Hi wily; If transmission lines were perfect and everything, there wouldn't be a problem. I would guess that some of the memory makers suspected that there was a problem, but there is always going to be a divergence of opinions among engineers. A good 90% of all engineers aren't worth the powder required to blow them to hell.
It is obvious to me that the Rambus technology isn't ready for prime time in the high or middle PC market. It is also obvious to me that the technology leaves too little margin for the various things that happen to a system, non -linearities being only one of them. I think that this opinion has long been common among the better engineers.
But it is always possible to find engineers to argue the other side. At almost all companies, the final decision isn't made by any particular engineer. Instead, when the engineers can't come to a decision, the two sides make up presentations, and show them to management. Management makes up their mind largely based on the credentials of the two opposing sides, as management can never understand the technology as well as the engineers in front of them. This lack of understanding isn't due to management being stupid, it is only due to the distance that management has from the problem. So management reduces the argument to a decision as to who has the better looking resumes.
One of the credentials that management is going to look at is the credentials of the companies offering the technology. If it is MMTI, management is going to ignore them, (as they should). If it is INTC, management is likely to listen to it. This is despite the fact that a large number of engineers at INTC were against this technology. The dissenting engineers at INTC weren't brought along to sell the technology, or if one of them was, he knew to toe the company line, like any good employee.
To a certain extent, management at INTC painted themselves into a corner. They agreed to go with the Direct Rambus specification before RMBS designed it. Then, when RMBS came up with a specification that wasn't so close to the edge, INTC decided that it wasn't enough of an improvement over current art to bother with. So RMBS was sent back to the drawing board until they could come up with a situation that was good enough (read was maybe a little over the edge). Somebody on this thread suggested this explanation, and it seems pretty reasonable to me.
One of the things that humans hate to do is to admit that they were wrong. People at INTC, RMBS, Samsung, Dell, etc., no different from normal people in this. Once management decided that this was the way to do things, they naturally rewarded engineers for the usual gung-ho attitude that management loves to hear from engineers. Something like "We are working to eliminate all issues on this technology" sounds a lot better than "We are beginning to suspect that this is not going to be manufacturable." So management rewards the engineers that give them what appear to be solutions, though what they actually end up getting is their hole dug deeper.
It's sort of like the double or nothing attitude seen among stock investors (and that I have suffered from once or twice myself). Several times, in my career, when I have given a realistic estimate of the situation to management, they have ignored it, or chosen someone else to do the project, someone who didn't know what they were getting into, or who deliberately misled management as to how hard something was to do. This isn't a matter of incompetence on the part of management, it is just the ancient problem of decision making in a murky and unclear world. What are the alternatives, voting?
There is no way of avoiding these sort of problems. They are inherent in the fact that human communication loses information the vast majority of the time. When communication doesn't lose information, we reckon it a miracle, and call it "learning". It really is kind of odd that humans haven't figured out a way of organizing in such a way as to avoid these kinds of hole digging, but once a big organization starts digging a hole, they just don't stop until they are forced to admit it.
By the way, the above explains also why RMBS and INTC (and their shareholders) continue to be optimistic in the face of what is obviously a major fiasco.
So sure, some of the engineers at memory companies saw the problems in advance. In fact, it is obvious now that the engineers at Micron knew what was up all along. But management signs the paychecks, makes the decisions, and cannot possibly truly understand the issues.
-- Carl |
