To: richard surckla who wrote (30600 ) 9/25/1999 4:27:00 AM From: Bilow Read Replies (1) | Respond to of 93625
Hi richard surckla; Re: Rambus has done it right and the problem lies elsewhere, with Intel, Dell, etc. I probably see the responsibility of a design engineer somewhat differently than the usual engineer. I see a design as a set of instructions for taking a pile of smaller parts and combining them into something useful. This is a lot harder than people think, because it is in direct, apparent, violation of the law of thermodynamics that says that entropy always increases. Okay, we avoid violating this law by taking energy from the sun or whatever. But it still a fact that the number of designs that create a useful product is infinitesimally small compared to the number of designs that decrease the total value of the component parts. This probably requires a simple example to explain. A simple design: Mix 1000 33 ohm chip resistors with 1000 47 ohm chip resistors by swirling them around in a dog food dish, using a blender set on medium for 30seconds. It is an obvious fact that the resulting collection of chip resistors is worth considerably less than the sorted resistors that it was combined from, due to the fact that time and effort will have to be expended to separate the resistors (the ones that remain unbroken) back into their source bins. Standard technician practice with such misexperiments is throw the whole lot away. This sort of economics is now being forced onto the motherboard manufacturers who now have to unsolder, unconnect, and sort back together any components they wish to save. This disaster is not some kind of unusual thing in the industry. It happens every day, all the time. But normally they manage to do it in a secret enough place that it doesn't leave me rolling on the floor with laughter. I was once at a place that, due to a small board change, ended up destroying 200,000 copies of a board I had designed. (Somewhere in layout, the board lost all its ground plane connections.) Of those boards, they had sadly stuffed 50,000. We were horrified when we found out, as the PCBs cost about $10 each, and they were being stuffed with $150 of parts each. This was a lot of money to that company at that time. But it was so horrible, like a caricature of the secret bete(sp?) noir that haunts every good designers sleep, that after five minutes, everybody burst out laughing. The sad fact is that partitioning the blame solves nothing. The Rambus requirements were beyond what was easily achievable. Blame Intel. Blame Rambus. Blame Dell. Blame someone else. Blame them all. At 800MHz, there will plenty of opportunity to root around and fix the blame on someone. The layout woman updated her resume and mailed it out, though she did not end up being blamed. Engineering wise, fixing the blame does no use. It is used only after companies have given up hope of dealing on good terms, when they have decided to p:$$ as hard as they can in each others faces. At that time, the lawyers come out, and everybody tries and figures out what will happen when they take the civil suit(s) to a jury or judge. The end result doesn't matter, it almost never pays to litigate these kinds of things, they settle. The fact is that nobody would have deliberately caused a design to be faulty. As far as absolving Rambus of blame, we should remember that their specifications are very, very, very hard to meet. If you build cars that can only be safely driven by people with the instincts and nerves of Mario Andretti, do you blame the drivers or the manufacturer when the wrecks pile up on suburban streets? It doesn't matter. -- Carl