News Nvidia Client Arm CPUs Are Not an Immediate Threat Claims Intel CEO

[TerryLaze]

Yeah... about that... I'm going to go lazy since apparently we're using wikipedia for sourcing so...

Intel has an architectural license just like Arm does, but they absolutely will not issue it anymore.

But also on the top and right of the page you can find this.
Open partly.... you need licenses for the advanced parts, as I said to make something that would be useful today you would have to buy a design from intel or amd.
Also :The i686 subset of the x86 architecture is therefore fully open.

Open	Partly. For some advanced features, x86 may require license from Intel; x86-64 may require an additional license from AMD. The Pentium Pro processor (and NetBurst) has been on the market for more than 21 years[1] and so cannot be subject to patent claims. The i686 subset of the x86 architecture is therefore fully open.

 

[WebDevDeadDrop]

But also on the top and right of the page you can find this.
Open partly.... you need licenses for the advanced parts, as I said to make something that would be useful today you would have to buy a design from intel or amd.
Also :The i686 subset of the x86 architecture is therefore fully open.

I could be wrong, but the last I heard ISA's are protected by copywrite and individual implementations in hardware are only protected by patent. So you can build a Pentium Pro, but you can't just build a x86 processor that uses the ISA features available on the Pentium Pro in custom silicon.

 

[TerryLaze]

I could be wrong, but the last I heard ISA's are protected by copywrite and individual implementations in hardware are only protected by patent. So you can build a Pentium Pro, but you can't just build a x86 processor that uses the ISA features available on the Pentium Pro in custom silicon.

So you can copy the copyrighted ISA verbatim you just can't copy it if you change it?! How does that make sense, either you can copy the ISA or you can't.
Also if anybody would hold the copyright of the x86 ISA then they would not allow it to be copied so you wouldn't be able to copy the patent because the copyright would be a sub part of the patent.
Copyright is active for 70 years instead of the 20-21 years they state for the 686...so you wouldn't be able to make one for another 50 years.

 

[WebDevDeadDrop]

So you can copy the copyrighted ISA verbatim you just can't copy it if you change it?! How does that make sense, either you can copy the ISA or you can't.
Also if anybody would hold the copyright of the x86 ISA then they would not allow it to be copied so you wouldn't be able to copy the patent because the copyright would be a sub part of the patent.
Copyright is active for 70 years instead of the 20-21 years they state for the 686...so you wouldn't be able to make one for another 50 years.

You're in the wrong head space if you think copyright law and patent law are in a state where they make 'sense'. That is the current state of affairs though. You can in fact make a CPU exactly as it was that's out of patent, but you can't make changes to it that are still covered by other patents or copyright.

And frankly, there's a non-zero chance you could be sued for programing the decoder on the CPU if you used an exact copy of the Pentium Pro's decoder code because that's going to be covered by copyright as well. I'm not sure how well that would hold up in court, but software copyright is a thing.

 

[thestryker]

They still would not be able to do that if they would step on actual licenses.

What licenses? FPGAs just run the design and no product is being sold. Not to mention what they're running are relatively ancient designs which aren't likely covered with patent protection by the time the projects started.

Yes, they bought x86 CPU core designs and are legally able to produce them, that's what I said.
Maybe there is a language barrier or something but what do you think a x86 license is? would be?

Do you know what an architecture license is? It sure seems you don't have a clue at all here. AMD got their start in PCs cloning Intel chips and after the first one entered into a licensing agreement for microcode. AMD only had one period of time where they had licensed actual CPUs and that was because of IBM. I believe their last clone was in the 386 era and everything from then on was custom designs based on technology sharing. The technology sharing agreements all come with effectively refusal rights so if AMD were to be sold for example and Intel was not okay with it for whatever reason they could kill the agreement. Intel hasn't done any CPU design sharing after the initial IBM deal they were forced to take.

The history of x86 clones is relatively wild and involved a lot of reverse engineering. Transmeta for example never had a license because they never used the instruction sets just translated them. Cyrix reverse engineered their first designs, but didn't use any of the patented technologies and this whole saga ended in a ton of lawsuits and cross licensing.

But also on the top and right of the page you can find this.
Open partly.... you need licenses for the advanced parts, as I said to make something that would be useful today you would have to buy a design from intel or amd.
Also :The i686 subset of the x86 architecture is therefore fully open.

Open	Partly. For some advanced features, x86 may require license from Intel; x86-64 may require an additional license from AMD. The Pentium Pro processor (and NetBurst) has been on the market for more than 21 years[1] and so cannot be subject to patent claims. The i686 subset of the x86 architecture is therefore fully open.

This is a very incomplete way of looking at it that shows a complete lack of understanding. Licensing a design is the only effective way to make an x86 CPU today, but that's because Intel won't license the architecture. AMD and Via have sold designs fairly recently, but Intel hasn't since they were forced to. Outside of the technology agreements with AMD Intel has basically been forced into agreements (due to lawsuits) with the other companies who have them (well mostly had now).

If Intel wanted to they could package and license their x86 patents like Arm does with their architecture licenses. The only thing stopping them is that they very much don't want to.

The only thing I've not seen explained is how Centaur made x86-64 compatible CPUs, but my assumption is that due to the Intel/AMD cross licensing this was included in the Intel/Via agreement.

 

[bit_user]

what do you think a x86 license is? would be?

They still would not be able to do that if they would step on actual licenses.

I believe it boils down to some patents that are needed to make a performant implementation of the x86 ISA. Perhaps using a FPGA can circumvent many of those (you can't patent an ISA - only the fundamental methods for implementing parts of it), or perhaps Intel didn't deem it a sufficient competitive threat to pursue.

I could be wrong, but the last I heard ISA's are protected by copywrite

I've never heard of such a case. Until Oracle vs. Google, it was widely believed you couldn't even copyright an API. I've still never heard of an ISA being copyrighted, though (or, at least, such a copyright being successfully enforced).

And frankly, there's a non-zero chance you could be sued for programing the decoder on the CPU if you used an exact copy of the Pentium Pro's decoder code because that's going to be covered by copyright as well. I'm not sure how well that would hold up in court, but software copyright is a thing.

Oh, if you got a copy of the hardware design of the Pentium Pro and either synthesized it verbatim or even created a "derivative work", that's definitely a copyright violation.

 

[bit_user]

What licenses? FPGAs just run the design and no product is being sold.

You can sell the source code of a "soft core" which runs on a FPGA. This happens all the time. Or, you can sell cards or equipment which contains FPGAs that are running licensed IP. Either way, FPGAs aren't some magic loophole. Oh, and by the way, you don't need to sell something for patent infringement to occur.

When I was talking about a x86 "soft core" running on a FPGA circumventing certain patents, that's simply because the most natural way to implement certain parts of the ISA on a FPGA might differ enough from custom-designed silicon that the patents wouldn't apply. That's my theory, at least.

Not to mention what they're running are relatively ancient designs which aren't likely covered with patent protection by the time the projects started.

Or that.

 

[thestryker]

You can sell the source code of a "soft core" which runs on a FPGA. This happens all the time. Or, you can sell cards or equipment which contains FPGAs that are running licensed IP. Either way, FPGAs aren't some magic loophole. Oh, and by the way, you don't need to sell something for patent infringement to occur.

When I was talking about a x86 "soft core" running on a FPGA circumventing certain patents, that's simply because the most natural way to implement certain parts of the ISA on a FPGA might differ enough from custom-designed silicon that the patents wouldn't apply. That's my theory, at least.

Yeah I was watching football and making breakfast and thus failed to put the word "compatible" in there. Near as I could tell (the projects are all around a decade old or more) they're all custom designs intended to run the same instructions. The instructions would have been covered by patents, but those should have long expired.