It sometimes come to my attention that some people have a mangled view of CPU history. Now, I don't mean to be picky or smart*ss, since I was not paying all that close attention while it was going on myself, and it didn't seem very interesting at the time, but the misconceptions seem to give people the wrong idea. In that vein, I was so thrilled with my exposition 
in a comment at the news section that I decided to drop it into a thread, and hope to get more info.
> an agreement ... allows AMD to produce x86 CPUs despite the fact that
> Intel holds the patent for this technology. Without it AMD wouldn't be
>able to produce x86 anything...
Intel and AMD have cross-licensing of patents, although that does not give them the right to copy the others design. They have to do their own original. They can use the patents in their design.
But it is has been perfectly legal, I think, for any company to make a work-alike of anybody else's CPU since Intel lost in court to Zilog. Zilog was a company formed mainly by former Intel high-ups to functionally clone the 8080 (Intel's 8 bit CPU), only do it right. The first Zilog Z80 ran twice as fast (yes, believe it or not) as the Intel 8080, which highly POed Intel. It ran all the same op-codes as the 8080, and a few extensions.
(No reason to be surprised Zilog could double the speed. Back then Intel did not believe in pushing the edge. I suppose that is what motivated the employees to leave and start Zilog. Intel started pushing the speed edge with the 486, and really changed philosopy with the Pentium.)
The original Cyrix (not the same-named entity at VIA) 6x86, for instance, was produced as a clean-room design (that is, without any knowledge of the Intel chip internals) without any permission from Intel. Intel sued Cyrix over and over about patents (although not for cloning x86), but never won and never stopped Cyrix. (You don't have to win to crush a smaller competitor with legal fees.) Naturally it is simpler legally, and worth some money to get Intel's permission, if they were inclined to give it, but it isn't necessary.
BTW, Cyrix counter-attacked Intel legally on patents, and got a settlement out of Intel. Through a set of ownership transfers of Cyrix, VIA ended up getting the right to make chips for the PII socket, although AMD could not.
When the LSI business was pretty new, it was a practical necessity to have a (legally binding) second-source agreement with another LSI company. Every LSI company paired themselves with another. It was considered business suicide to use a unique chip from an only-source, because you would be SOOL if there were the slightest hiccup in delivery. AMD and Intel used each other as a second source. As such, both had complete access to the other's original designs, including masks. In those days Intel was not a big deal (believe it or not) and AMD was doing them a bit of a favor, in return for which Intel was supposed to make AMD-designed chips, although Intel screwed AMD and didn't. For instance, AMD used to make something called bit-slice chips (something like 1/4 or 1/8 of a CPU) for minicomputers, and other specialty chips, such as a fast hardware binary multiplier, so AMD was not a leach on Intel or a slouch. 286s, 386s, and most 486s from AMD were minor reworks of Intel designs, and generally ran at higher clocks. By the 486, Intel wanted to get out of the contract and stop AMD. AMD won some and lost some in court, but ultimately started producing 486s of their own design. By the time of the Pentium, AMD was not doing well on the design front, and bought Nexgen, a company that already was producing almost-x86 chips based on a RISC back end, a style of design Intel adopted later with the Pentium Pro/II/III/4. Some K5's, the K6s, and the Athlons come from this.
The 8080 vs Z80 conflict was no more cut-throat than others. The original Apples from Apple used a 6502 (from MOS?) rather than the 6800 from Motorola. The 6502 was a near duplicate of the 6800, with some improvements. There was a 1 to 1 correspondence between the opcodes, IIRC. It is almost funny that Apple went to Motorola chips for the Lisa/Macintosh when Apple decided to copy the Xerox Star design. It is even funnier that for years Apple was suing MS for copying Apple's copy of the Star, until Apple fell on hard times and couldn't afford the legal fees. Apple was in a position like countless other little companies that Apple crushed for stepped on its toes, only Apple didn't settle with the little guys, just put them out of business. If Gates was like Jobs, Apple would be out of business. Jobs put Steve Wozniak, the actual sole designer of the original Apple and co-founder of Apple, out of business when Woz started a company essentially unrelated to Apple's product.
Now back to the patents. While AMD was developing the successor to the K6, Intel decided to terminate AMD with prejudice My understanding is that AMD licensed the DEC EV bus for the Athlon because, unlike with the Pentium, Intel did not license the PII bus, so AMD couldn't use it. I don't know how that squares with cross-licensing. It doesn't. AMD licensed the DEC EV bus while DEC was on the ropes and about to expire. Not long after, in an odd turn of fate, DEC was picked up by Intel (I believe) when it finally gave up. IAC, it was believed that AMD could not survive without being plug-in compatible with an Intel chip, the reason being AMD would have to use a separate chipset and a mobo unique to AMD. What mobo company in its right mind would make a mobo only for AMD CPUs? None. What chipset maker would make a chipset only usable with AMD CPUs? None. So that was the end of AMDs x86 business.
But AMD developed a chipset of its own and a mobo design that used the same parts, or readily adaptable, as Intel used in PII mobos. AMD gave away the chipset design to whomsoever would like to use it. And the total mobo design is such that a straight-forward revision changes a PII mobo into an Athlon mobo. When Intel reverted from a slot cartridge back to a socket, so did AMD. When Intel changed chip packages, so did AMD. It was so neat that practically every mobo maker eventually did an AMD mobo to pick up a few extra sales. That turned out to be fortuitous for them when Intel terminated the PII/PIII unexpectedly early, when Athlons seemed about leave them behind, but mobo makers could continue with the mobo design now using Athlons.
But that was later. Advance rumors and benchmarks of engineering samples had the Athlon FPU at a pathetic half the Pentium Pro/II, although the real thing turned out to have the most killer x86 FPU ever produced by anyone up to then. AMD had lined up several chipset makers for Athlon mobos, but only AMDs expensive original, meant as a prototype, was ready at the Athlon intro. VIA, which was later to make a fortune for years as the premier Athlon chipset maker, preferred to work on the Intel version, where the money was. Really good and competitive Athlon mobos took a while to arrive, as Taiwan manufacturers gradually realized the Athlon was going over, unexpected to them, or to anyone really, and there was money to be made.
I'm not sure what point I'm trying to make. Its not a total history. Just pieces where I see people have gotten a strange angle on it due to things not brought up much.
For instance, I see no evidence that Intel is trying to keep AMD weak but still alive. AMD has pulled off amazing manuevers, which no one had any right to expect to succeed, to stay in the CPU business. If Intel's manuevers had worked, as they had every reason to think they would, AMD would be gone.
Intel does not own x86, although they invented it, because no one can own an instruction set, unless they have changed the law since the Zilog decision. There are lot of things people can't own, which the lawyers and lawmakers are now trying to make ownable, what they are calling intellectual property. When intellectual property (IP) was called "an idea" no one thought you could or should own an idea. No one owns "Calculus", without which science as we know it could not be carried out, and it would be unfortunate if they did.
AMD or Zilog or MOS or MS or Apple are not the only ones to ever try to horn in on somebody's business. People have done it in every field, not especially computers. The only thing unusual is the extent of protection the computer field tries to claim under the law.
I hate to be an AMD fanboy, which I am to some extent. I think Intel is a great company that has done some remarkable things. The Pentium was a milestone and the Itanic will probably be too once they get it all worked out. It always looks like folly when you reach like Intel is doing. Unfortunately the Pentium was, and the Itanic no doubt will be, out of my price league. Therefore a couple steps back is what I settle for, and AMD has always been a far better value there. To me that makes AMD a great company even if their best had never beat Intel's best. (But when AMD was making 486s, they smucked Intel and cost less.) AMDs chips have not been sub-standard in comparison to Intel. They are first quality, beautifully conceived and executed. I wish Intel designs were as elegant, which they seldom have been. To me, what AMD is doing with 64 bit is what Intel would have gotten done a few years ago if they were like the Intel of past years, and it would be going smoother if Intel had. You don't wait until everyone has 4G of memory before you make a CPU that can use it properly. Anyway, the historical point I guess I am trying to get across is that AMD is not some second rate, johnny-come-lately company that tries to imitate Intel and can't. Intel just has grown to very large proportions. And the whole Intel vs AMD thing is not even close to unique in computer history.
Yeah, I know it's too long.
in a comment at the news section that I decided to drop it into a thread, and hope to get more info.> an agreement ... allows AMD to produce x86 CPUs despite the fact that
> Intel holds the patent for this technology. Without it AMD wouldn't be
>able to produce x86 anything...
Intel and AMD have cross-licensing of patents, although that does not give them the right to copy the others design. They have to do their own original. They can use the patents in their design.
But it is has been perfectly legal, I think, for any company to make a work-alike of anybody else's CPU since Intel lost in court to Zilog. Zilog was a company formed mainly by former Intel high-ups to functionally clone the 8080 (Intel's 8 bit CPU), only do it right. The first Zilog Z80 ran twice as fast (yes, believe it or not) as the Intel 8080, which highly POed Intel. It ran all the same op-codes as the 8080, and a few extensions.
(No reason to be surprised Zilog could double the speed. Back then Intel did not believe in pushing the edge. I suppose that is what motivated the employees to leave and start Zilog. Intel started pushing the speed edge with the 486, and really changed philosopy with the Pentium.)
The original Cyrix (not the same-named entity at VIA) 6x86, for instance, was produced as a clean-room design (that is, without any knowledge of the Intel chip internals) without any permission from Intel. Intel sued Cyrix over and over about patents (although not for cloning x86), but never won and never stopped Cyrix. (You don't have to win to crush a smaller competitor with legal fees.) Naturally it is simpler legally, and worth some money to get Intel's permission, if they were inclined to give it, but it isn't necessary.
BTW, Cyrix counter-attacked Intel legally on patents, and got a settlement out of Intel. Through a set of ownership transfers of Cyrix, VIA ended up getting the right to make chips for the PII socket, although AMD could not.
When the LSI business was pretty new, it was a practical necessity to have a (legally binding) second-source agreement with another LSI company. Every LSI company paired themselves with another. It was considered business suicide to use a unique chip from an only-source, because you would be SOOL if there were the slightest hiccup in delivery. AMD and Intel used each other as a second source. As such, both had complete access to the other's original designs, including masks. In those days Intel was not a big deal (believe it or not) and AMD was doing them a bit of a favor, in return for which Intel was supposed to make AMD-designed chips, although Intel screwed AMD and didn't. For instance, AMD used to make something called bit-slice chips (something like 1/4 or 1/8 of a CPU) for minicomputers, and other specialty chips, such as a fast hardware binary multiplier, so AMD was not a leach on Intel or a slouch. 286s, 386s, and most 486s from AMD were minor reworks of Intel designs, and generally ran at higher clocks. By the 486, Intel wanted to get out of the contract and stop AMD. AMD won some and lost some in court, but ultimately started producing 486s of their own design. By the time of the Pentium, AMD was not doing well on the design front, and bought Nexgen, a company that already was producing almost-x86 chips based on a RISC back end, a style of design Intel adopted later with the Pentium Pro/II/III/4. Some K5's, the K6s, and the Athlons come from this.
The 8080 vs Z80 conflict was no more cut-throat than others. The original Apples from Apple used a 6502 (from MOS?) rather than the 6800 from Motorola. The 6502 was a near duplicate of the 6800, with some improvements. There was a 1 to 1 correspondence between the opcodes, IIRC. It is almost funny that Apple went to Motorola chips for the Lisa/Macintosh when Apple decided to copy the Xerox Star design. It is even funnier that for years Apple was suing MS for copying Apple's copy of the Star, until Apple fell on hard times and couldn't afford the legal fees. Apple was in a position like countless other little companies that Apple crushed for stepped on its toes, only Apple didn't settle with the little guys, just put them out of business. If Gates was like Jobs, Apple would be out of business. Jobs put Steve Wozniak, the actual sole designer of the original Apple and co-founder of Apple, out of business when Woz started a company essentially unrelated to Apple's product.
Now back to the patents. While AMD was developing the successor to the K6, Intel decided to terminate AMD with prejudice
My understanding is that AMD licensed the DEC EV bus for the Athlon because, unlike with the Pentium, Intel did not license the PII bus, so AMD couldn't use it. I don't know how that squares with cross-licensing. It doesn't. AMD licensed the DEC EV bus while DEC was on the ropes and about to expire. Not long after, in an odd turn of fate, DEC was picked up by Intel (I believe) when it finally gave up. IAC, it was believed that AMD could not survive without being plug-in compatible with an Intel chip, the reason being AMD would have to use a separate chipset and a mobo unique to AMD. What mobo company in its right mind would make a mobo only for AMD CPUs? None. What chipset maker would make a chipset only usable with AMD CPUs? None. So that was the end of AMDs x86 business.But AMD developed a chipset of its own and a mobo design that used the same parts, or readily adaptable, as Intel used in PII mobos. AMD gave away the chipset design to whomsoever would like to use it. And the total mobo design is such that a straight-forward revision changes a PII mobo into an Athlon mobo. When Intel reverted from a slot cartridge back to a socket, so did AMD. When Intel changed chip packages, so did AMD. It was so neat that practically every mobo maker eventually did an AMD mobo to pick up a few extra sales. That turned out to be fortuitous for them when Intel terminated the PII/PIII unexpectedly early, when Athlons seemed about leave them behind, but mobo makers could continue with the mobo design now using Athlons.
But that was later. Advance rumors and benchmarks of engineering samples had the Athlon FPU at a pathetic half the Pentium Pro/II, although the real thing turned out to have the most killer x86 FPU ever produced by anyone up to then. AMD had lined up several chipset makers for Athlon mobos, but only AMDs expensive original, meant as a prototype, was ready at the Athlon intro. VIA, which was later to make a fortune for years as the premier Athlon chipset maker, preferred to work on the Intel version, where the money was. Really good and competitive Athlon mobos took a while to arrive, as Taiwan manufacturers gradually realized the Athlon was going over, unexpected to them, or to anyone really, and there was money to be made.
I'm not sure what point I'm trying to make. Its not a total history. Just pieces where I see people have gotten a strange angle on it due to things not brought up much.
For instance, I see no evidence that Intel is trying to keep AMD weak but still alive. AMD has pulled off amazing manuevers, which no one had any right to expect to succeed, to stay in the CPU business. If Intel's manuevers had worked, as they had every reason to think they would, AMD would be gone.
Intel does not own x86, although they invented it, because no one can own an instruction set, unless they have changed the law since the Zilog decision. There are lot of things people can't own, which the lawyers and lawmakers are now trying to make ownable, what they are calling intellectual property. When intellectual property (IP) was called "an idea" no one thought you could or should own an idea. No one owns "Calculus", without which science as we know it could not be carried out, and it would be unfortunate if they did.
AMD or Zilog or MOS or MS or Apple are not the only ones to ever try to horn in on somebody's business. People have done it in every field, not especially computers. The only thing unusual is the extent of protection the computer field tries to claim under the law.
I hate to be an AMD fanboy, which I am to some extent. I think Intel is a great company that has done some remarkable things. The Pentium was a milestone and the Itanic will probably be too once they get it all worked out. It always looks like folly when you reach like Intel is doing. Unfortunately the Pentium was, and the Itanic no doubt will be, out of my price league. Therefore a couple steps back is what I settle for, and AMD has always been a far better value there. To me that makes AMD a great company even if their best had never beat Intel's best. (But when AMD was making 486s, they smucked Intel and cost less.) AMDs chips have not been sub-standard in comparison to Intel. They are first quality, beautifully conceived and executed. I wish Intel designs were as elegant, which they seldom have been. To me, what AMD is doing with 64 bit is what Intel would have gotten done a few years ago if they were like the Intel of past years, and it would be going smoother if Intel had. You don't wait until everyone has 4G of memory before you make a CPU that can use it properly. Anyway, the historical point I guess I am trying to get across is that AMD is not some second rate, johnny-come-lately company that tries to imitate Intel and can't. Intel just has grown to very large proportions. And the whole Intel vs AMD thing is not even close to unique in computer history.
Yeah, I know it's too long.
Facebook
Twitter