- Nov 23, 2001
- 6,712
- 142
- 106
why hasn't intel's x86 patent expired yet ?
- Thread starter Soulkeeper
- Start date
they already have? I guess anyone is free to build a 386 or even 486 compatible x86 CPU. Good luck selling that tho 
I've always wondered how a 486 would fare if built on a 32nm process so I kinda hope someone will do it
I've always wondered how a 486 would fare if built on a 32nm process so I kinda hope someone will do it
As the previous poster mentioned, the original patents have expired by now but aren't relevant as you can't build a modern processor that will actually run currently shipping code without access to the newer patents (i.e. SSE 1-4, x86-64), which are still valid.
- Nov 23, 2001
- 6,712
- 142
- 106
so to be technically correct it isn't the "x86 patent" at all, it's patents they have on the "extensions" to the x86 technology
so if amd just said take your SSE* and ... and brought back 3dnow
something tells me it can't be that simple tho
so if amd just said take your SSE* and ... and brought back 3dnow
something tells me it can't be that simple tho
It's a little more complicated. The original x86 CPU was only 8-bit, and you have tons of additional patents associated with each new generation of CPU to cover all the new tech introduced with them.
There's already enough market penetration of code that requires some version of SSE that producing an x86 chip without those extensions would be financially suicidal.
There's already enough market penetration of code that requires some version of SSE that producing an x86 chip without those extensions would be financially suicidal.
- Nov 23, 2001
- 6,712
- 142
- 106
Originally posted by: jones377
I've always wondered how a 486 would fare if built on a 32nm process so I kinda hope someone will do it
Larrabee?
Originally posted by: Idontcare
Originally posted by: jones377
I've always wondered how a 486 would fare if built on a 32nm process so I kinda hope someone will do it
Larrabee?
Not quite what I had in mind.
no, that is wrong, the original x86 cpu (8086) was a 16-bit CPU, a cheaper version the 8088 was released which was a 16-bit cpu with an 8-bit MEMORY bus (much the the later 386SX was a 32-bit cpu with a 16-bit memory bus).
Hmmm provided you didn't need any special extentions (SSE, 3DNOW, X86-64 etc) you could probably still build a 32-bit x86 CPU that might even be able to be usable with a lot of today's software...no sure how the performance would be however due to the lack of extentions.
Originally posted by: jones377
Originally posted by: Idontcare
Originally posted by: jones377
I've always wondered how a 486 would fare if built on a 32nm process so I kinda hope someone will do it
Larrabee?
Not quite what I had in mind.
IIRC the socket 7 IDT Winchip's where based around a modified 486 architecture, while they were crap at 3D gaming (the FPU was even slower than the Cyrix M1 (6x86) and M2 (6x86MX) based CPU's), they where pretty competitive with the Pentium MMX and early K6 as far as windows performance went.
Nemesis 1
Lifer
- Dec 30, 2006
- 11,366
- 2
- 0
You know this just doesn't seem right. I can't think of one good reason intel is doing this . That has any sound logic to it. We already know that Intel is moving towards Sandy Bridge. @ 32nm. Sandy is a strange beast. It rather confuounded me infact.
Intel went to a new processor formate with AVX on sandy 4 operand core 3+1= 4 the 1 is for FMA.. FMA will not be active on sandy. Were x86 is ported to sandy via intel compiler. What hit me was this isn't the natural evolution of x86 this is something else.
The problem I am having is AMDs Bulldoozer will do FMA. This will be a 3 operand core. Amd has said only certain Risk processors will do FMA. So This allows Intel to 2x anything that can be vectored.
So Sandy forces AMD to Come up with FMA for X86 + Engineer that very special X86 risk backend. So Intel is in fact moving away from X86 to Vector computing with X86 ported.
So this lawsuite to me looks like Intel wants to open the X86 market to all. Taken AMD to court might allow this. In the mean time AMD has to come up with FMA for X86 . Which intel can use. When Intel loses in court. All X86 will become public domain. Bring in More competion to drive the Market. All the While Intel sets with its new vectoring processor thats their IP. Able to use everthing everone comes up with and still have a priority cpu with AVX. Than Add in FMA and intel is set to roll.
Its Intels best interest to allow all to make x86 cpus now. The more the better.
This is what intel is really after. AMD going to have it all. Problem is . So well everone else. Intel is clever. This time around. A fox they are but which type of fox.
Intel went to a new processor formate with AVX on sandy 4 operand core 3+1= 4 the 1 is for FMA.. FMA will not be active on sandy. Were x86 is ported to sandy via intel compiler. What hit me was this isn't the natural evolution of x86 this is something else.
The problem I am having is AMDs Bulldoozer will do FMA. This will be a 3 operand core. Amd has said only certain Risk processors will do FMA. So This allows Intel to 2x anything that can be vectored.
So Sandy forces AMD to Come up with FMA for X86 + Engineer that very special X86 risk backend. So Intel is in fact moving away from X86 to Vector computing with X86 ported.
So this lawsuite to me looks like Intel wants to open the X86 market to all. Taken AMD to court might allow this. In the mean time AMD has to come up with FMA for X86 . Which intel can use. When Intel loses in court. All X86 will become public domain. Bring in More competion to drive the Market. All the While Intel sets with its new vectoring processor thats their IP. Able to use everthing everone comes up with and still have a priority cpu with AVX. Than Add in FMA and intel is set to roll.
Its Intels best interest to allow all to make x86 cpus now. The more the better.
This is what intel is really after. AMD going to have it all. Problem is . So well everone else. Intel is clever. This time around. A fox they are but which type of fox.
You're saying that Intel wants to open the X86 market to all, so it decides the best way is to sue the biggest licensee, and after losing in the court and paying the costly trial bills, x86 will become public domain -> goal achieved? They own the IP rights, why not just make it public domain if they want that?
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
Hm. Even earlier than I thought. But, smaller = more per wafer. So thats good.
Nemisis: Wut? For what possible purpose would Intel want to invite nV into the x86 market? And why would them losing the case "AMD can continue making x86 chips" open the market to anybody else?
It's still a multi-billion dollar capital investment to make a CPU fabrication plant, then as mentioned there's no guarantee your product will sell.
Are instruction sets patent-able? I did a patent search and didn't find very much in the way of instruction set patents. So the original question about the x86 patent, is there such a thing? I would think that since an opcode is essentially a mathematical algorithm, it wouldn't be covered by a patent...?
* Not speaking for Intel Corp.
* Not speaking for Intel Corp.
DrMrLordX
Lifer
- Apr 27, 2000
- 21,637
- 10,855
- 136
The short answer is that you can patent just about anything if the USPTO lets you get away with it and if your lawyers can stand up to challenges to your patents down the road.
I am not prithee to Intel's patent portfolio (or that of anyone else, really), but I suspect that the instruction sets themselves are nested in patents covering hardware implementations thereof (or vice versa: patents covering instruction extensions are nested in vague patents covering theoretical hardware implementations thereof). Probably. Observe this patent of AMD's:
http://www.patentstorm.us/patents/6014739.html
The patent covers an x86 instruction extension (x86-64 I think), but if you observe the claims and full text, you'll see that they go into more than just the instruction set.
In the end it's all lawyer-speak and it's pretty messed up. Also, it is interesting to see that this patent was filed back in 1997!
Engineers avoid reading patents for lawsuit reasons. If you'd said it was an Intel patent up front that may have helped.
My understanding is that patents cover implementation of the instruction set. The claims in the AMD patent DrMrLordX linked are almost all "A microprocessor comprising: whatever". From the AMD patent:
Broadly speaking, the present invention contemplates a microprocessor comprising a register file and an instruction decode circuit. The register file includes a standard register set comprising a plurality of standard registers and an extension register set comprising a plurality of extension registers.
It's all about the processor that has the extra registers and the new fancy instruction decoder, not the actual new registers.
For what it's worth, the "background of the invention" and "summary of the invention" sections of AMD patents (and probably others) are usually pretty readable if you ignore the overuse of "plurality" and "comprising" / "comprehending" / "contemplating"
. It's mainly the "claims" section that is unreadable legalese.I'm not a lawyer, and I'm not speaking for any company.
Nice link...I counted (BOE) 8100+ AMD patents there, and 8800+ Intel Patents.
It's easy to see why the 2 companies have become so dependent on each other for technological advancement, and why the "nuclear option" of either company pulling the plug on licenses will never come to fruition.
senseamp
Lifer
- Feb 5, 2006
- 35,787
- 6,195
- 126
Maybe someone can explain to me what the big deal about FMA is?
http://en.wikipedia.org/wiki/Multiply-accumulate
It's been around on CPUs for almost 20 years. I mean it's new to X86, but pretty much every other major ISA out there has had it for a very long time, even Intel's own Itanium.
http://en.wikipedia.org/wiki/Multiply-accumulate
It's been around on CPUs for almost 20 years. I mean it's new to X86, but pretty much every other major ISA out there has had it for a very long time, even Intel's own Itanium.
TRENDING THREADS
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 10K
-
Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)- Started by Vattila
- Replies: 13K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
Facebook
Twitter