Those are the kinds of things I would think AMD would want to be going for by bargaining that they don't need to negotiate with Intel over giving them prolonged access to amd64 instruction set and that they'd be fine with ARM. But knowing that AMD has a perpetual x86 license is great news which gives them even more bargaining power. I'm not a lawyer or anything but I kind of understand the situation. But I am somewhat aware of the fact that designing an architecture isn't just about making the fastest thing you can, it's about not having to pay ridiculous fees to license things from your competitors.
You have it backwards. The AMD64 instruction set is AMD's baby, not Intel's. Intel had to pay AMD for the licence, not the other way around. They had to do it because back in the day, the AMD64 Athlons were pummeling the Pentium-4's into the pavement.
In response, AMD accused Intel of breaching the patent cross licensing as well, and went to invalidate Intel x86-64 (AMD64) license.
Then both AMD and Intel signed a new x86 cross-licensing agreement in the 2009 settlement, which was extended to GF, and that finishes the day 12 of Nov of 2014.
It is more about Project CARS' developers complaining because they have to change their way to make things. Instead relying in a single powerful core and ignore the rest of the hardware, now they have to develop a wide engine that uses all the available cores. They also have to learn to use the new memory model and to offload to the GPU some compute tasks.
In response, AMD accused Intel of breaching the patent cross licensing as well, and went to invalidate Intel x86-64 (AMD64) license.
Then both AMD and Intel signed a new x86 cross-licensing agreement in the 2009 settlement, which was extended to GF, and that finishes the day 12 of Nov of 2014.
Yeah umm no.
AMD's perpetual x86 ISA license was given to them by a CA court judge, Intel can't terminate it. That was Intel trying to buffalo AMD and get them into another expensive protracted legal battle when their financial resources were strained. They would of went to court over the issue and there is little doubt that AMD's license would of been validated. GF doesn't design CPU's, they just manufacture them. AMD is who design's the CPU and thus use's the licensed x86 instruction set, they never transferred it.
Really Juan your reaching here and punching well above your weight class.
For your own education, here is some background on the numerous legal battles between AMD and Intel.
By 1994, Intel had navigated the arbitration along its long journey up to the Supreme Court of California. The Supreme Court awarded AMD “a permanent, nonexclusive and royalty-free license to any Intel intellectual property embodied in the Am386” and “a two-year extension of certain patent and copyright licenses . . . related to the Am386.” AMD v. Intel, 885 P.2d at 998–99.
AMD has the rights to anything that existed inside the 80386 to and that includes the 32-bit x86 ISA. Intel can not, under any condition, terminate or change it. What is expiring is the cross licensing agreement that allowed each of them to use specific processor related IP from each other. This is AMD64 / EMT64 for Intel and SSE / AVX for AMD.
Also before you type anything else in, remember the western legal system is based heavily on precedents. The prior legal battles have made it clear that AMD has a right to design and manufacture CPU's compatible with Intel's x86 instruction set. This right has been established to be wholly independent of Intel, meaning Intel doesn't have to agree to anything and AMD would still maintain it's right.
BTW there is one other company that has the right to produce x86 CPU's that are wholly independent of Intel and that's Via. Cyrix successfully clean room reverse engineered several Intel products and thus developed their own x86 compatible ISA. Via purchased Cyrix from National Semiconductor and part of that purchase was those designs which enabled Via to win court cases against Intel for their right to design x86 "compatible" CPU's. They eventually entered into a similiar license agree with Intel that AMD did where they got access to the Intel developed extended instruction sets. That agreement should be expiring here shortly (can't remember if it was 2013 or 2014), not sure if Via will pursue a different part or not. Regardless of what Intel wants Via could still design CPU's that are "compatible" with the x86 ISA.
When 100 million Americans are living below or at the poverty line
big core and big GPU is not the money bringer
it is simple but effective designs like APU that will bring home the bacon
That is why Intel is frantically trying to catch up with their IGP
X86 AMD should be around Sandy with Kaveri or better 10%-15% slower than latest Intel i5 chip
After Excavator the real deal starts 6core(18core under HSA) 8 core(24core under HSA) APu
will be on the seen .
that's my opinion
but Mantle, HSA, those are well ahead of Intel's grasp ATM
and those will be a driving force going forward
Mantle is said to provide 20 - 50% improvement
HSA offers bigger gains
Excavator will offer full context switching
AMD is far from out and can easily gain 10- 15% of market from Intel should at ces14 they display a Kaveri with all 12 cores humming under HSA it will help their cause immensely
Steamroller should be a nice bump up over Piledriver we will soon find out
The x86-32 was patented around 1985 and lasts 20 years. Anyone can make an x86-32 cpu but they cant include sse, avx, or fma without buying the rights. Thats what the cross license is for. This gives intel access to AMD's patents such as AMD-64 extension, hence cross-license. Originally AMD had to pay intel somewhere around 8% royalty to intel for every cpu sold. Intel took that money to block sales which led to the lawsuit.
What Paladin said sounds exactly like what my bud who worked over at AMD told me (Now Intel Eng). So, You can't deny the facts. It's all just a crazy circle though. I guess it's better that we discuss this instead of ARM. lol
AMD's perpetual x86 ISA license was given to them by a CA court judge, Intel can't terminate it. That was Intel trying to buffalo AMD and get them into another expensive protracted legal battle when their financial resources were strained. They would of went to court over the issue and there is little doubt that AMD's license would of been validated. GF doesn't design CPU's, they just manufacture them. AMD is who design's the CPU and thus use's the licensed x86 instruction set, they never transferred it.
Really Juan your reaching here and punching well above your weight class.
For your own education, here is some background on the numerous legal battles between AMD and Intel.
By 1994, Intel had navigated the arbitration along its long journey up to the Supreme Court of California. The Supreme Court awarded AMD “a permanent, nonexclusive and royalty-free license to any Intel intellectual property embodied in the Am386” and “a two-year extension of certain patent and copyright licenses . . . related to the Am386.” AMD v. Intel, 885 P.2d at 998–99.
Nice to know that you are not only an 'expert' engineer but also an 'expert' in the law. Now continue reading the paragraph at the point where you stopped:
However, this result left several outstanding lawsuits unresolved, including a 1991 antitrust suit that AMD had brought against Intel in the U.S. District Court of the Northern District of California. The two companies had battled each other in court for seven years and had spent over $100 million when they finally settled all remaining lawsuits in 1995. AMD received a perpetual license to the microcode found in Intel’s 386 and 486, but also agreed to not copy any other Intel microcode. Instead, AMD would develop its own chips in the future.
AMD received a perpetual license to 386/486 microcode, but microcode != x86 ISA.
The "permanent, royalty-free license to all Intel patents in the Am386" sentence that you quote was eliminated after Intel appeal:
In paragraph 5 of the award, the arbitrator granted AMD a permanent, royalty-free license to all Intel patents, copyrights, and other intellectual property used in the Am386. The rights granted to AMD in paragraph 5 included both the right to make the Am386 and the right to have others make the Am386 for AMD. In paragraph 6, the arbitrator extended for two additional years, with respect to the Am386 only, a 1976 grant of certain patent and copyright licenses from Intel to AMD that the 1982 AMD-Intel agreement had extended until 1995. (Because paragraph 5 grants AMD an indefinite license to all Intel patents and copyrights used in the Am386, it is not immediately apparent what additional rights AMD gained by paragraph 6.)
After the arbitrator made his award, AMD petitioned the superior court to confirm the award and Intel petitioned to have the award corrected by deleting paragraphs 5 and 6. The court confirmed the award as made.
Intel appealed. On appeal, the Court of Appeal concluded that the arbitrator had exceeded his powers in fashioning the relief in paragraphs 5 and 6 because those paragraphs did not draw their essence from the AMD-Intel contract and lacked any rational nexus to it. The Court of Appeal corrected [9 Cal.4th 394] the award by deleting paragraphs 5 and 6 and confirmed the award as corrected, determining that it was unnecessary to vacate the award because the correction did not affect the merits of the decision.
AMD had to re-negotiate the x86 license. The famous x86 cross-licensing settlement was signed in 2001. AMD got x86-32 license plus some related licenses and Intel got x86-64 license plus some related licenses. This settlement was broken in early 2009
AMD knows very well that Intel holds the x86 license and their official response to Intel was:
Intel's action is an attempt to distract the world from the global antitrust scrutiny it faces. Should this matter proceed to litigation, we will prove that Intel fabricated this claim to interfere with our commercial relationships and thus has violated the cross-license.
AMD remains in full compliance with the cross-license agreement. And as we've stated all along, the structure of GLOBALFOUNDRIES takes into account all our cross-license agreements. We will continue to respect Intel's intellectual property rights, just as we expect them to respect ours.
Again, we believe that Intel manufactured this diversion as an attempt to distract attention from the increasing number of antitrust rulings against it around the world. With a ruling from the European Commission and a U.S. trial date looming, and investigations by the U.S. FTC and NY Attorney General, the clock is ticking on Intel's illegal practices - and yet with its dominant monopoly position it still tries to stifle competitors.
The AMD/Intel cross-license agreement is a two-way agreement, the benefits of which go to both companies. Intel leverages innovative AMD IP critical for its product designs under the cross license. This includes AMD patents related to 64-bit architecture extensions, integrated memory controller, multi-core architecture, etc.). The cross-license is very much a two-way street.
In fact, we informed Intel that their attempt to terminate AMD's license itself constitutes a breach of the cross-license agreement, which, if uncured, gives AMD the right to terminate Intel's license
AMD's perpetual x86 ISA license was given to them by a CA court judge, Intel can't terminate it. That was Intel trying to buffalo AMD and get them into another expensive protracted legal battle when their financial resources were strained. They would of went to court over the issue and there is little doubt that AMD's license would of been validated. GF doesn't design CPU's, they just manufacture them. AMD is who design's the CPU and thus use's the licensed x86 instruction set, they never transferred it.
Really Juan your reaching here and punching well above your weight class.
For your own education, here is some background on the numerous legal battles between AMD and Intel.
By 1994, Intel had navigated the arbitration along its long journey up to the Supreme Court of California. The Supreme Court awarded AMD “a permanent, nonexclusive and royalty-free license to any Intel intellectual property embodied in the Am386” and “a two-year extension of certain patent and copyright licenses . . . related to the Am386.” AMD v. Intel, 885 P.2d at 998–99.
Nice to know that you are not only an 'expert' engineer but also an 'expert' in the law. Now continue reading the paragraph at the point where you stopped:
However, this result left several outstanding lawsuits unresolved, including a 1991 antitrust suit that AMD had brought against Intel in the U.S. District Court of the Northern District of California. The two companies had battled each other in court for seven years and had spent over $100 million when they finally settled all remaining lawsuits in 1995. AMD received a perpetual license to the microcode found in Intel’s 386 and 486, but also agreed to not copy any other Intel microcode. Instead, AMD would develop its own chips in the future.
AMD received a perpetual license to 386/486 microcode, but microcode != x86 ISA.
The "permanent, royalty-free license to all Intel patents in the Am386" sentence that you quote was eliminated after Intel appeal:
In paragraph 5 of the award, the arbitrator granted AMD a permanent, royalty-free license to all Intel patents, copyrights, and other intellectual property used in the Am386. The rights granted to AMD in paragraph 5 included both the right to make the Am386 and the right to have others make the Am386 for AMD. In paragraph 6, the arbitrator extended for two additional years, with respect to the Am386 only, a 1976 grant of certain patent and copyright licenses from Intel to AMD that the 1982 AMD-Intel agreement had extended until 1995. (Because paragraph 5 grants AMD an indefinite license to all Intel patents and copyrights used in the Am386, it is not immediately apparent what additional rights AMD gained by paragraph 6.)
After the arbitrator made his award, AMD petitioned the superior court to confirm the award and Intel petitioned to have the award corrected by deleting paragraphs 5 and 6. The court confirmed the award as made.
Intel appealed. On appeal, the Court of Appeal concluded that the arbitrator had exceeded his powers in fashioning the relief in paragraphs 5 and 6 because those paragraphs did not draw their essence from the AMD-Intel contract and lacked any rational nexus to it. The Court of Appeal corrected [9 Cal.4th 394] the award by deleting paragraphs 5 and 6 and confirmed the award as corrected, determining that it was unnecessary to vacate the award because the correction did not affect the merits of the decision.
AMD had to re-negotiate the x86 license. The famous x86 cross-licensing settlement was signed in 2001. AMD got x86-32 license plus some related licenses and Intel got x86-64 license plus some related licenses. This settlement was broken in early 2009
AMD knows very well that Intel holds the x86 license and their official response to Intel was:
Intel's action is an attempt to distract the world from the global antitrust scrutiny it faces. Should this matter proceed to litigation, we will prove that Intel fabricated this claim to interfere with our commercial relationships and thus has violated the cross-license.
AMD remains in full compliance with the cross-license agreement. And as we've stated all along, the structure of GLOBALFOUNDRIES takes into account all our cross-license agreements. We will continue to respect Intel's intellectual property rights, just as we expect them to respect ours.
Again, we believe that Intel manufactured this diversion as an attempt to distract attention from the increasing number of antitrust rulings against it around the world. With a ruling from the European Commission and a U.S. trial date looming, and investigations by the U.S. FTC and NY Attorney General, the clock is ticking on Intel's illegal practices - and yet with its dominant monopoly position it still tries to stifle competitors.
The AMD/Intel cross-license agreement is a two-way agreement, the benefits of which go to both companies. Intel leverages innovative AMD IP critical for its product designs under the cross license. This includes AMD patents related to 64-bit architecture extensions, integrated memory controller, multi-core architecture, etc.). The cross-license is very much a two-way street.
In fact, we informed Intel that their attempt to terminate AMD's license itself constitutes a breach of the cross-license agreement, which, if uncured, gives AMD the right to terminate Intel's license
AMD and Intel signed a new x86 cross-licensing settlement. It finishes the day 12 Nov 2014.
Not only the x86 ISA is a complete mess, but the x86 licenses/patents are a complete pain.
By all that AMD is migrating to ARM. Like it or no :lol:
@juanrga:
Evidently you are expert at neither of those fields:
Microcode is the x86 ISA. It's the instruction sets the microcode contains that define the ISA.
The paragraphs you cite are the excerpts from the portion where Intel appealed because the original ruling gave AMD access to all Intel proprietary licensing, and the rights to reverse engineer. The right to clean room reverse engineer was eventually rescinded by the appellate courts. So AMD cannot reverse engineer Intel hardware anymore...as the judge who ruled in the initial ruling judged that the IBM contract between AMD and Intel gave AMD rights to reproduce all Intel hardware as if they were Intel. The appellate court (on the second appeal mind you...the first appeal was shot down) came in and said that the ruling was overstepping boundaries and the original agreement with IBM did not merit such considerations for AMD. They ruled that AMD had to develop proprietary hardware of their own, but had access to the original x86-32 microcode a.k.a. the x86 ISA.
AMD is not moving to ARM anytime soon, and x86 is not going anywhere out of the consumer sector anytime soon either.
As for Intel threatening AMD:
The only threat that Intel can leverage is the use of extensions newer than those AMD already has a license for, or in layman's terms, extensions after AVX2. In order to compete AMD would need to be able to access those extensions so they can continue to run compatible code. However, AMD can produce x86-64 ISA CPUs that use up to AVX2 instruction sets indefinitely. Intel was trolling AMD, and if they breach the agreement, AMD can literally pull their license for x86-64 to Intel and instantly render all their future products incompatible with 64 bit instruction sets. In this instance, AMD literally has Intel over a barrel, and Intel is aware of it, but they are trolls anyway..
Reread all of what you posted in it's original context.
Also, might want to google processor microcode wikipedia entries, or whatever else you can find. Basically, x86 ISA without microcode is nothing. The ISA does not define hardware designs, rather, it defines what instruction set compatabilities the hardware must have to conform to the ISA. This is why AMD and Intel hardware can be so drastically different and still function as x86 CPUs, because the hardware itself is irrelevant as long as you can produce the desired outcome from the instructions processed.
In fact, GCN can actually run up to SSE4 x86-64 code if you want to get technical about it. If you think AMD is concerned enough about their licensing to switch to ARM entirely, then why did they make even their GPUs capable of running x86 instruction sets? Answer: because they're not concerned at all.
Also, might want to read up on ASM, because it is the most basic use of those microcode instructions we are discussing, and really opens up how everything "works"...
Now that we've established you're reaching desperately, please...go back to speculating about Kaveri, as your "ARM is going to rule the world" nonsense wore out it's welcome long, Long, LONG, ago...
Pretty much, this is how Cyrix was able to produce x86 "compatible" CPU's without requiring a license from Intel. They clean room reverse engineered Intel CPU's and developed their own microcode that was compatible with the x86 ISA. Intel tried to sue them into oblivion but the court held that it was independently developed work and didn't copy any of IP of Intel's. So Intel got really evil and tried to claim that the socket interface and data bus where theirs and only they could authorize manufacturers to produce compatible sockets. That didn't go over well since they had previous made those available to third parties in an effort to promote their own products. So next generation they went to a propriety slot design with a propriety bus interface and specified that it could only be made to support Intel branded CPUs. AMD was forced to continue on the previous socket 7 interface and thus super socket 7 was born. Eventually AMD went to their own slot and data bus design which was revolutionary for it's era as it used DDR technology (EV6 they got from DEC alpha).
So now the only thing Intel and AMD have similar is external x86 ISA compatibility. Everything else on their products is completely different.
I think the real question is if Juaranga is sent here to forum slide. He is constantly doing it.
That HSA video @ S|A forums highlights exactly what I've been saying all along, AMD needs a platform which allows multiple chips to communicate with each other over something that has more bandwidth than PCIe.
The entire slide just flat out came out and said they were using infiniband, exactly like I was hinting at, and Sernox has also had my back on this before by mentioning that it's possible to do Hypertransport over Infiniband.
http://insidehpc.com/2013/12/22/hsas-clusters/
This video (from S|A forums) blatantly states that AMD is actually going to do everything I was saying or at least something similar
1. Drop focus on servers and instead move to HPC and workstations (which is the goal of this entire presentation)
2. Create a way for dCPU, dGPU (AMD mentioned multiple APUs but there's no reason why that can't be big dCPU + dGPU)
3. Cease focusing on big x86 cores for servers and migrate to something that allows for HPC to migrate to desktop.
2015 we will be on a new high end AMD platform with infiniband and we will probably be able to add additional APUs, CPUs, and GPUs into expansion slots.
ARM is good enough in servers. It's not good enough for HPC, workstations, or even general purpose x86 computing (excluding workflows that exist only in browser).
As I've been saying. Juaranga has confused "we don't think x86 big cores are the right job for servers" coming from a company that has hit rock bottom for CPU market share in servers, with "we're abandoning our failure of a plan of releasing x86 server CPUs for 4% of the market and then jamming that architecture into desktop and mobile".
IT should be obvious why this isn't on any roadmaps yet. It's still more than likely being worked on and it would make Intel's CPU + PCIe slots for expansion to add sound cards and video cards platform look archaic.
I think I will enjoy reading this thread in a few years, much like how I like to stalk seronx's posts from years ago with everyone going "lol he's crazy xD WHERE DOES HE GET THIS" and then him being 80% right.
AMD's perpetual x86 ISA license was given to them by a CA court judge, Intel can't terminate it. That was Intel trying to buffalo AMD and get them into another expensive protracted legal battle when their financial resources were strained. They would of went to court over the issue and there is little doubt that AMD's license would of been validated. GF doesn't design CPU's, they just manufacture them. AMD is who design's the CPU and thus use's the licensed x86 instruction set, they never transferred it.
Really Juan your reaching here and punching well above your weight class.
For your own education, here is some background on the numerous legal battles between AMD and Intel.
By 1994, Intel had navigated the arbitration along its long journey up to the Supreme Court of California. The Supreme Court awarded AMD “a permanent, nonexclusive and royalty-free license to any Intel intellectual property embodied in the Am386” and “a two-year extension of certain patent and copyright licenses . . . related to the Am386.” AMD v. Intel, 885 P.2d at 998–99.
Nice to know that you are not only an 'expert' engineer but also an 'expert' in the law. Now continue reading the paragraph at the point where you stopped:
However, this result left several outstanding lawsuits unresolved, including a 1991 antitrust suit that AMD had brought against Intel in the U.S. District Court of the Northern District of California. The two companies had battled each other in court for seven years and had spent over $100 million when they finally settled all remaining lawsuits in 1995. AMD received a perpetual license to the microcode found in Intel’s 386 and 486, but also agreed to not copy any other Intel microcode. Instead, AMD would develop its own chips in the future.
AMD received a perpetual license to 386/486 microcode, but microcode != x86 ISA.
The "permanent, royalty-free license to all Intel patents in the Am386" sentence that you quote was eliminated after Intel appeal:
In paragraph 5 of the award, the arbitrator granted AMD a permanent, royalty-free license to all Intel patents, copyrights, and other intellectual property used in the Am386. The rights granted to AMD in paragraph 5 included both the right to make the Am386 and the right to have others make the Am386 for AMD. In paragraph 6, the arbitrator extended for two additional years, with respect to the Am386 only, a 1976 grant of certain patent and copyright licenses from Intel to AMD that the 1982 AMD-Intel agreement had extended until 1995. (Because paragraph 5 grants AMD an indefinite license to all Intel patents and copyrights used in the Am386, it is not immediately apparent what additional rights AMD gained by paragraph 6.)
After the arbitrator made his award, AMD petitioned the superior court to confirm the award and Intel petitioned to have the award corrected by deleting paragraphs 5 and 6. The court confirmed the award as made.
Intel appealed. On appeal, the Court of Appeal concluded that the arbitrator had exceeded his powers in fashioning the relief in paragraphs 5 and 6 because those paragraphs did not draw their essence from the AMD-Intel contract and lacked any rational nexus to it. The Court of Appeal corrected [9 Cal.4th 394] the award by deleting paragraphs 5 and 6 and confirmed the award as corrected, determining that it was unnecessary to vacate the award because the correction did not affect the merits of the decision.
AMD had to re-negotiate the x86 license. The famous x86 cross-licensing settlement was signed in 2001. AMD got x86-32 license plus some related licenses and Intel got x86-64 license plus some related licenses. This settlement was broken in early 2009
AMD knows very well that Intel holds the x86 license and their official response to Intel was:
Intel's action is an attempt to distract the world from the global antitrust scrutiny it faces. Should this matter proceed to litigation, we will prove that Intel fabricated this claim to interfere with our commercial relationships and thus has violated the cross-license.
AMD remains in full compliance with the cross-license agreement. And as we've stated all along, the structure of GLOBALFOUNDRIES takes into account all our cross-license agreements. We will continue to respect Intel's intellectual property rights, just as we expect them to respect ours.
Again, we believe that Intel manufactured this diversion as an attempt to distract attention from the increasing number of antitrust rulings against it around the world. With a ruling from the European Commission and a U.S. trial date looming, and investigations by the U.S. FTC and NY Attorney General, the clock is ticking on Intel's illegal practices - and yet with its dominant monopoly position it still tries to stifle competitors.
The AMD/Intel cross-license agreement is a two-way agreement, the benefits of which go to both companies. Intel leverages innovative AMD IP critical for its product designs under the cross license. This includes AMD patents related to 64-bit architecture extensions, integrated memory controller, multi-core architecture, etc.). The cross-license is very much a two-way street.
In fact, we informed Intel that their attempt to terminate AMD's license itself constitutes a breach of the cross-license agreement, which, if uncured, gives AMD the right to terminate Intel's license
AMD and Intel signed a new x86 cross-licensing settlement. It finishes the day 12 Nov 2014.
Not only the x86 ISA is a complete mess, but the x86 licenses/patents are a complete pain.
By all that AMD is migrating to ARM. Like it or no :lol:
@juanrga:
Evidently you are expert at neither of those fields:
Microcode is the x86 ISA. It's the instruction sets the microcode contains that define the ISA.
The paragraphs you cite are the excerpts from the portion where Intel appealed because the original ruling gave AMD access to all Intel proprietary licensing, and the rights to reverse engineer. The right to clean room reverse engineer was eventually rescinded by the appellate courts. So AMD cannot reverse engineer Intel hardware anymore...as the judge who ruled in the initial ruling judged that the IBM contract between AMD and Intel gave AMD rights to reproduce all Intel hardware as if they were Intel. The appellate court (on the second appeal mind you...the first appeal was shot down) came in and said that the ruling was overstepping boundaries and the original agreement with IBM did not merit such considerations for AMD. They ruled that AMD had to develop proprietary hardware of their own, but had access to the original x86-32 microcode a.k.a. the x86 ISA.
AMD is not moving to ARM anytime soon, and x86 is not going anywhere out of the consumer sector anytime soon either.
As for Intel threatening AMD:
The only threat that Intel can leverage is the use of extensions newer than those AMD already has a license for, or in layman's terms, extensions after AVX2. In order to compete AMD would need to be able to access those extensions so they can continue to run compatible code. However, AMD can produce x86-64 ISA CPUs that use up to AVX2 instruction sets indefinitely. Intel was trolling AMD, and if they breach the agreement, AMD can literally pull their license for x86-64 to Intel and instantly render all their future products incompatible with 64 bit instruction sets. In this instance, AMD literally has Intel over a barrel, and Intel is aware of it, but they are trolls anyway..
Reread all of what you posted in it's original context.
Also, might want to google processor microcode wikipedia entries, or whatever else you can find. Basically, x86 ISA without microcode is nothing. The ISA does not define hardware designs, rather, it defines what instruction set compatabilities the hardware must have to conform to the ISA. This is why AMD and Intel hardware can be so drastically different and still function as x86 CPUs, because the hardware itself is irrelevant as long as you can produce the desired outcome from the instructions processed.
In fact, GCN can actually run up to SSE4 x86-64 code if you want to get technical about it. If you think AMD is concerned enough about their licensing to switch to ARM entirely, then why did they make even their GPUs capable of running x86 instruction sets? Answer: because they're not concerned at all.
Also, might want to read up on ASM, because it is the most basic use of those microcode instructions we are discussing, and really opens up how everything "works"...
Now that we've established you're reaching desperately, please...go back to speculating about Kaveri, as your "ARM is going to rule the world" nonsense wore out it's welcome long, Long, LONG, ago...
/educational section
Nice to see you back to educate us. The problem is that you are completely wrong once again.
You confound microcode with macrocode. The assembly instructions of the x86 ISA are macrocode. AMD FX-8350 processor has different microcode (AMD own microcode) than Intel i7-3770k processor (Intel own microcode) and, however, both processors run the same macrocode: x86.
Maybe you would read the wikipedia links that you provide. Pay attention to this part:
A processor's microprograms operate on a more primitive, totally different and much more hardware-oriented architecture than the assembly instructions visible to normal programmers. In coordination with the hardware, the microcode implements the programmer-visible architecture. The underlying hardware need not have a fixed relationship to the visible architecture. This makes it possible to implement a given instruction set architecture on a wide variety of underlying hardware micro-architectures.
In his response to Intel threatening, AMD recognized that Intel holds the x86(32) patent, whereas AMD holds the x86(64) patent. This is clear in the official response, which I reproduced but you didn't understand. Maybe the next quote is more clever:
AMD believes it has a strong position if the situation turns into a patent slugfest. While it's true that Intel holds the x86 patent, AMD noted to Ars that it has a number of patents of its own, including some related to the functionality of integrated memory controllers, the x86-64 instruction set, and x86 multicore configurations. The company also hinted that it may hold patents regarding the creation of an integrated CPU+GPU product on a single die—the so-called "Fusion" parts that now appear on the roadmaps of both companies.
Your discussion about AVX2 is ridiculous. The x86 cross-licensing settlement was signed in 2001 and refreshed with a new settlement in 2009. AVX2 instructions were introduced in 2012.
No. GCN doesn't requires a x86 license from Intel. Who knows what new misconception is in your head!
I think the real question is if Juaranga is sent here to forum slide. He is constantly doing it.
That HSA video @ S|A forums highlights exactly what I've been saying all along, AMD needs a platform which allows multiple chips to communicate with each other over something that has more bandwidth than PCIe.
The entire slide just flat out came out and said they were using infiniband, exactly like I was hinting at, and Sernox has also had my back on this before by mentioning that it's possible to do Hypertransport over Infiniband.
http://insidehpc.com/2013/12/22/hsas-clusters/
This video (from S|A forums) blatantly states that AMD is actually going to do everything I was saying or at least something similar
1. Drop focus on servers and instead move to HPC and workstations (which is the goal of this entire presentation)
2. Create a way for dCPU, dGPU (AMD mentioned multiple APUs but there's no reason why that can't be big dCPU + dGPU)
3. Cease focusing on big x86 cores for servers and migrate to something that allows for HPC to migrate to desktop.
2015 we will be on a new high end AMD platform with infiniband and we will probably be able to add additional APUs, CPUs, and GPUs into expansion slots.
ARM is good enough in servers. It's not good enough for HPC, workstations, or even general purpose x86 computing (excluding workflows that exist only in browser).
As I've been saying. Juaranga has confused "we don't think x86 big cores are the right job for servers" coming from a company that has hit rock bottom for CPU market share in servers, with "we're abandoning our failure of a plan of releasing x86 server CPUs for 4% of the market and then jamming that architecture into desktop and mobile".
IT should be obvious why this isn't on any roadmaps yet. It's still more than likely being worked on and it would make Intel's CPU + PCIe slots for expansion to add sound cards and video cards platform look archaic.
I think I will enjoy reading this thread in a few years, much like how I like to stalk seronx's posts from years ago with everyone going "lol he's crazy xD WHERE DOES HE GET THIS" and then him being 80% right.
There are many "real questions such as why people here insists on labeling as fake official roadmaps, whereas others present their own fantasies about 10-core SR FX CPUs coming in 2015 as "my friend at AMD said me". Another real question is why you insist on mispelling my nickname. It is not "juaranga".
We know that AMD needs to comunicate different processors. We also know the answer is not hypertransport and this is why AMD has moved to PCIe with Berlin and Seattle.
AMD is not droping focus on servers. In fact all the talk presented new server processors such as Toronto and Cambridge.
AMD mentioned "multiple APUs", because as I said before laws of physics make unreliable (dCPU + dGPU) above certain scale and you have to move to an APU design. That is the reason, why both AMD and Nvidia are using APUs for their most advanced designs.
AMD will be migrating from x86 to ARM. They already said that ARM will win in the server space. We know that ARM will win in the HPC space also and then will migrate to desktop. I wrote about this two months ago. I repeat my guessed roadmap:
First servers, then HPC, latter laptops, and finally desktops.
Nice to see this thread going in circles, repeating the same mistakes, ignoring the same corrections, reposting the same fallacies/fantasies about big CPUs, and rediscovering stuff said months ago. I suppose some here are doing it on purpose.
Anyone has something to comment on nVidia GameWorks? it sounds like something really really Evil... can anyone share any visions on the matter?
It seems that Nvidia is taking a page out of Intels evil playbook with this tactic... If you can't beat them then cheat! I still can't understand why anyone would ever support these 2 companies (intel & nvidia) and their sleazy underhanded practices!
Anyone has something to comment on nVidia GameWorks? it sounds like something really really Evil... can anyone share any visions on the matter?
It seems that Nvidia is taking a page out of Intels evil playbook with this tactic... If you can't beat them then cheat! I still can't understand why anyone would ever support these 2 companies (intel & nvidia) and their sleazy underhanded practices!
Yea this is a reason why I support AMD and there open development procedures...
But people buying expensive computer parts really can't afford to spend based on principle and you can't blame them.
Anyone has something to comment on nVidia GameWorks? it sounds like something really really Evil... can anyone share any visions on the matter?
It seems that Nvidia is taking a page out of Intels evil playbook with this tactic... If you can't beat them then cheat! I still can't understand why anyone would ever support these 2 companies (intel & nvidia) and their sleazy underhanded practices!
Amd seemed to have learned from the old days though, I heard nvidia 'didn't want the consoles' because 'margins were too small' more like Amd were ging to undercut them no matter what and getting all those developers etc on board before releasing hsa apus, if this was 8 years ago they would have just released kaveri and said 'but it would perform like this if software utilized it fully' with no software fully using it of course.