Rambus (RMBS): Trial Notes 4/12/06 Part 1 Well, clos...

Reply Private New

Replies (1) Next 10 Prev Next

Send PM Follow Ignore

Followers	8
Posts	1020
Boards Moderated	0
Alias Born	08/31/2003

calbiker

Re: None

Wednesday, 04/12/2006 6:51:53 PM

Wednesday, April 12, 2006 6:51:53 PM

Trial Notes 4/12/06 Part 1

Well, closing arguments is all that’s left. Jury will start deliberating Friday for half a day.

I’ll start with some jury instructions. We’ve had discussions on DLL infringement. From a previous Whyte ruling, it appeared DOE (doctrine of equivalents) could not be used for DLL. And I had stated Rambus is hosed. It now appears DOE is the only argument R can use on DLL and Read Request. This is very good news. Either we misinterpreted the original ruling or it was changed, or Cal doesn’t know what he’s talking about. Ha. At least I feel vindicated for my previous negative comments on DLL infringement. But I wasn’t worried, even with LOLo’s questioning Cal’s calibration. HA.
Cal did get a re-worked calibrated heart. Resting heart rate = 50 bpm! wink

Just a note from the ending. Looks like Rambus employee Woo is in the hot seat. Ha. Hynix wanted to enter into evidence a slide from the Japan Rambus Developer Conference after the jury was dismissed. Stone says they can’t do that. Woo was on the witness list. He would have been called if Hynix was to use this slide as evidence. Stone wants to re-open the case if the slide gets into evidence. Life is not easy for Whyte. Ha. Whyte will think about it. The slide describes an increasing performance gap between processor performance and memory performance (speed). Horowitz just got through testifying that they solved the performance gap. That would be a powerful closing statement for Furness.

Horowitz was on the stand when Furness questioned him about the slide. He never seen it. He can’t say anything about the slide because no info is given. He also made a comment about the difference between a marketing person and a ‘real’ engineer (or something to that effect). And yes, we’ve seen the difference here. LOL.

Just a little background. In 1990 there were two performance throttling issues: latency and bandwidth. The R inventions solved the BW problem. Latency still remains a problem. We don’t know what Woo was thinking when he made the slide.

This is probably the bad boy.
http://www.rambus.com/events/Track1_1XDR_Rambus_Woo.pdf

Getting back to the start.

Horowitz defined the terms listed in a architecture overview.
R. channel: set of wires to connect the memory to controller (notice he didn’t use ‘Bus‘)
R. DRAM: Special high speed interface
R. protocol: relates to interface: var. block size, access time register, arbitration

R started with 500 MB/s devices, now are up to 2.5 GB/s. Factor 5 improvement.

Effects of different features?
Need to add DLL when clock times become shorter.
Access Time Reg. - most systems require it to operate synchronous memory. A fixed delay time would not work well. 2 parts would have to be built. [Some systems require (for example) a delay of 2, some a delay of 3. Both chips would have to be built it the delay is not programmable.]

S: Did you have a roll in the patent application?
H: Yes, wrote more than Mike. He wrote it as he would write a technical paper. Wrote most of the text in the first patent application.
S: Had a roll in writing claims?
H: No. A skill I didn’t have.
S: What was the goal writing the spec?
H: Describe all ideas we came up with. Never occurred all ideas had to be used together. [An argument against H contention that they don’t use the narrow bus]

S: Describe multiplex bus.
H: Don’t need to use a multiplex bus. Cites pg. 12: “In the preferred implementation 8 bits … but a person skilled in the art can use 16 or other data lines.”
S: What’s preferred embodiment?
H: The best way to use all features (thought at that time). Rambus DRAM is also not the preferred embodiment.

Horowitz then scanned the original patent application to point out features that are at issue in this court.
He talks about: prior art, Elxi (?) , that they did not invent (DLL, dual edge clock, multiplex), clock scheme, bus interface, protocol and high speed wires, large block transfers (variable burst length), registers providing control information (one such is access time reg.), remove skew (DLL), multiplexing (used in dual edge clocking (not bus multip.)), dual edge clock. [There may have been more]

S: Did you include the word ‘DLL’
H: No, Don’t know why now.

H: The goal was to increase BW. As a result, a narrow bus could be used.

Fruness cross to follow. Damn, that was only 1.5 pages of 7. Don't know if we'll get this done.