Intel Threatens To Revoke AMD's x86 License 476
theraindog writes "AMD's former manufacturing division opened for business last week as GlobalFoundries, but the spin-off may run afoul of AMD's 2001 cross-licensing agreement with Intel. Indeed, Intel has formally accused AMD of violating the agreement, and threatened to terminate the company's licenses in 60 days if a resolution is not found. Intel contends that GlobalFoundries is not a subsidiary of AMD, and thus is not covered by the licensing agreement. AMD has fired back, insisting that it has done nothing wrong, and that Intel's threat constitutes a violation of the deal. At stake is not only AMD's ability to build processors that use Intel's x86 technology, but also Intel's ability to use AMD's x86-64 tech in its CPUs."
Re:Fuzzy on x86 IP (Score:5, Informative)
I believe it's not the core x86 instructions, but rather all the various MMX and SSE extensions that have been tacked on in the past 10-15 years. And as mentioned in the summary, AMD's x64 extensions are at stake, too.
Re:if they do that (Score:5, Informative)
Since x86_64 is a superset of x86, would this mean AMD couldn't even sell x86_64 based chips either?
Funny thing is that AMD licensed/agreed to share their x86_64 arch back to Intel.
So essentially it's:
"I'll let you play with mine if I can play with yours."
Now a 3rd party (loosely affiliated with AMD) is playing with Intel's x86, and that wasn't part of the agreement.
Re:if they do that (Score:3, Informative)
No, but most users don't need to. Microsoft does, and Microsoft has no reason to want any one other firm to be indispensable to PC vendors the way Microsoft is. So, if the AMD cross-licensing agreement goes away and there isn't serious competition for Intel in the x86 world, I'd expect Microsoft to start supporting alternatives.
Re:if they do that (Score:3, Informative)
Re:if they do that (Score:5, Informative)
It IS just a matter of passing a different CPU flag. MS discontinued the MIPS, PPC, and Alpha versions because there was not only no demand for it, but the few people who bought it tied up lots of MS customer service time bitching that X86 programs didn't run on MIPs/PPC/Alpha.
Windows is no more married to X86 than Linux or OS X. In fact, I can tell you where to get a fairly modern Windows Kernel running on a PPC chip in pretty much any electronics store: The XBox 360.
The NT kernel was designed from the ground up to be portable. The only real reason it's currently only supporting X86 is because that's the only place there's any sort of demand. If X86 dies (And it won't. AMD and Intel both have lots to lose, though AMD more than Intel here), Microsoft will port over to PPC (Or whatever), throw on an emulation layer, and probably take the opportunity to break a whole bunch of crappy stuff in Windows that's maintained simply for backwards compatibility.
Re:if they do that (Score:3, Informative)
Agree, ARM has been gaining grounds due to it's low (as none) power consumption when idle. So long backward compatibility tough.
Re:if they do that (Score:3, Informative)
Intel has 18-wheeler-truckloads more resources for marketing than AMD will ever hope to garner. While there will always be the minority that will seek alternatives, Intel has the power to win them over, whether it'd be through financial incentives, equipment "giveaways" or brute-force, corporate style.
If AMD loses its x86 license, I'll speculate that AMD will have to choose the three obvious paths:
I don't want to see AMD go down, but it's kind of sad to know that Intel has the power to do exactly that.
Re:if they do that (Score:3, Informative)
It also seems a bit unclear as to why Sun continue to develop and produce the SPARC
Because there is a demand for it, and it does things that x86 doesn't. 8 cores * 8 threads = awesome virtualization abilities. The ability for SPARCs to scale up in a linear fashion to > 100 cores in a single general-purpose SMP box positions it in the high-end datacenter realm, where PPC is, but not x86. Plus, SUN isn't going it alone. Fujitsu is on the SPARC bandwagon with them.
Re:Intel will license it (Score:1, Informative)
You are apparently ignorant of history.
Intel licensed x86 to AMD originally because Intel was unable to keep up with demand.
When Intel decided to end that relationship, AMD refused to stop making x86's, and sued Intel to keep the right to do so.
AMD actually LOST that case, but AMD and Intel were told by the courts to make a license that worked, and AMD was forced to pay Intel for court costs. They renewed the license in 2001.
AMD has now breached the license. Intel has no responsibility to keep AMD in business. Intel can get another foundry to make x86 CPUs. There's no law against being a monopoly.
Natural law is against being a failure like AMD.
Re:if they do that (Score:2, Informative)
If x86 dies, which it is in the process of doing, Microsoft will port Windows to run on SPARC, ARM, PPC, whatever comes next.
You must be new. There already was a PPC port and a speculated Sparc port as well. That was, what, 10 years ago?
Lead, follow, or get the hell out of my way. (Score:3, Informative)
But the big draw of windows is the inertia of 1,000,000 one-off apps that businesses have written. Microsoft would be scared of people moving to another architecture just because if people were making a (painful) switch anyway, they might look at the alternatives.
If you wrote the damn app, then learn to recompile it and move on to whatever/whomever is going to be pimping procs next month or next year. If you're that worried about your legacy apps, then learn to use virtualization.
Moores law didn't get to be a "law" by playing nice and waiting around. Lead, follow, or get the hell out of my way.
Inertia is as fast and powerful as the people behind it. Adapt or die. It's that simple.
Re:"open" patent licencing as remedy to monopoly (Score:5, Informative)
No, their not. Abusing a monopoly position is.
I can certainly patent sexwidget and have a perfectly legal monopoly as the only company in the world producing them. Only if I try to force people to do other things not directly related to my sexwidget in order to get access to them is it considered abusing my monopoly status. In other words, if I try to force retailers to purchase other products like sexfoo & sexbar as a requirement for being able to sell sexwidgets, I'm abusing my monopoly.
Re:if they do that (Score:3, Informative)
The Cell is PPC based, can be faster than a Core i7 in some respects and is available cheaply in quantity.
If you are willing to order sufficient quantity, IBM can crank out fast and cheap PPC chips quite easily... They did it for both Sony and MS with their respective games consoles.
Emulation on the other hand will always incur a performance hit, sometimes quite a substantial one... Tho it helps if the CPU is designed to handle it. When the Alpha was still fairly new, you could run x86 emulation on it and actually outperform the real x86 hardware that was available at the time, but unfortunately Alpha was never made in sufficient quantities to push the prices down.
Re:if they do that (Score:3, Informative)
Today's chips, at their core, look a lot like RISC chips. They do a lot of work to hide that, translating x86 ops to native ops. I'd like to see a chip that can run in a x86 'translated' mode and a 'native' RISC mode, much like was done with 32bit/64 bit.
Except wouldn't that potentially be slower? More data would need loading off the disk into memory, and from memory into cache, since the RISC translation is (usually) larger.
Don't get me wrong, there's a lot 'wrong' with IA32, but I'm not so sure it would be worth using a 100% RISC ISA as an alternative. Complex instructions can be beneficial.
Re:if they do that (Score:4, Informative)
Not all virtualization requires hardware extensions. In fact, VMware was doing it long before Intel and AMD added virtualization support to their processors. VMware pulled this off by doing dynamic translation, where the virtual machine monitor would transparently rewrite native x86 into virtualized x86 code. For the most part this was just doing a straight copy, and perhaps rewriting some jump addresses. Privileged code that runs in the OS kernel had to be rewritten as something equivalent that would run fine in an unprivileged process.
This really isn't so different from running .NET or Java code. The code starts out compiled to a virtual instruction set, and the JIT compiler translates this on the fly to something that can run natively on the CPU.
This is also how Rosetta worked in Mac OS X to run PPC apps on an x86 processor. XBox 360 does a similar thing to run old XBox games, since the 360 uses a PPC processor but the old XBox was x86.
Sure, you take a performance hit in doing this, but the apps generally get rewritten to run natively eventually, and the ones that don't end up being old enough that they run faster on modern hardware even with the extra translation layer.
Re:if they do that (Score:3, Informative)
Or someone who never owned an Alpha system.
Strictly speaking you COULD run x86 apps on them, but the performance was abysmal.
Re:if they do that (Score:1, Informative)
Does AMD have any architectures not based on x86 to fall back on?
SPARC is open source.
http://en.wikipedia.org/wiki/Sparc [wikipedia.org]
"As a result of SPARC International, the SPARC architecture is fully open and non-proprietary."
http://en.wikipedia.org/wiki/Sparc#Open_source_implementations [wikipedia.org]
"OpenSPARC T2, released in 2008, a 64-bit, 64-thread implementation conforming to the UltraSPARC Architecture 2007 and to SPARC Version 9 (Level 1). Source code is written in Verilog, and licensed under many licenses. Most OpenSPARC T2 source code is licensed under the GPL. Source based on existent open source projects will continue to be licensed under their current licenses. Binary programs are licensed under a binary Software License Agreement."
Re:Intel will license it (Score:4, Informative)
Intel wasn't actually convicted of abusing its monopoly status. It wasn't a monopoly. And it wasn't convicted.
It settled with an economic commission (none of these things are courts) and at that point decided it was cheaper to pay the fine (less than $50 million; about an hour's pay to Intel) than to fight it in a court.
In the settlement Intel admits no wrongdoing, and the Japanese assert none.
Re:if they do that (Score:4, Informative)
I'd like to see a chip that can run in a x86 'translated' mode and a 'native' RISC mode, much like was done with 32bit/64 bit.
Already ready to use. The Transmeta Crusoe processor does this on the fly. Of course, now they're owned (or is that pwned?) by Novafora [novafora.com] so your guess is as good as mine whether this will survive.
Re:if they do that (Score:2, Informative)
You inverted Microsoft and Intel (Microsoft is the $140 billion company).
Intel basically drives investment in fab technology, I wouldn't harsh on them too much.
Have you ever used an Itanium box? (Score:2, Informative)
Have you ever used an Itanium box?
Jeez, what an awful piece of cr*p. Sound of a vaccum cleaner, performance slower than an equivalent x86, Mhz for Mhz (timeframe: 2000/1). Well maybe no in benchmarks, but if you had a box, side by side, both running Whistler, you couldn't tell the difference.
I had an early pre-release Intel box (well, I had several) plus pre-release Visual Studio and compilers. I ported a 2,000,000 line C++ CAD app from 32 bit MFC to 64 bit MFC. We did the port, but the box did not sing. It was horrible.
5 years earlier I'd used Sun's Windows emulation environment running Windows apps on Sparcstation pizza boxes. That was better.
Itanium is much more of a dead platform than x86.
I don't know how expandable SPARC is, in terms of future bandwidth, but if its available its a reasonable legacy bet, given Sun have the emulator software.
Real shame they dropped Alpha. That was a good platform. Ahread of its time. We had one in our office early nineties, running Digital UX. Sometime in 90-94. That thing was fast, compared to the competition.
ARM would be excellent though, I'd love that to happen. Same platform for desktop, mobile, embedded, low power, high performance. All we need is multi-core (sorry, haven't followed it closely enough to know if that is the horizon).
I've only used 2 machines in my life that have sounded like vaccuum cleaners:
1) Motorola Exorciser, 6809 development system with 8" floppies
2) Intel development Itanium box (several of).
Both were [polite]not very good[/polite].
Re:if they do that (Score:3, Informative)
emulation works to a point but
1: it's very hard to get it perfect
2: you generally lose a lot of performance. This is not an issue when emulating really old stuff but trying to emulate x86 on the comparitively slow arm is going to give terrible performance.
Sometimes you can get away with it. Apple did a pretty good job all considered. Sony screwed up pretty badly imo (even thier PSone emulation has bugs and thier partially software PS2 bc on the european PS3 was pretty terrible at least with ratchet and clank, the american PS3 with bc had pretty much all the PS2 hardware inside sidesteping the emulation problem). I haven't tried the XBOX 360 myself so I can't comment on that.
I doubt some small netbook vendor would have the resources to do this well even if there were suitable (as in faster than the intel chips they are trying to replace) CPUs on the market.
Re:Intel will license it (Score:4, Informative)
You are apparently ignorant of history.
You apparently can't even be bothered to read the wikipedia entry on AMD [wikipedia.org].
Intel licensed x86 to AMD originally because Intel was unable to keep up with demand.
AMD was a second source for the 8086 and 8088 because IBM demanded two sources, not because Intel couldn't make enough.
AMD refused to stop making x86's, and sued Intel to keep the right to do so. AMD actually LOST that case,
AMD was the one who challenged the x86 license cancellation and won the case in arbitration, and after numerous appeals it was upheld b the California Supreme Court.
They renewed the license in 2001. AMD has now breached the license.
Given that the licensing agreement isn't public, your analysis is clearly pulled straight from your rectum.
Intel has no responsibility to keep AMD in business.
The amusing thing about cross licensing agreements is that they cross. You can't really cancel half a contract. If Intel forces AMD out of the x86 CPU market... then Intel is out of it too, unless they intend to use something other than EMT64, which is a licensed implementation of AMD's proprietary AMD64.
Natural law is against being a failure like AMD.
Oh, I see. Your an Intel fanboy. That explains it.
who moderated this fool up so high?
Re:if they do that (Score:2, Informative)
Consoles are a special case though. They are sold at a loss and subsidised by the cost of the games.
Usually when a new generation console is introduced they have back compatibility by essentially including bits of the old console - i.e. the first PS3s, or by some hokey emulation code - the XBox360. The PS3 dropped the legacy hardware to cut prices and become profitable. The XBox360 could only play 13% of XBox games [daggle.com], which is actually quite an achievment considering how real time consoles are and that the CPU and GPU were totally different.
So manufacturers talk about back compatibility as a marketing bullet point. It's not really true though. I don't think it matters - people that care about old games will have the old console anyway and they can play them on that.
PCs are different to this - people have loads of software which they absolutely want to use when they buy a new machine. And compatibility break will cripple sales of a new OS.
Re:if they do that (Score:3, Informative)
Except that they didn't lose backwards compatibility. Do you even know about Universal Binaries, Rosetta, or OpenStep's legacy of hardware (and even operating system) independence?
Re:Dont see why they need a licence (Score:3, Informative)
The patent (U.S. Patent No. 4,814,976) covered a specific way of implementing four instructions. If you implemented them in a different way, then you would not be infringing, but I don't know of any companies that did this (by the way, the patent expires this year, so expect to see more complete MIPS implementations start appearing). A few licensed the patents. Most implemented almost-MIPS architectures, where the four relevant instructions were omitted. GCC and a number of other compilers have switches that allow you to generate code which doesn't use these instructions and an operating system can (very slowly) catch the illegal instruction exception and emulate them for legacy code.
Re:if they do that (Score:3, Informative)
Only L1 cache actually make a distinction between instruction cache and data cache. And AFAIK the instruction cache actually stores the translated instructions, so there isn't going to be any savings from x86 code being more dense. But maybe you meant that the space you save for x86 instructions stored in L2 and L3 cache are worth more than the overhead of translation.