Forced Certitude

Add Your Comments

First, let me repeat what I said a few days ago.

We follow the principle of prudence. We know that some people base their buying decisions on what we say. We also know that people often are concerned with whether or not anything they buy today can be upgraded tomorrow.

So we try our best to try to figure that out, and if we have any doubt that a product will be able to do something, now, or in the future, we’re going to express that doubt. . . .

If you buy something, and six months later you’re crashing left and right and then you find out that you need a new mobo, you wouldn’t be too happy with someone who said that it “could” work, would you?

Then again, it may work fine. Right now, we don’t know. When you don’t know, it is more prudent to go with the answer that you know works than the answer you’re not too sure about.

The TweakMax Article

There are times you write something not to provide an answer, but to provoke one.

This is what I actually said initially.

The Asus motherboards, on the other hand, use an older generation of voltage regulator; the CS5322. You can see the datasheet for it here. If you’d like to comparison between the 5303 and 5322, look here. It can get down to 1.1V, too, but while I’m not an electrical engineer, it doesn’t seem to be as advanced as the other chips. Not even going to pretend I understand 95% of the datasheets, but if you look at Figure 1 in both datasheets, the one for the 5303 refers to 60A, while the one for the 5322 refers to 35A.

The 5322 may well be good enough to handle a Palomino, but if I had doubts about any of these voltage regulators handling a Palomino, it would be this one, and I’d get something definitive from Asus before buying one.

I didn’t say it wouldn’t work; I didn’t say it couldn’t work. I just had some doubts about it, and suggested prospective buyers seek verification from Asus on this subject.

Well, the author of the TweakMax article apparently doesn’t think you should do that, and wrote an article to “prove” the A7V133 was Palomino-capable.

Now if he had written an article that proved the A7V133 was Palomino-capable, we would have put that on the front page a long time ago, and I would be happy that someone did.

Unfortunately, he didn’t.

“Listen to me, I Work in the Field”

It has been my experience that when somebody starts off making a vague authority reference (“I work in the IT industry” is one of my favorite), watch out.

People who really know what they’re doing show it in their explanation, or say specifically why they should be an authority (“I engineer MOSFETs”), and then they back it up with hard data (“The MOSFETs and all other voltage circuitry used in the A7V133 are designed to handle up to 60A, see ______.”)

If that article had that kind of proof, all I’d be saying now is “Thank you.”

That doesn’t happen.

Voltage

A minor point, but any recently made voltage regulator goes down to 1.3V. That’s been part of the VRM standard for some time now. VRM was recently revised to bring the minimum voltage down to 1.1V. Although I have no
idea where the author got 1.6V from; it’s extremely unlikely we’ll see a Palomino dip below 1.3V, simply because it would drop the overall power requirement far below any reasonable expectation for Palomino. It might dip a bit below 1.5V, but as indicated above, this should not be a problem for any VRM out there.

However, as I indicated in the article, voltage is not the end of the story; amperage is.

Amperage

You see the figure “46A” tossed around in these discussions. If you’ve wondered where it comes from, it comes from the Abit press release for the KT7.

Not the most authoritative source in the world, but AMD has been mum, and it’s the best we have right now. We’ll probably not see AMD techdocs on Palomino on the AMD website until they’re ready to come out.

However, AMD has not been mum about the maximum speed of the Palomino in .18 micron, that’s going to be 1.7Ghz, contrary to what the author states. At the beginning of 2002, we’ll get a .13 micron shrink, and the Palomino will turn into a Thoroughbred starting at about 2Ghz.

The discussion of the difference between two and three phase power regulators is correct. The issue is not “Can a two-phase power regulator work under these conditions?” The issue is “Is this particular two-phase regulator supposed to work under these conditions?”

Leap of Faith

The author then presents his only piece of evidence that the A7V133 will work with a Palomino, essentially saying, “I saw some A7V133s work fine with a TBird running at 1.4-1.5GHz, and that would have chewed up an equivalent amount of power.”

Now that’s certainly a piece of evidence which would indicate that at least some A7V133s may be capable of the deed. That’s not conclusive proof.

Conclusive proof is a statement that the power regulator is rated up to, say 50A. That’s a statement which means, “our products are designed to handle up to this, and if they don’t, it’s our fault.”

That’s a far cry from “I saw a couple work.” OK, we know two or three or maybe a few more of them work. That could well be because of the cushion built into the components.

Maybe once Palominos come out, some will work, and some won’t. Maybe most will work, and a few won’t. We’ve seen that a number of times with mobos, the manufacturer will not certify that an older mobo will work, but most do. Key word: Most. Not all.

Right now, based on my testing of KT boards, I wouldn’t say anything is sure to work all the time, even if it’s supposed to. Take the multiplier, for instance. We’ve had boards where our multiplier worked, and yours didn’t. We’ve had boards where our multiplier didn’t, and yours did.

When the items that are certified to work often don’t, don’t you think you should be extra careful about those items where there is no promise?

Now it could well be there’s no problem at all with the A7V133, and they really are designed to handle the current required by Palominos. From what I see, it could end up being any of these possibilities.

We don’t know for sure which of these possibilities it will be.

We know Palominos are going to use less power. Just how much less, we’re not sure of. We also don’t know what voltage they’ll use. The range appears to be somewhere between 1.4V-1.6V, but these are educated guesses.

The only figure we know for sure is that 46A figure, and we don’t even know for sure at what Mhz that’s rated for.

Famous Last Words: “I’m Sure It’s Fine”

Quoting from the article:

In their article they mention that in the PDF it says 35A at 1.6V. Well I just wanted to clear this up, the 35A is just a demo schematic. The IC PDF itself mentions no limit on the Amps it can handle. Asus could have easily used different components that can handle more power, which I am sure is the case.

I looked at the two schematics. One showed what presumably would be a typical use of the circuit. The one used by the Asus shows 35A. The other, later one shows 55A. This is an indicator to me that the first probably is less capable than the second.

Is that conclusive proof the circuit can’t handle 50A? Absolutely not, and I never said it was. What I said was, if I had doubts about any of these voltage regulators handling a Palomino, it would be this one, and I’d get something definitive from Asus before buying one.”

Does the author present any evidence which proves these circuits will be fine? Does he look at the MOSFETs and find out that they’re all rated at 50 or 60 or 100A?

No. All he does is says that Asus could have used different components, and he’s sure they did.

Not a shred of proof. Just his word. Based on? “I saw a couple of them work.”

Do you think that explanation would cut it in any engineering project?

Learning the Scales

We then get an observation about how hot these MOSFETs are supposed to get. The author states that he did some (unknown) measurements on one, and came up with a nice cool temperature, and figured that a few more watts would only change things a few more degrees.

Now if he had tested temperatures at 1.2Ghz (just how we don’t know), and then 1.5Ghz, and saw that, he might have a point, but he didn’t do that. He did not determine how the temperature scaled on these MOSFETs.

However, somebody else did. To quote the Abit press release:

When future processors reach a 40A workload, ABIT’s 3-Phase Power will stay a cool 70° Celsius, while the transistors on other motherboards climb to 90°. When CPUs reach a 46A workload, ABIT’s 3-Phase Power will remain within the bounds of safety at 88°, while other power solutions hit 120°and higher! At 120°, not only will you have an excess of heat in your case; you will also risk damaging the mainboard’s PCB, destroying the transistors (rendering the board useless), or even damaging the CPU.

Ignore Abit’s claims about the temperatures from the competition for the moment. Just look at how hot Abit’s cool power solution gets: 70° and 88°C. Not 31° or 35°C as the author claims.

If even Abit’s temperature reading (which are probably the lowest possible) are more than 40°C above the author’s; there obviously is something very wrong with the author’s measurements as a guide, isn’t it?

Also notice that Abit’s temperature reading scales quite differently than the author’s; it jumps up 18° (and the competition’s 30°C). Abit’s measurement indicate that MOSFETs temperature climb at a much higher rate than the author’s estimates.

There’s also two issues here. Overclockers are usually concerned about how much heat radiates from a component. That can adversely affect other components. However, if the inside of the component itself gets too hot, that could be a real problem.

Again, could be. May well not be (not even Abit said it was a definite problem), but could be. The voltage regulator is rated at 150°C. Athlons are rated at 90°C. How many of you have Athlons running great at 80°C?

It could well be the inside of the MOSFET gets much hotter than the outside, but just because you can’t measure the temperature inside doesn’t mean you can ignore it. The MOSFET trying to operate sure won’t.

“I Don’t Know, But I’m Sure”

“While I can’t give an actual Amp limit for the A7V133, it should have no problems at all at up to 50A.”

If you can’t give an actual amp limit for the A7V133, you got no business saying it should have no problems at all up to 50A, because you don’t know. You’ve given absolutely no reason for that belief outside of seeing a couple work.

I’m Sure vs. I Don’t Know Enough Yet

This is not Ed saying, “It can’t work” vs. the author saying, “It will work.” This is Ed saying, “We’re not sure” vs. the author saying, “I am sure.”

Unfortunately, at least some people walked away from reading my article thinking I did say, “it won’t work.” Let me repeat, loud and clear, I did not say and am not saying the A7V133 won’t work with Palomino. I am saying we don’t have conclusive proof that it can handle the power, and you should find out from Asus whether or not it does.

We just don’t know at this point. We still don’t know.

As per the author, if he is so sure it will work, why don’t you ask him if he’ll buy you a new motherboard if Asus announces in a few months that the A7V133 won’t officially support Palominos? He should have no problem doing that if he’s so sure,

Pigs will fly before that happens.

Why in God’s name would you prefer to take the word of somebody who ultimately is not responsible for the consequences of being incorrect over hearing from the company making the product that is responsible for the consequences of being incorrect?

Wouldn’t it be much better to find out from Asus whether or not it will work than to rely on somebody saying, “I’m sure it’ll work, ’cause I saw it happen?”

Admittedly, there’s not a ton of certitude in this field, and maybe that’s something we should start demanding. Things come out of left field often enough. Maybe AMD will do something weird with Palominos and render all current mobos obsolete. I can’t swear they won’t do that; I greatly doubt it, but I’m not sure.

I do a lot of inferencing, and what I hope is intelligent speculation. However, when I do, I don’t say I’m sure about it. I’ll say “looks like” “appears” “likely” “probably.” These words mean something; they mean I’m not sure.

So far, the track record is pretty good, around 75-80%, but obviously far from perfect. The computer industry can be pretty secretive, and if you need to make plans for the future, sometimes you have to read tea leaves.

Nor am I arguing you should buy an Abit or IWill over an Asus. It might look that way, and might turn out that way, but what I really want to see is for Asus to answer the question. So should you.

Now you may want an Asus. You may think the evidence presented by the gentleman is good enough for you. It may well end up fine. I’m here to tell you it might not, and it’s not a sure thing at this point, no matter what others might say.

If you can’t afford to make an incorrect decision, either get the answer from the responsible people, or wait until the answer is set in stone.

Possibilites, Probabilities, Certitudes

You’re a policeman. Somebody’s just been stabbed to death. You find somebody a couple blocks carrying a collection of butcher knives.

What do you do?

Do you stop looking and hang the guy on the spot? No, maybe he’s a butcher coming home from work, and he brings his high-quality knives he bought home so the other butchers won’t swipe them on the job

Let’s say he tells you that, and you find out that story’s true. Do you just let him go at that point? No, he might be a butcher who decided to try his slicing and dicing skill on the decreased for a change of pace.

In either case, you look into it more. You just don’t jump to conclusions, and stop right there.

We have a situation here where we don’t know if A7V133 is innocent or guilty. So we don’t say “Yes.” We don’t say “No.” We look into it more. If for whatever reason, there’s no more time to do it, you say, “We think this, but . . . .” We don’t say, “We’re sure.”

Email Ed


Leave a Reply

Your email address will not be published. Required fields are marked *