Hi jim kelley; Re: "A question for you...who invented the techniques incorporated into SDRAM and DDR if Rambus did not?
"
This gets back into the subject of how much genius it takes to make DRAM into SDRAM.
Almost all that you need to do to convert a DRAM into SDRAM is to add registers to the inputs and outputs. This requires about 10 minutes with a schematic capture tool. All you do is add registers to the inputs and outputs. This was essentially IBM's idea, though Texas Instruments started working on it 1988.
As soon as you add registers to a DRAM, you start to think about adding clock stages to the internal paths in the DRAM, so as to make it a pipelined device. It's obvious how to do this, provided you don't require that the logic work with different CAS pipeline delays.
Hitachi solved that problem, and filed for a patent on it around 1988. Their solution gave a DRAM that had two different pipeline delays for CAS, a total of either 2 or 3 clocks, and I believe it is the solution now in use throughout the industry.
Given that your DRAM has two different CAS latencies, you need to store a bit of control data to determine which latency to use. Since the industry had already passed control data to VRAMs (which were a niche specialty DRAM with an extra read/write port) around 1986, the engineers simply used the same technique with SDRAM.
When Rambus wrote their 1990 patent on RDRAM, they couldn't claim the controllable latency because it was a (painfully) obvious application of two things that were already patented, namely the selectable latency, and control bits inside DRAM.
In other words, Rambus didn't teach the industry Jack-$h:t about controllable latency. And their patents reflect that, they only cover controllable latency for their own weird bus.
The German case apparently involves the old Rambus patent documents from 1991. These patents don't claim anything in SDRAM. If they did, Rambus would have asked for royalties in 1991. Instead, Rambus had to continue their patents until they got language from the (over worked) patent examiners that could reasonably be held to cover SDRAM.
Here's a prediction for the German patent case:
IT'S GOING TO BE OVER MUCH MORE QUICKLY THAN EXPECTED, AND RAMBUS IS GOING TO LOSE.
That is to say, after the judges get the case, there is going to be speculation on how long before the ruling comes down. When that speculation arrives, I will update this prediction with a more precise time.
And when Rambus loses, all the losers that are hanging onto this sinking ship are going to blame German mendacity for the loss. And then Rambus will lose in the other European locales, and the same tired excuses will be trotted out. Same with the Micron and Hyundai cases.
Get started early, maybe you can come up with some new and different excuses for this one.
-- Carl |
