Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Elmer who wrote (66159)	7/20/1999 1:33:00 AM
From: Jeff R	Read Replies (6) \| Respond to of 1585470

This is the history of the AMD Microcode lawsuit and other issues that plague AMD. AMD reversed engineered the 80287 Math Coprocessor and copied the microcode. Actually they contracted with another company to to do the design. AMD didn't want to use their technology exchange credit for the 287 design. AMD and Intel had a technology exchange agreement. AMD gave Intel the rights to a couple of chips, (.i.e Hard Disk Controller and QPDM graphics chip). Intel refused these chips and complained they weren't what they were asking for and were not compatible with Intel's process. Later it was found in court arbitration that Intel had purposely refused the devices so AMD would not have enough points to claim access to the 386. Subsequently. AMD went off and reversed engineered the 386 in Austin Texas. AMD made the AM386 static with lower power with a slightly higher frequency. In parallel, AMD requested arbitration to resolve the technology exchange dispute and claimed damages. Also Intel claimed breach of contract on a few small items. After AMD introduced the AM287, Intel immediately sued for copyright infringement. AMD had the patent licenses but claimed they had access to the microcode too under a separate agreement. True to AMD operations the lead litigator left AMD's representative law firm and the law firm lost the case. At this point of time the AM386 was history and AMD was just starting to sample the AM486. There was dark cloud over AMD. AMD could be liable for billion dollars of revenue and profits of the AM386 and AMD was unable to ship the AM486. Jerry Sanders immediately stated that a clean room 486 was already in the works. This was not true at the time and VP Bob McConnel eventually lost job. AMD spend the next 9 months trying to develop a clean room microcode version. AMD was never successful in developing one which ran all the code. They failed partly because AMD had no real documentation of 486 micro-architecture. They had product engineers trying to write microcode because all the design engineers were exposed. AMD eventually won the Arbitration hearing which awarded right to the 386 technology as a remedy for Intel's breach of contract. This really had no effect since the AM386 was completely ramped down. If was found through a Cyrix/Intel litigation that Intel had generated and internal memo which stated that one Intel executive believed that AMD had the rights to the microcode. Once this document was exposed which Intel knew about but didn't turn over, the 287 litigation case was thrown out and retired. As soon as the 287 court case was thrown out, AMD started shipping the AM486 with Intel's original microcode. AMD disabled the ICE function but the original microcode was still there. Eventually AMD won the 287 microcode case. AMD hired a new law firm which made intel executives look like chimps. AMD explicitly paid Intel for rights to microcode. I recall a few hundred thousand dollars. Intel claimed that those right only were for the the microcomputer system and not for firmware within the chips. AMD at the time of the agreement AMD was in both the chip and and microcomputer business with (Advanced Micro Computer). AMD won the retried case unanimously. At this point the cloud was lifted. Intel later sued claiming that AMD did not have the rights to the ICE code. AMD claimed that the pin was disabled but Intel claimed that the code was still there. Since AMD was already shipping millions of 486, AMD eventually settled with Intel on the arbritration ruling and paid Intel about 60 million and agreed that the could continue shipping the 486 for a couple of months until the fixed their mask sets which would interrupt production. AMD had also gained rights to make 20 percent of their production in outside fabs. Just when AMD thought they would introduced the 133 MHz version Intel brought out the 90, 100 and 120 MHz Pentium which completely obsoleted the AM486. The AM486-133 could only complete with the Pentium-75. In this agreement, AMD gave-up their rights to Pentium technology. AMD was working on the K5 design. The K5 was late because this was the first true AMD design and AMD's engineers didn't really understand the full extent of the x86 compatibility issues. Intel had years of verifications suites, etc and knew the architecture. The K5 design was more a risk design which was susposed to go fast. The original K5 design has a shallow pipeline stage and was inefficient in executing commonly used x86 instructions like string moves and segmentation reloads. The bottom line it the K5 design sucked. It was late, the frequency was low, it was a much larger die size than Pentium, and didn't offer a clock to clock advantage. The floating point unit was much slower as compared to the Pentium. The K5 had to be re-architected in some areas and completely re-layed out and just to get the frequency up. The original K5 barely made 60-75 MHz. AMD came up with PR rating since their integer super scalar unit was now fixed. Finally AMD's die size was considerably larger than Intel's tried and proven Pentium and just made it to 100 MHz in frequency in 0.35 micro process. Eventually AMD made version at 117 MHz for the PR150 rating. The later K5 design teams made a heroic effort to catch up but the Pentium was far ahead. Sensing the melt down in Austin Texas, AMD merged with Nexgen with had the NX686 on the drawing boards and had demonstrated their advanced architecture. The NEXGEN team basically saved AMD processor business with their timely execution. The K6 had twice the transistors as the K5 but was smaller. Because K6's design was originally designed for an IBM process, AMD attempted to duplicate the design in 0.35 AMD thought they had the recipe working in the SDC but when the process was transferred to Fab 25 there seemed to infighting and bickering so the process was not transferred but re-done. The Austin's fab engineers had this inflated ego because they were getting all the millions of $$$ for Fab 25. What they failed to realize was to transfer SDC's recipe exactly. The Texas engineers ineptness and NIH are the major reasons AMD has a poor reputation in manufacturing.