Space for Rent? Anandtech Rambus “Editorial”

Credibility and integrity

Why do you visit a website like ours, or for that matter, anybody else’s? Usually, it’s because you think you’ll learn something there. That means you have some faith in the person or persons you are reading.

Credibility is the only asset a writer has. He or she need not always be right; we are all human. Nor may the person always be fair as you might see it; the reasonable can disagree on many things.

To be credible, you must personally exercise your good faith and honest judgment. It is yours, and yours alone. You cannot lend or sell or rent it. If you try, all you do is lose it.

I am afraid that is just what happened in AnandTech’s Rambus DRAM: Uncovering Facts & Burying Rumors.

Op-Ed For Hire

Here’s the key paragraph in the piece:

“In an effort to help clear up some of the major misconceptions about RDRAM, we were contacted by Rambus and asked to meet with them under the assumption that they would be able to answer pretty much all of our questions and give us an update as to exactly where Rambus as a company and RDRAM as a product stands today and where they see themselves and their product(s) in the near future.”

Dangling modifier aside, they wanted an ad in editorial clothing. And that’s what they pretty much got. It sure looks an awful lot like a cut-and-paste between RAMBUS advertising along with some ifs, ands and buts.

You may say, “It wasn’t an article, it was an editorial.” Sure, but whose?

Figleafs

There is nothing wrong with Rambus trying to get its message across. Nothing even wrong with writing editorial comment. You see that in magazines all the time, but do you know what? There’s something marked on each and every page of such “editorials”: ADVERTISEMENT.

Do you take them just as seriously as regular editorial material? I think not.

Nor does a pro-RAMBUS opinion necessarily bother me. I would be just as perturbed if someone did the same with DDR memory. My only biases are for high performance at a good price, and against the opposite.

If you saw a book review of a John Grisham novel, and the reviewer let John Grisham review the merits of his own book without mentioning the praises were coming from John Grisham, how honest is that? How credible?

Sure, you can interview John Grisham about the book, but that doesn’t pretend to be a book review like this pretends to be an editorial evaluation.

Just whom do you think more likely to have inspired a comment like, “But never has there been a situation where so much hatred for a particular company and their products has been so freely expressed and accepted alike by end users and members of the media.” A beleaguered employee of the despised company, or an allegedly impartial journalist?

Wow, we aren’t buying and dislike $15 a gallon milk just because it’s overpriced. We’re committing a hate crime.

I have to wonder, though, what’s next? Are we going to see editorials dictated by Intel on how wonderful those 810 and 820 motherboards really are? Will 3dfx get an editorial denouncing the full color Nazis? Will Via get to prove that it’s just a hate crime by the BX Klan to call the Apollo Pro lousy? Where does it end?

Just how many more “injustices” need to be addressed? And just what do the “victims” have to do to have you address them?

Inside jokes?

Just who do you think more likely to have inspired a comment like, “Until this point there hasn’t really been a reason not to like Rambus and this “wonderful” RDRAM technology they’ve brought with them,” An employee of the despised company, or an allegedly impartial journalist?

Yes, “wonderful” is in quotation marks. This implies sarcasm. It is truly rare that one is sarcastic about one’s own article. Especially when it is supposed to be an expression of your own personal opinion?

Why? Why spend pages regurgitating the words of RAMBUS personnel, then imply you don’t mean it?

(See Firing Squad and see the rather close similarities between what the Rambus exec says and what’s in the article. Actually, the RAMBUS exec was a bit more honest and forthcoming than the article.)

This Isn’t Bob Woodward

The numerous dubious comments unchallenged in this piece are an abdication of editorial judgement and journalistic responsibility. No surprise coming from the Rambus folks, but this was supposed to be written by journalists, not stenographers.

I’ve spent some time recently looking closely into RDRAM compared to DDRAM. There was nothing in that article that couldn’t have been found on the Web, and unfortunately, there was much on the Web that should have but wasn’t in that article.

Just a few highlights:

We’re Just Friends

“Intel’s relationship with Rambus, although a very public one, is limited to one specific thing: Intel is a Rambus partner that has a license to produce Rambus memory controller devices”

I saw this and decided to see what I could find in the SEC reports. Found this in five minutes:

“In January 1997, the Company granted a warrant to Intel Corporation for the purchase of 1,000,000 shares of Rambus common stock (the “Intel warrant”) at an exercise price of $10.00 per share. The warrant will become exercisable only upon the achievement of certain milestones by Intel relating to shipment volumes of Rambus-based chipsets (the “Intel milestones”).”

That’s limited? If RDRAM becomes the memory standard (at which point they’ll certainly meet the milestones), a share price of $500-$1,000 is a quite reasonable estimate. I guess they forgot about something potentially worth a half-billion to a billion dollars. I do all the time, don’t you?

Never mind that if Intel does muscle RDRAM through, if selling ten million chipsets is worth that kind of money, how much might one or two hundred million chipsets packed with RDRAM be worth in warrants issued by a grateful RAMBUS?

1.6 X 2 is greater than 2.1 X 2 if I never do the second multiplication

First we get told we suddenly need 3.7Gb/sec bandwidth, more than three times more bandwidth than we currently have. Never mind nothing, not even Quake, terribly strains what we do have. Maybe it will in a couple years.

Then we look at next generation RDRAM and DDR, and, oops, neither of them make it. Double oops, DDR actually does better. Uh-oh, better start talking about dual-channel RDRAM, oops, that doesn’t make it either.

Whatever you do, don’t multiply DDR’s faster rate by 2 to figure out what dual-channel DDR could do. That would show dual-channel DDR could meet “The 3.7Gb Challenge” (if indeed that’s so important) and dual RDRAM can’t, and who’s running this show, anyway?

Instead, we forget all about counting bandwidth and start the “Almighty Pin Count”. Nasty old SDRAM all of a sudden has too many expensive pins, and it just becomes obscene when we talk about DDR and that dual-channel.

How much do you think memory slots cost to make? The whole board costs about $100, and I doubt most of that goes into memory slots. Never saw any great RDRAM board bargains, have you?

I’m sure it’s not the easiest thing in the world to handle all those traces. Then again, it’s probably not too easy building traces that handle much higher speeds, either. At first glance, doesn’t look too disproportionately different to me, but I could be wrong.

Understand the main difference between SDRAM and RDRAM. SDRAM is like a fountain, it pushes a lot of water fairly slowly. RDRAM is like a fire hose. It pumps less water much more quickly. You measure the total water coming out in this case, it comes out about the same.

If you want the fountain to gush more water, unless you want to widen the fountain, you probably have to pump water faster. If you want the fire hose to, unless you can figure out how to make water go even faster, eventually you have to make the hose wider.

When you read about RDRAM having more headroom than SDRAM, remember that if you keep pumping water faster out of the fountain and widen the fire hose, eventually you end up with the same thing.

This is not to say DDR will be easy to do and that there aren’t legitimate difficulties, there are. But we don’t have them spelled out here. Nor do we hear a word about the difficulties of going to quad-pumping as opposed to dual-pumping with RDRAM.

Instead, we get (comments in bold mine):

“Once you reach a certain point (What point?) you’re going to have to start adding more layers to the motherboard’s PCB (Why?), which begins to increase the cost of the motherboard at a pretty hefty rate. Some motherboard manufacturers such as ABIT won’t even consider a motherboard design that employs a design with more than 4 layers; chances are that if you said that an 8-layer design (Where did you come up with that figure?) was required in order to have dual SDRAM banks, they would opt for another solution.”

We’re given absolutely no reason outside of an assertion to believe we need a PCB board twice as deep, just a tossed-out figure. Why don’t we question these assertions by the not-most-unbiased people in the world when it comes to DDR? And gee, Mr. Anand, if getting to 3.7Gb a second was as important as you said, wouldn’t it be worth some extra pins and tricky traces to more than reach that goal, as opposed to falling short with RDRAM?

And why do we need bricks for mobos moving high-speed SDRAM signals around, but not for even higher speed RDRAM signals that, since they have fewer pins, should be moving more data per trace?

Where’s the Beef?

If I’m sitting with a bunch of RAMBUS people, all I want is one line of questioning. I can even boil it down to one question:

“Why hasn’t there been anything vaguely near the amount of RDRAM needed by Intel over the last months, and what the hell makes you think it’s going to be any different next time around when Willamette will requires tons of it?”

I get that answer and all the details I need to support it, or I tell you I got a song-and-dance.

That’s the reason why RAMBUS costs what it does. Not because it costs outrageously more to make, but because very little of it is being made, and whatever is being made is going into OEM machines (at a much lower price than what you see at Pricewatch).

He asked the “What.” He didn’t ask the “Why.” The “What” is spilled milk. The “Why” is what matters.

If RDRAM is so wonderful, why aren’t the memory manufacturers making very much of it? That is the question, and he didn’t ask it. Intel went through the whole 820 fiasco because it wasn’t around. We’ve found ourselves overclocking BX boards or buying Via boards because it wasn’t around. If there isn’t enough around early next year, Willamette is dead meat.

It’s just one of the biggest if not the biggest PC stories of the year 2000, and Round 2 commences next fall.

So if I’m sitting in a room with these people, I want numbers. I want details. I want explanations as to why Intel was left holding the bag. I want to know who is making RDRAM PC memory, who is going to make it, and how much of it can they make now, and six months from now. I want to know how long it takes and how much it costs to convert an SDRAM production line to RDRAM. I don’t want to hear cutesy chip counts, I want to know how many RIMMs can be made, both now and in the fall/winter.

Of course, any answers will only make sense if before you walk into that meeting, you know how big the PC memory market is, and estimate just how much RDRAM is going to be needed to make Willamette fly.

Let’s just say that by early next year, if RDRAM isn’t approaching 50% of total PC memory production, Willy and Intel are in big trouble.

Besides that, I would like the RAMBUS people to explain to us all just how you encourage memory manufacturers to convert to making your products by suing them or implicitly threatening to sue them. I want to know why Intel is using DDR for servers.

And if I don’t get answers, that’s what I’m going to report.

What do we get here instead? Just a vague, “Not much will be made now, more will be made later” answer. No details, no production figures, nothing to give an inkling as to what had happened and what will happen. In short, no work.

Whose Website is it, anyway?

I really got the impression this article was like being given a school assignment to write an essay on “Why I Like Hitler.” From some odd comments here and there, you get the impression the author is trying to signal “I don’t really mean this.”

But this is not a homework assignment. Nor is it an assignment given to you by the boss that you have to do or be fired. So why was it written, and written the way it was?

Why not write an article comparing both DDR and RDRAM, pointing out the merits and demerits of each as you see them? Why just write this from the Rambus perspective? Couldn’t the misconceptions just as easily been addressed from an independent point of view? Why was there so little questioning on issues that should have been questioned? If you can’t spare the time or effort to do the job right, then don’t do it at all.

I have no idea what may have transpired. I do know, though, that I wouldn’t have written anything like that. I would have been very averse to put myself under any obligation, past, present or future. I absolutely would not let anyone dictate what had to go into an article, snide asides notwithstanding.

Actually, I pretty much did that a few days ago (“Willy and RAMBUS”). I didn’t need anyone traveling to or from Rambus headquarters to do it, didn’t need editorial assistance, didn’t write a word under obligation to anybody, past, present or future.

But I’ll tell you what I did need to do. I needed to judge what I saw, cross-check and not accept just anyone’s word. For better or worse, my name, not someone else’s, not Rambus’, was on that, and that means something to me.

A lot of you may be quick to say the author was bought off in some way. I won’t, first because we don’t know that. It could just be lousy work and judgment, but then, that’s not too good either, and needs just as much feedback.

Secondly, even if it were true, it’s the abdication of personal responsibility that’s important, not the size or even the existence of any bribe. If Judas Iscariot betrayed Jesus for free, would Jesus have said, “Cool?”

I will not toss the first stone, but rather the first question: What happened? What were the exact circumstances and conditions under which Rambus arranged this? Any freebies, now or in the future? Any RDRAM banner ads or sponsorships coming soon to a website near you? Any other quid pro quos?

Am I doubting your integrity? Sure I am, and so are many others, including quite a few people in your own forums. The article practically begs for it.

I’m at least giving you the chance to clear yourself and explain just how this abject abdication of the responsibility that comes when you attach your name to something came about.

The more I look at it, though, the less it looks like bribery and the more it looks like half-assed work and bad judgment. You were clearly unprepared. You clearly did not do your homework before going in there. You clearly had not thought through the issues, and you were not ready to ask the tough key questions that needed to be asked. You were far too willing to just accept what was given you; you were sold a bill of goods, and you tried to pass it off as your own. And you clearly were not too honest in how you presented the material.

This is not good. You let the people who come to your website expecting good work down. It is because of them that you get the ad revenues, but if you are not credible, they and the ads will go elsewhere. They deserve better than that. If other obligations leave you too busy, have someone else do it, or wait until you can do a good job. This was worse than nothing.

Folks, read the article, and make up your own minds. If you agree with me, or even if you don’t, you should go to Anand Forums and make your opinion known. People can always make mistakes, the key is to keep them from happening again.

As always, you can write me, too.

This is not a very honest industry. It will not get any better unless we insist it gets better.

Email Ed


To put it mildly, I’ve gotten a lot of response on the subject.

Just a couple points:

Some of you concluded I’m anti-Rambus and anti-Intel. Wrong. What I am against are extremist arguments which largely ignore
all sides of the truth. To the extent that was the case in the article, I went after it. Had the article been written the same way, but it was Micron doing the calling and Samurai DDR the subject, I would have blasted every bit as much.

I didn’t conclude money or favors had to have changed hands for that article to have been written. I only mentioned it as a possibility. As I said in the article, it probably was a case of bad judgment. It was so bad that it left an impression in many minds that the gentleman had to have been paid off in some way, and they didn’t need me to come to that conclusion.

I simply asked the circumstances under which this article was put together, and the gentleman declined to give an answer. This is not conclusive evidence of a quid-pro-quo***, it probably is more the case of refusing to admit error or poor judgment. That’s not as bad, though hardly good.

None of us are perfect, and we nor our minions should conduct ourselves as if we were. For instance, I made a mistake in saying that Willamette was going to require the next generation of RDRAM to function. Several of you pointed that out to me, and after looking at it more closely, you’re right. I see how I made that mistake, (and it is key to understanding much of this) but nonetheless it was made, and it would be no less wrong if I played ostrich.

It’s Easy To Bash; Let’s See You Do Better

It is a lot easier to criticize a piece than to write one. The big problem with the RDRAM/DDR situation is that it has become polarized, with people taking one side or the other. Truth always is the first casualty in this kind of fight. The entire issue has become almost completely muddled by the mis- or even in some cases dis-information being presented by BOTH sides.

I had not planned on writing about RDRAM/DDR after my Willamette/RAMBUS piece because the essential pieces are not in place yet, and won’t be, for some time yet.

However, RAMBUS PR offensive notwithstanding, it probably is a good idea to take a look at this issue now and separate truth from propaganda. So I will.

Beginning early next week, I will begin a series of articles on RDRAM/DDR: Discussing the differences in how it works, a brief history of the players, the production and cost issues, along with various technical issues like latency and granularity.

I can tell you this right now; the two sides are pretty evenly matched. They both have strong and weak points. Anyone declaring inevitable victory now is simply a fool. RDRAM is no inevitable winner in this contest, but neither is DDR. Either one could win and the decisive steps are still quite a bit down the road.

Now that I’ve gotten ideologues on both sides mad :), if there are any particular points you think are important, write me. I don’t plan to get bogged down in tiny technical details, though, so if you tell me RDRAM (or DDR) rules because it takes one less clock cycle than the other for one particular action, please understand if I find that less than critical. 🙂

I try to answer all my email (and if I’ve missed a couple of you here and there, I’m sorry), but for this, if you do send me something, please understand if I don’t respond beyond a “thank you” unless I have further questions to ask.

Thanks to all of you who liked the article and wrote me, and thanks to all of you who disagreed and said why.

The issue is not who is right; it is what is right.

***One rightly could ask at what point dependence on testing samples or advertising affects editoral integrity. There is an inherent conflict-of-interest in reviewing merchandise or receiving advertising revenue provided by the merchandisers, and it is a subject I will pursue fairly shortly.

Email Ed


Be the first to comment

Leave a Reply