per your link) is nothing but a rumor and that KT333 board will start showing up within 2 weeks.

 

[meaty]

Enough with the hardware upgrades lets get some software(games imo) that compliments the hardware we already have.
Unless you really need to run q3 @ 400000x240000 lol.

 

[NicColt]

Legion Hardware says

"I will be using the Promise RAID patch, as I will be testing all drives via the Promise FastTrak TX2 ATA100 RAID controller"

"From all my testing so far I have come to the conclusion that there is in fact absolutely nothing wrong with the ?VIA PCI BUS?.

"In terms of stability and reliability I found the VIA chipsets to be some of the best in this department. I was a little disappointed with the nForce PCI RAID performance as it tended to be very inconsistent giving results with massive fluctuations. When testing VIA?s chipsets I found them to be flawless, giving perfect results first time (this was only once the Promise patch was installed). I could almost see NVIDIA having the need for such a patch."

"To cap off this article I think it?s fair to say that all seven chipsets performed very closely. Part 1 of my VIA PCI Bus Investigation has clearly demonstrated that VIA?s PCI Bus is able to perform on par with similar chipsets from Intel, SiS and NVIDIA"

"Burst Rates are where this all started! Supposably the VIA chipsets have very poor burst rate performance from what I have been reading, but hey don?t believe everything you read Anyway from what I found the i850 was again the winner with a burst rate of 93506. Surprisingly (well I wasn?t all that surprised) VIA claimed the second, third and fourth positions over SiS, NVIDIA. Nevertheless all burst rates were very close with only slight fluctuations."

>Yes, if someone sells me a product that is supposed to be able to send 133MBytes/sec over a bus

Other components also HOG the bus such as Video card >more and more< Sound cards, NIC's, name it. I'm sure that if someone did test on those they would find problems with their burst also, and not only on Via chipsets but on all chipsets. Even if you buy a car that does 170Mph it doesn't mean that you can drive it in traffic at 170Mph.

and don't forget that Legion Hardware tested this on a P4 1.7Ghz P4X266, P4X266A, 645, i845, i850 compared to an Athlon XP 1600+ for Via chipsets.

Now how are you gona trust, Tech Channel or Legion Hardware.

 

[IntelConvert]

That's a pretty good find Nic. However, after reading the article there are a few questions on my mind:

First off, besides just the drive with WinXP, why didn't they test 'data drives' on the VIA IDE controller? Seems to me that they limited the extent of the test appreciably by just utilizing the one Promise RAID controller (with Promise's patch).

Secondly, they are very aloof as to specifying the test configurations. None of the Athlon motherboard brands/models were disclosed and some of the other components used were not adequately specified. IMHO, that's shoddy test reporting!

Thirdly, that was only Part 1 of their study, so I for one will wait for Part deux before deciding who and what I'm going to believe.

Cheers...

 

[NicColt]

>'data drives' on the VIA IDE controller?

becasue the 'apparent' flaw that Tech Channel talked about only happens on certain Raid controllers.

>aloof as to specifying the test configurations. None of the Athlon motherboard brands/models were disclosed

It doesn't matter since they were testing the chipset, not the board itself.

 

[WetWilly]

Now how are you gona trust, Tech Channel or Legion Hardware

At this point, I'd trust tecChannel hands down. I could add to what Intel Convert said, but I'll narrow it down to two issues:

1) The smaller problem I have with taking Legion Hardware's article as gospel is that it's being taken out of context here as addressing the entire VIA/PCI issue. I shouldn't be surprised because Legion Hardware themselves state the context then promptly take the article out of that context. The second page of the article mentions that the context of that article is purely PCI (IDE) RAID performance. There are other PCI issues with high-end sound cards. tecChannel tested a high-end sound card (Creamware Luna II) with and without the unofficial latency patch. To be fair, tecChannel didn't test the RPP with the sound card either, but they did illustrate the PCI issue did exist and was resolved in that specific instance with the unofficial patch. Didn't see Legion Hardware do the same; they didn't address high-end sound cards at all.

2) The bigger problem I have is that while Legion Hardware may have provided a plethora of benchmarks, they used only one PCI RAID controller - the Promise FastTrak TX2 ATA100 RAID. tecChannel used three - Promise, Advance, and Acard. Worse yet, they're attempting to bench maximum PCI/IDE bursting and they're only using a ATA/100 card. tecChannel used ATA/133 controllers. Not to mention SCSI as well, which Legion Hardware hasn't touched yet. Sure, Part II's on the way, but I don't see how anything conclusive can be said about VIA's PCI issues from Part I.

3) Here's a quote from the last page of the article:

From all my testing so far I have come to the conclusion that there is in fact absolutely nothing wrong with the ?VIA PCI BUS?. Of course I cannot be 100% sure of this from only these few RAID tests, however they do prove that burst rates aren?t a problem.

They haven't tested ATA/133, they haven't tested SCSI, they haven't tested high-end sound cards, they cannot be 100% sure of this from these few RAID tests, but Part I "prove(s) burst rates aren't a problem?" I don't think so.

 

[IntelConvert]

Very well put WetWilly. I would only add re: Nic's remark "It doesn't matter since they were testing the chipset, not the board itself", IMO full disclosure is everything in any impartial/objective test - i.e., what else didn't they disclose??? It should be up to the targeted audience/reader as to what is relevant and what is not!

 

[WetWilly]

why didn't they test 'data drives' on the VIA IDE controller?

Like I'd posted a while back, the problem doesn't exist with IDE controllers on V-Link southbridges, so the patch does zero for the onboard IDE controller's performance.

It doesn't matter since they were testing the chipset, not the board itself.

I couldn't disagree more. Considering that nForce is finally getting polished, I'd say it's significant who manufactured the nForce mobo and what revision of the BIOS the board has. I wouldn't take ANY nForce board with a figurative 1.0 BIOS as the barometer of nForce performance. Likewise, what version of the nForce system drivers were being used? The original release version? Pre-release? Latest version? I suppose you could make assumptions, but it's just another beloved patriot in the aura of authority this article is trying to generate for itself.

IntelConvert

It should be up to the targeted audience/reader as to what is relevant and what is not!

Exactly. It's one of the many shortcomings of web "journalism" and I use that term rather loosely. It's rather easy to lower your standards/loosen your filters and run through web articles without fully absorbing their point, valid or invalid. Let's face it - some "journalism/objective review" is nothing but filler so there's something between JPEGs of benchmark results.

If this article wasn't trying/being used to make a specific point, I would have scanned it, noted they failed to mention the equipment they used which means the results have rather limited usefulness, noted it was a expanded yet limited subset of the tecChannel article, probably ignored any mention of Part 2, and moved on to the next thing. Using it to not only prove a point, but to discount a lot of users' valid complaints means it's time to look a bit more critically, and the article (at least Part 1) doesn't stand up very well at all under scrutiny.