Wrote an article yesterday about MSI problems. Wrote an article the day before that decrying certain websites for not knowing about updated BIOSes
This morning, I see an updated article from Anandtech about it. This article does everything possible to convince you
that some improperly soldered components are the sole reason for a reported performance boost.
This can get a little confusing, so let me lay it out.
You have three possible BIOSes here:
You have two board revisions:
One with a component bridging R127, and no component bridging R126. Let’s call that “unfixed.”
One with a component bridging R126, and no component bridging R127. Let’s call that “fixed.”
Therefore, you can have six possible combinations.
A fixed board running Version 1.00
A fixed board running Version 1.09
A fixed board running Version 1.10
An unfixed board running Version 1.00
An unfixed board running Version 1.09
An unfixed board running Version 1.10
What the Anandtech review provided were benchmarks from fixed boards running the various versions of BIOSes. What they did not do is provide benchmarks from the unfixed boards.
Why is this important? If you do both, you get to find out how much of the improvement is due to the BIOS, and how much of
it is due to the “fix.” You would think anybody would want to measure that. They didn’t. Very strange.
Did they have something to say on the issue? Yes, they did.
With a working AMI Flash Utility, we took the board NewEgg supplied us with and flashed the 1.0B9 and later the 1.0B10 BIOSes. Both of these revisions resulted in the same problems we had with the original board MSI sent us, we could not install Windows 2000.
Hmmm. Most of the Anandtech benchmarks use Win98, anyway. Seems simple enough to run a few memory benchmarks in Win98, doesn’t it? But they didn’t. Hmmmm, why?
It was very clear to us at this point that a simple BIOS fix wasn’t going to be the answer to our problems. At the same time, there were reports of users flashing their BIOSes to the 1.0B9 and 1.0B10 releases and not noticing any performance improvement at all.
Oh? One of the proclaimed premium computer hardware testing sites suddenly starts not running tests based solely on what some users say? LOL!!!
Unfortunately for Anandtech, I have been scouring around for results, and have found quite the opposite.
Now what I don’t know in most cases, yet is whether or not those results are coming from “fixed” or “unfixed” boards. Not all cases, however.
If you take a look here (it’s the seventh post in the thread, by WouterKCS), we do get a measurement of the increase
in performance from an “unfixed” board.
Hmmmmm. Pretty big improvement for a BIOS change that isn’t supposed to improve anything, isn’t it? Matter of fact, if you look at the percentage improvment and compare it to the percentage improvement shown in the Anandtech review, the BIOS change
accounts for most of it.
Even more damning is the piece that started it all. Go here. Scroll down to March 13, 2001, to the piece entitled, “MSI K7TPro266 remodelled improves performance.”
The person who originally did this knew how to do this right. He benchmarked with an unmodified board using version 1.00,
then measured with the unmodified board using version 1.09, and only after that he modified the board and measured the modified board using version 1.09.
This is what he came up with in Sandra:
Unfixed version 1.00: 458/525
Unfixed version 1.09: 548/685
Fixed version 1.09: 601/725
Sure looks to me the BIOS does more for the memory score than the soldering fix, doesn’t it?
Of course, that’s only two examples. However, I already have an email from somebody who has verified that he has an “unfixed” board, and he also reports a considerable improvement in memory scores from the latest BIOS.
What Happened Here?
You have three possibilities, neither good:
1) Anandtech did incredibly sloppy, shortsighted work or
2) Somebody or bodies didn’t want to admit being wrong about BIOS improvement.
3) Everything I’ve cited is dead wrong, and so am I.
The first is bad enough. The second is worse. See the next section on what I want to do about door number three.
We have no idea whether or not MSI will recall “unfixed” motherboards. Let’s hope they do, but assume that they don’t.
If you believe that the only way to get your memory speed up to par is to get out a soldering iron, like that article would have you believe, you’re a whole lot more likely to do that than if you know you’ll get most of the improvement without doing so. This is extremely delicate work highly hazardous to the health of your motherboard. It’s not something you present as the only option when there are others available.
If it’s the first, I know a website that needs to get its act together.
If it’s the second, and looking “right” is more important than caring about any possible consequences to the audience, that looks to be irresponsible and a breach of faith.
Somebody has some explaining to do. Which is it?
What Needs To Be Done Now
First, we need to determine how many KT266Pro motherboards have this problem, and how many don’t. Only a couple people have reported to me so far, and they say they have “unfixed” boards. To ensure that possibility number three isn’t true, we also need to confirm beyond a shadow of a doubt that the BIOS upgrade does provide significant improvement.
If you have an MSI K7TPro266, if you could, email me with the following:
1. Where you got it from and when.
2. What version of the board do you have?
3. If you have some software that identifies motherboards, what version does it say you have?
4. Find R126 and R127, (they’re located near the Southbridge). Once you find them, see whether the resistor is on R126, R127, or both, and tell me what you have.
5. If you could, please get SiSoft Sandra memory scores using BIOS 1.00 AND BIOS 1.09 or 1.10 (version 1.09 provides very high memory scores but does appear to be pretty unstable, 1.10 provides lower scores but more stability), and send me the numbers (screen shots would be even better if possible).
6. If you go around forums, and know of people who have this board, please inform them of this article (it’s www.overclockers.com/articles380) and ask them to consider providing this information.
In all likelihood, you don’t want to get out the soldering iron. That probably invalidates your warranty, which would be a bad thing indeed if there eventually is a recall. Also remember that memory performance doesn’t influence overall system performance very much.
After we get a clearer idea as to whom and how many are affected by this, and give MSI a little time to do the right thing, we’ll see where you go from there.