I’m looking at these numbers, and in one way, they make sense, in others, they don’t at all.
With one exception, the scores I got were from “unfixed” boards, which MSI apparently is saying is the shipping
version. Certainly, the vast majority of the boards have the resistor at R127.
It’s very clear that going from 1.00 to 1.09 BIOS jumps memory scores up significantly (though not consistently) on these boards. Just as clear 1.10 brings them most of the way back down again. A few representative samples:
CPU Speed: 1200Mhz
Comments: All using CL=2, 4-way Interleaving, 1nt command
CPU Speed: 950Mhz
Comments: Memory running CL=2.5, speed SPD, 4-way interleave (BIOS reports 266MHz at startup)
You can see the BIOS all by itself made a big difference going from 1.00 to 1.09. If any of these people swapped resistors, I’d bet the numbers wouldn’t jump up all that much, maybe 7-10%.
On the other hand, so does a resistor at R126. The one score we got from someone with a resistor there was:
CPU Speed: 1267Mhz
Obviously a big jumpup at 133Mhz from the results shown by the person with the resistor at R127. I would bet if the person went to a newer BIOS, this time around it would be the BIOS that wouldn’t increase the numbers too much.
So saying that one of the other does everything, or even most all the time, seems quite incorrect. Both have a significant impact, and for one person taking these two actions, the one that will look most important to you will be the one you do first.
Since it seems like the resistor at R127 and BIOS 1.10 is the current MSI “standard”, one should consider the platform to be that for any comparisons.
Per memory scores:
I’m not seeing a whole lot of consistency in the degree to which these numbers vary. The only consistent trend is
that increasing FSB has a disproportionate effect on memory scores. Running your memory faster is the
best way to improve your memory scores.
Some of this is no doubt due to different memory settings, which brings us to an important point. If memory is to be considered a key benchmark, (and for better or worse I think it’s going to be for at least the first round of DDR testing), a review site should state precisely what settings were used to achieve that benchmark.
This can be more of a problem than you think. The same settings won’t work for everybody. For instance, many have found
that under BIOS version 1.09, using the “1T command” rendered the system unstable. However, there’s a few out there running happy as a clam with it, and using “1T” seems to be fine under version 1.10.
Actually, just about any problem a few or more people have reported has been no problem for at least some.
Some have reported problems installing Win2K, some have not. Installation problems or boot problems may be caused by the placement of PCI cards. However, these are not the “normal problems” encountered with other Via boards. A few have reported that certain cards will only work in slot 1, and if they’re anyplace else, you go nowhere.
Some are having nightmares with an SBLive card which go away upon removal, some aren’t.
Some are reporting that the Promise controller doesn’t work with BIOS version 1.09 and 1.10. I’ll need to look more into that, but if generally true, that’s obviously a big problem.
Do you know what this board reminds me of? It reminds me of the Abit KT7 when it first came out, which isn’t exactly a recommendation. It either worked or didn’t. After a lot of effort, solutions were found for most situations, but even then, too many of them just didn’t work right.
Why One Is Not Enough
This whole MSI situation shows why the old review process doesn’t work any more. I’m not talking about other people or places, the process doesn’t work.
If you have just one board, does the board really have problems installing Win2K in general, or is it just that particular board? If you only have one board, how could you possibly tell which is the case?
You got a new BIOS and a new place for the resistor. If you try changing the BIOS first, that looks like the biggest improvement. If you try changing the resistor first, that looks like the biggest improvement. Unless you’re a soldering slut, you’re probably not going to try it both ways with one board.
If you don’t happen to use the Promise controller, how are you going to know it’s broken? If it’s broken using a certain type of hard drive, how are you going to know if you don’t use that kind of hard drive? What if this isn’t a 100% situation, but a particular item works on 85% of the boards, but not the other 15%?
Reviewing a particular item is easy. You do the installation, run a few programs, and write down the results.
Reviewing how a product is doing “out in the wild” is awful. It’s literally a jungle out there. All different kinds of equipment, all different skills levels, all different kinds of results, out of which you have to make some sort of sense. It’s a much harder, much more time-consuming affair.
It’s also the one that’s needed.
How many times have you bought a product and ran into a problem not mentioned in any review on the subject? I bet you have at least once. I know I have. I know some of you have run into that problem based on our own reviews. And I know that we didn’t say anything about it because either we didn’t have the problem, or didn’t test a particular combination of equipment.
Or the opposite can happen; we get a product we don’t like for very good reason, and you end up buying it, and you don’t have our problems at all.
Neither situation helps you very much, does it?
I don’t know if this situation bothers anybody else, but it sure bothers us.
A New Way To Review
We don’t believe just taking a piece of equipment like a motherboard and testing it is good enough anymore. That doesn’t mean we won’t take motherboards and test them; we just won’t base the review solely on our findings. We’ll look around and
see what others are running into and incorporate that into the review. We suspect we’ll talk more about what what we find elsewhere than what we find ourselves.
This also opens up another possibility.
May I suggest to you that one can “review” a product one doesn’t even have?
Before you reject that notion out of hand, just what do you think the CPU Database is? It’s a virtual review mechanism for CPUs. We certainly haven’t had every single type of CPU listed there, but we can certainly look at the data and give you a pretty good idea about how well a type of processor we’ve never seen, heard, touched or tasted before can overclock.
In a number of areas, it would be better to have the equipment, I’m not disagreeing with that. However, what’s better, basing your opinion about a product based upon an untypical board, or basing it upon a wide variety of opinions about a particular product. Which is more likely to find more of the potential problem areas?
I’ve had a few people tell me essentially, “you can’t say anything about the MSI board, you don’t have one.” Well, having just one and basing everything on just one can be worse than not having one at all. If half the boards out there work, and half don’t at some task, how can one board work and not work at the same time. That’s what it would have to do, wouldn’t it.
Far better to look at a range than just one.
We’re going to test this idea out over the next week or so. We’re going to scour around finding out what we can about the MSI board without actually having one, and see what we come up with. Maybe it will work, maybe it won’t, but when you have something’s that broken, at least try to fix it.