Well look out, I had problem working with cubase and VSTi, Nuendo.... I done everything to make it work and it always become shit! At last I found what the real problem was..............

two points:

1. You people talking about intel chipsets being more stable are either

a. lying

b. lying and never used an intel chipset platform before

c. stupid

2. if you dump the via ide driver & go to the standard dual PCI IDE controller provided by *any* windows version, including XP, your disk scores will improve 50%. I've been doing this since MVP3 all the way to 266a with never a problem, only a score increase. Note however, that a friend of mine did this on an AbitKT7a (133a) board and started having HD corruption problems. YMMV

I've been keeping an eye on this issue as reported on HardOCP and it seems that an interesting side effect of the 'patch' is that 3d performance suffers. It seems possible that VIA is not only completely aware of this issue, but actually likes it or did it on purpose. If giving higher PCI priority to the video card results in higher gaming benchmark scores on review sites and lower general PCI bandwidth (something hard enough to find that it took the whole community years to figure out), that's a trade VIA--and almost any other company fighting tooth and nail for an edge over the opposition--would be happy to make.

I'm so sick of VIA lies.

[q]Do these PCI issues only come with overclocking?[/q]

No. In fact, overclocking would likely present a work around for the issue. Why? Because as you increase the FSB you raise the PCI bus speed as well. Raising the PCI bus speed increases the theoretical maximum and thus should open up the VIA chipet to outputting more data. Albeit still less than a competitors chipset clocked at the same speed.

This of course assumes that whatever is wrong with VIA's implementation doesn't cause some sort of wall to be errected. What I mean by that is that the bandwidth limit ends up being an absolute due to the error versus being just a overal lower output value. I don't think the error behaves in this way, but I can't say for sure either.

#52 I am looking to by a Soyo Dragon + with Athlon 1900+
for a Digital Audio Workstation. I am also planing to use the built in RAID 0.

Your comments and this thread make me second guess my plan. Are you overclocking your system? Do these PCI issues only come with overclocking?

[q]I was reading this article and the comments, and pondering this Via patch this guy whipped up. I was intrigued, so I slapped it on. Over the next three days, my formerly perfect KT7A system devolved into insta-death hell. Random lockups, with or without cause. Got one just now as I was writing the first draft of this comment.

My vote is for the Via patch to suffer and die. YMMV.[/q]

http://bbs.pcstats.com/viahardware/messageview.cfm?catid=19&threadi...

Yes, in the VIA Hardware.com forums that discuss the patch it seems that some SBLive users (particular 5.1 users) are having difficulty with the patch.

HardOCP has also reported one difficulty from users that have talked to them. A uninstall worked for that individual though.

Nothing's perfect.

George has purchased a SBLive! 5.1 to find out what's going on.

This sounds like prime BS, but I swear it's all true.

I was reading this article and the comments, and pondering this Via patch this guy whipped up. I was intrigued, so I slapped it on. Over the next three days, my formerly perfect KT7A system devolved into insta-death hell. Random lockups, with or without cause. Got one just now as I was writing the first draft of this comment.

My vote is for the Via patch to suffer and die. YMMV.

P.S. - Uninstalling the damned thing didn't fix my machine, it was doing it still, right up to the System Restore I just did. Good thing I built a point before installing that junk.

[q]Not that it is a great loss, but doesn't that remind you of Intel's i820 fiasco two years ago?[/q]

Not really. The third slot on the i820 just didn't work period. All three slots on the nForce work; just one requires single sided memory.

And instead of not posting or constantly crashing if you do not adhere to the instructions in your manual, it simply clocks down to insure stability.

I personally don't see why people have a problem with it. It's almost like the problem resolves around a issue of control. Since people can't disable the save your ass feature of the chipset, they get upset.

I can't at the moment think of anything remotely similar to the above. The closest thing that comes to mind is the KG7-RAID and it's limitations on buffered versus unbuffered memory and load. Of course it will just plain crash if you over do it.

Monaco:

Not to be rude or mean, but it's quite obvious you don't what you are talking about.

The fact you think the third slot of the nForce is "dead" alone shows that you haven't been paying much attention.

Originally Posted by Monaco
Ryu I was sure you were joking about nForce. Now it seems that you are not. All I can say is...????

nForce still looks like a joke to me. It\'s designed for OEMs, not us. The memory slot problems and the worthless integrated video seem to prove that point. The theoretical bandwidth is pretty awesome....but it\'s still just theoretical bandwidth. I haven\'t seen any tests that were too impressive- mostly due to that hog of a video card that is built in. Use it, and your scores suck. Disable it, and it\'s still sitting right in the middle of your bandwidth.

I\'m sure that some day relatively soon, nForce will kick serious ass- but not yet.

Also- I believe the dead 3rd slot on nForce boards falls into the same category of dumbass errors as VIA\'s PCI bus problems. An error that the company says is a \"feature\". Both VIA and nVidia deserve a public whipping for shit like that.

Originally Posted by ronald
Still, if they are right, nForce is only limited to two memory slots for optimal performance. Not that it is a great loss, but doesn\'t that remind you of Intel\'s i820 fiasco two years ago?

[q]What about these alleged memory slot problems with the nForce?[/q]

I think the referenced link:

http://www.amdworld.co.uk/nfmemory.htm

pretty much sums up my feeling on the subject matter of "SuperStability" mode and more.

[q][i]The Via KT266A platform offered better tweaking and could be taken further by the use of the FSB adjustment peaking above what we could achieve with the NVIDIAŽ Nforce platform. But has the results reveal to the VIA KT266A had to be pushed harder above its manufactured state to achieve a competitive performance.[/q][/i]

[q][i]Now has we have already stated the NVIDIAŽ Nforce is still a young design and has a great platform to build on and our opinion for what its worth is stick with the NVIDIAŽ Nforce platform until proven otherwise .The MSI K7N420 Pro NVIDIAŽ Nforce motherboard has been thoroughly tested since October 2001 here and with in that period two other KT266A motherboards have passed through for testing and each time so far the K7N420 Pro has proved to a better solutio/q][/i]

[q][i]The NVIDIAŽ Nforce is what we would call the BMW of motherboards wether you like that phrase or not is for you to decide but after many hours of using the NVIDIAŽ Nforce (MSI K7N420 Pro) and still waiting for hint of unstabilty to arise we would gladly argue that memory performance is the best at present in any AMD platform and until proven different we will beg to differ and we fully expect to see more performance with the release of DDR333 which once again we hope to show you very shortly.[/q][/i]

It simply doesn't get much prettier in any other part of the review for the KT266A. They are even choosing nForce over the vaunted Epox 8KHA+ and Abit KR7-RAID.

Originally Posted by ronald
What about these alleged memory slot problems with the nForce?

http://www.gamepc.com/reviews/hardware_review.asp?review=nforceddr&...

http://www.amdworld.co.uk/nfmemory.htm

[qForce has compatability problem/q]

Got link?

[qackluster performanc/q]

Not sure what review you've been reading. It's pretty much identical if not faster than the KT266A in terms of performance.

[q]and horrid overbussing potentia/q]

As opposed to a overclocked KT266A that performs slower than a stock nForce?

http://www.aceshardware.com/read.jsp?id=45000232

[q]I'll take an Epox 8KHA+ or an Asus A7M266-D, thanks much.[/q]

Your loss on the former, your win on the latter.

#59 pompus #57 right

nForce has compatability problems, lackluster performance, and horrid overbussing potential. I'll take an Epox 8KHA+ or an Asus A7M266-D, thanks much.

[q]Take your pick. If you are really a tech freak like the rest of us, then your decision is already made.[/q]

Right! nForce all the way.

;)

[q]Ryu - I'd suspect it's more that they don't appreciate him complicating the 4-in-1 team's job. What if the new 4.38 4-in-1s come out tomorrow and rock on all systems, except those who had this guy's 'fix' applied previously? How much do you think it'd cost them to run down the problem? I'd bet it would be pretty costly.[/q]

It is a possibility, but I find it a unlikely reality. His tweaks make a positive improvement. The worst case scenario to the conditions of his changes is add-in device that fails to cope well with the adjustment he has chosen(which seems unlikely since he manages to make add-in devices work better; check the FAQ). In most cases the register settings he is manipulating either need to be at his choices or at some point will/could be.

It seems to me he is unliked because he represents a press relations problem when he manages to fix problems for them. It has the side effect of making VIA look inept.

"Well, why couldn't VIA do this?"

In reality it might be a mix of A & B, but we do have to remember this is the press relations department who has the attitude with him. Not the driver writers.

Check out the rest of his FAQ though. Interesting reading.

hey Intellimole, about your post #39...

I see what you are saying. You also make my point for me. VIA does have the fastest hardware. Period. You've read the reviews as much as I have, I'm sure. 1900+ XP and DDR kicks ass, and that's that.

Unfortunately that ass-kicking comes at a price- you have to baby it, and watch out for weirdness or malfs. All a part of being on the bleeding edge. You don't expect a top fuel dragster to run flawless quarters, every time, do you? Nope- every once in a while it's gonna crash and burn. That's the price you pay for going all out.

Intel/SiS is slower and very stable, VIA is faster and not so stable. Take your pick. If you are really a tech freak like the rest of us, then your decision is already made.

Ryu - I'd suspect it's more that they don't appreciate him complicating the 4-in-1 team's job. What if the new 4.38 4-in-1s come out tomorrow and rock on all systems, except those who had this guy's 'fix' applied previously? How much do you think it'd cost them to run down the problem? I'd bet it would be pretty costly.

http://www.networking.tzo.com/net/software/readme/faqvl019.htm

George Breese (the maker of the referenced tweak):

[q]
[b]7.4 What is your involvement with VIA?[/b]

I have no involvement with VIA Technologies Incorporated. As far as I know, only three members of the VIA marketing team know that I exist, and two of those have been unfriendly in past discussions.
[/q]

I guess they don't like him saving their bacon.

[q]Thats where most of these posts stop. They dont tell you all the shit they went through to get these computers to that point.[/q] Okay, I built two VIA/AMD systems this year, and I'll tell you exactly what I did and what problems I had:

MSI K7T-Turbo-R mobo with Duron 850, 512MB Crucial RAM (2 DIMMS), twin Seagate Baracuda III HD's in RAID 1, Voodoo V5 (which I happened to have lying around), Panasonic 8x4x32 CD-RW, Sony 52X CD, Intel 10/100 NIC, generic floppy, USR 56K modem, in an Antec SX-830 case, using onboard sound. I put the graphics card, one sitck of memory and the CD on the 'board and fired it up; success. I then loaded W2K, added the VIA 4-n-1's (but for the IDE driver, which was Microsoft's); success again. I did a 48-hour burnin with multiple apps at default BIOS settings, then added the rest of the hardware and started pushing the OC potential. When I reached the limit, I repeated the burnin. At no time did i have a problem requiring "months of hell and testing and switching devices and troubleshooting" -- the system worked fine the first time through, and OCing was the only trial-and-error process (exactly as it should be).

The users (friends of mine) ran this box for six weeks without rebooting once; then a lightning strike fried the mobo, CPU, modem and graphics card. I replaced them with another copy of the mobo, modem, an Athlon T-bird 1.33GHz, and a Hercules GeForce MX, re-using the other components (which I thoroughly tested) and letting it run at stock speed (I'll treat 'em to an OC speedup later, when they can appreciate it). Everything worked fine, first time through. I added W2K SP2 (which had released in the meantime), and my friends proceeded to again run the system endlessly with almost no problems -- the [inl/i] problem being the crappy bundleware from the CD-RW (Roxio), which I replaced with Nero (which fixed the problem completely). There have been no other problems.

To me, that doesn't count as "having problems" -- it counts as a typical build, and it was one of the easiest I've had. My secret: I chose good components, I read the manuals, and I followed the recommended build sequence.

Now you tell me where the problems are...

[q]why hasn't this bandwidth limiatation shown up in benchmarks of via-based motherboards? Either for the kt266a or the p4x266? When reviewers are posting hard-drive benchmarking scores, or testing the raid setups, shouldn't this bandwidth problem affect those scores?[/q]

tecChannel's benchmark test were capturing the burst speed of a 5400RPM ATA133 hard drive. To conduct these tests they used a Promise ATA133 controller on the PCI bus.

You should be able to see it in other places. In fact:

http://www.storagereview.com/jive/sr/thread.jsp?forum=1&thread=21269

Storage Review forum users talk about poor STR (sustained transfer rates) in RAID 0 from their VIA boards.

So, I think perhaps the reason reviews didn't see is due to ignorance more than anything else. We assumed the scores received were the best available and that appears not to be the case.

[q]Or do the run-of-the-mill benchmarks of disk subsystems not stress the bandwidth of the pci bus as much as video editing would? Or is this an issue where on-board raid or ide controllers do not map to the pci bus?[/q]

Starting with Apollo Pro 266 VIA implemented a hub architecture this means the VIA IDE controllers are no longer mapped to the PCI bus. You can tell you have a hub architecture by seeing if your southbridge is a 8233 or 8233A. These controllers will NOT see the effects of this PCI bug. The 686A/B, and the older 596A/B on the other hand will see the effect.

http://www.tech-report.com/reviews/2001q4/p4-chipsets/index.x?pg=2

High-Point contollers either on-board or on a PCI card, Promise controllers on a PCI card or on-board, CMD Technology controllers, or even 3Ware controllers, you will feel the effect of the PCI bug. And if you are using any of the afforementioned adapters chances are you are using RAID as well.

Video editing just compounds issues because you need high STR and you usually also have a PCI card that is doing crunching through the data to complete a task you have assigned.

http://www.pinnaclesys.com/ListProducts.asp?Langue_ID=7

Here's a great example of video editing at an extreme: HDTV feed:

http://www.storagereview.com/jive/sr/thread.jsp?forum=1&thread=20948

Well Spoken RYU... well spoken

There is no doubt that VIA has got/had/probably-will-keep-having problems. I think this is a very serieus problem that I can only hope VIA address', I've continually purchased and used VIA based motherboards over the past couple of years, and there has not been one board that I did not have to "do something extra" to just to make it work the way it should have worked when I purchased it. This fact is very wrong, and does not bode well for VIA's track record (in my opinion of course).

You see so many post "I've got such and such and it works great, you guys must suck and not know what you doing if yours is messing up"... thats just simply not true, there are to many damn combinations of hardware to say such an ignorant thing.

For instance, right now I've got two IWILL KK266 based computers up and running with an uptime going on 2-3 months running FTP, HTTP, remote desktop and many other varies servers and services with alot of game playing, CD burning, filesharing programs, IRC, and both running Genome@home (like Seti) and there overclocked as far as they will go. These computers have been working great and I'm 100% satisfied with there performance and stability. Thats where most of these posts stop. They dont tell you all the shit they went through to get these computers to that point. I went there months of hell and testing and switching devices and troubleshooting to get them to that point. Doing all that "Tech stuff" happens to be my bag so its not that bad for me. But most of the problems were so rediculesly stupid, forcing me to scour the internet looking for incompatibilities, pathes, tricks and the like all to make it work like it should have a day or two after I pulled it out of the box.

Anyway, the point is not "Can you have a good computer that works fast and stable based on a VIA chipset?"... oh no my friends, the real question is "Why is VIA hiding its problems, not dealing with there problems ASAP and forcing users to go out of there way to make there product perform correctly?"

Ryu -
That was as lucid an explanation of where you're coming from, and why you think this issue matters, as one could have expected. I have only one question (probably due to my lack of tech know-how) -- why hasn't this bandwidth limiatation shown up in benchmarks of via-based motherboards? Either for the kt266a or the p4x266? When reviewers are posting hard-drive benchmarking scores, or testing the raid setups, shouldn't this bandwidth problem affect those scores? Or do the run-of-the-mill benchmarks of disk subsystems not stress the bandwidth of the pci bus as much as video editing would? Or is this an issue where on-board raid or ide controllers do not map to the pci bus?

Let's get a few things straight here.

1. I don't like VIA. I'm not some chump who bought one board and got burned. I've owned ten of their boards starting with the MVP3 and working all the way up to the KT133A. Two: I take care of over 100 KX133 systems at my job. I know VIA systems very well and I know I don't like them and I would put money where my mouth is to say that if you worked with me, you would have an equal distaste for them.

You don't have to like my opinion. I provide the story and the facts. You are more than welcome to reach your own conclusion. I must assume though that if you come here to TR to read our news stories you are also after our commentary and opinions. You've got them now and now comments are your battlegrounds for dissent or agreeance.

2. I am not anti-AMD and never have been. The K6-2 and K6-3 series kicked more ass than Dennis Hopper hopped up on smack. And if AMD had bothered to build a competitive chipset around them I feel they would have proven very dangerous competitors to the PII and PIII. My current system is a nForce board an a Athlon XP 1800+ and quite frankly this MSI nForce board has proven a worthy competitor (in terms of stability, as it simply blows away my old system in performance) to the i815EP Black Pearl + PIII 1GHz board I retired into my secondary system.

3. I didn't write this article to bag on VIA for being screw-ups. It's a hopeless and futile excercise to try and say they are the lone (&()*-ups of the industry. As many have noted the i850MCH has difficulty with bus master cards performing certain operations thus effectively limiting it's bandwidth. The 760MPX seems to have a USB 1.1 bug, nForce has it SuperStability function (if you call that bug), and we haven't heard anything bad about ALi or SiS; yet.

My point and my desire was made clear in my final paragraph. I don't feel VIA ever gives it's user the straight story. It is **MY** perception that they lie about the facts of their difficulties and then attempt to use their press relations to conviently re-write history after the event passes into obscurity.

I want them to do like Intel and release specification sheets. If Intel has the balls to detail every bug their chipsets have then I fail to see why VIA, NVIDIA, SiS, etc cannot do the same. Those specification sheets are also a great way to disprove the myth that sometimes crops up around a errata, which is why I mentioned the release of VIA specification sheets would be a great way to make me eat crow.

If you want an example of how bugs get blown into myth, just look at how many different stories about the i850 bug were told in this thread alone.

VIA making their specifcation sheets available removes the apperance of impropriety. And that is what I want, I want to quit wondering whether or not VIA is lying. The fact that they continue to hide their difficulties only makes me more supicious of what problems of performance or hardware the community has not yet discovered.

4. Some people asked, "Well, what gives who will this PCI performance issue effect?"

Potentially everyone. Take a look at those bandwidth scores. Your ATA 100 hard drives in burst, your two RAID-0 ATA-66 hard drives in burst, even sustained transfer rates in RAID, high end video or music digital editing, etc, etc.

And to that effect I think #52 illustrates things quite clearly. There are quite a few things that heavily tax the PCI bus and this issue limits those individual from completing either their hobby or their job in a timely or reasonable fashion.

These are the kind of difficulties that create FUD in certain industries and the last thing AMD or it's partners need is more FUD. I've had e-mails in the past about video editing companies saying that AMD is a poor choice. Well, given what we know about the above VIA chipsets, is it any damn wonder those companies give that tag line?

It's time to get brutally honest and it's time for VIA to start setting their sights higher than, "good enough."

#43, my mistake, that comment was intended for #40 :)

http://www.theinquirer.net/21120105.htm

Here is the problem-I bought a Soyo Dragon + (raid ) and I can't use it for the applications I bought it for. I wanted to use the raid 0 to do high speed video editing and playback. It won't do it. When you playback a video off the raid array, it pauses and "hiccups". If you copy the same file to the IDE 0 drive, it plays perfectly. You don't have to be doing anything very "high-end" for this to matter. It sucks because I spent a lot of money on a board I can't use for my applications.

AG #46 here.

http://developer.intel.com/design/chipsets/specupdt/29071501.pdf

Page 12, errata 5.

" No fix planned "

Havent looked at the i850 one (links got a bit wrong BTW) but i think youll find it the same.

Seems like we are ALL Via 'beta testers'