In reply to: 2bStealthy who wrote msg# 235 Date:12/23/2004 1:56:27 PM

Post #of 237

2b, you could be right. I was looking at the court documents and on December 20th, Woody Higgins, the lawyer who filed the patents, will be allowed to testify which is a set-back for the defense. The defense argued that Higgins shouldn`t be allowed to testify because of client-council privelege with Moore, but the judge ruled in favor of PTSC.

I also have a copy of the letter written to Woody Higgins by Russell Fish dated September 12, 1992. Fish writes about how the patent office has split the patents and he suggests that ``Chuck`` and himself should prosecute the patents separately. Fish also says that Moore`s patents are really not all that important. I think they will settle or be acquired.

Also, is it a coincident that Higgins is allowed to testify on the December 20th and Donald Bernier resigns on the same day due to ``time constraints`? Are you serious? Let me know what you think.

The letter is as follows:

To: Woody Higgins

cc: Helmut

From: Russell Fish

Date: September 12, 1992

Subject: ShBoom Patents

GENERAL

The patent office has split our one ShBoom patent claim into the following:

1. Triple bus multiplexing.

2. Multiple instruction fetch.

3. DMA coprocessor.

4. Automatic sensing of number of memory chips.

5. Fish Clock.

6. Merged stack/register architecture.

7. Automatic polynomial generation.

8. ``ShBop`` monolithic CPU & DRAM.

9. Method for prefetching

10. Stack caching.

One possible ramification of ten separate patents is that each of these ten ideas had a single inventor. To the best of my recollection 1, 2, 5, 6, & 8 were mine, and 3,4,7,9, & 10 were Chuck`s. #10 did have significant input from me. If you ask him, I believe you will get the same answer on all 10.

In my estimation, 1,2,5,6,& 8 are also the most useful and valuable potential patents.

The prototype chip used Moore`s patent #3 for the coprocessor. The new chip uses my coprocessor/DMA design which takes a different approach and achieves superior results.

Moore #4 is a neat idea, but not essential.

Moore #7 was not implemented in the prototype nor in the current chip.

Moore #9 has been modified in the current chip and the patent is probably not essential to ShBoom.

Moore #10 was not implemented in the prototype and the current chip uses a different technique jointly invented by Shaw, McClurg, and myself.

ANALYSIS OF SIGNATURE

ShBoom gets its speed from #2 and #6. Multiple instruction fetch gets around the Von Neumann bottleneck. Merged register/stack architecture allows single cycle math and logic operations without a pipeline.

You cannot build ShBoom or anything like it without these two patents. These two patents are complimentary and increase the speed of ShBoom by from two to three over a similar computer which doesn’t use these techniques.

#1 is a cost reduction and power reduction idea. By sharing pins, wires, and drivers three ways. power consumption is reduced by two thirds, chip area is reduced by 50%, and package cost is reduced by 70%.

Without #1 you can make a part which runs as fast as ShBoom. but it will cost three to four times as much and burn three times the power.

#5 and #8 work together and are the key patents necessary to build a fast cheap merged CPU/DRAM. The Fish Clock enables performance optimization which is only possible when memory and CPU reside on the same piece of silicon. Neither of these is used in the current ShBoom.

PRIORITY

#2, multiple instruction fetch, is the key patent. I suggest that we prosecute it first. From a competitive standpoint it is a SPARC killer.

#6 is important but less likely to be copied since it goes against the prevailing wisdom of computer architects. Very few computer architects (outside of Burroughs/Unysis) understand the implications of an arithmetic stack. No microprocessor architects have given it a second thought. I would prosecute it next.

#1 is important for building a cheap ShBoom, but it is only significant if you have #2 and #6. And if you have # 2 and #6 you probably don`t need #1. #1 is potentially important when combined with a memory architecture I designed for alliance (but still own the rights to). I would do #1 next.

#5 and # 8 are only significant if we ever decide to build or license a CPU/DRAM which uses the techniques. I suggest we prosecute these last.

THE MOORE PATENTS

When we were defining the original ShBoom architecture, Moore was focused on designing a theoretical machine. I was focused on making and selling damn fast, dirt cheap chips. For this reason, my ideas are probably the more commercially valuable.

All of the Moore patents are either not used in the in the current chip (#3, #4, #7, &10), are a mathematical curiosity (#7), have been superceded (#3 & #10), or can be easily designed out if necessary (#9).

In the interest of economy, we might even leave it to Chuck to prosecute his own patents. He also might accept and even be happier with this arrangement.

http://www.investorshub.com/boards/r...