AMD CPU speculation... and expert conjecture

[jdwii]

AMD talks a little bit more about Seattle
Hot Chips 26: The real deal in ARM A57-cores SoCs
http://semiaccurate.com/2014/08/28/amd-talks-little-bit-seattle/
same old stuff that was already known. amd is very, very tight lipped about performance.
preliminary uncore for project skybridge is here.
amd chose 8x relatively weak cores instead of more powerful x86 cores because of the type of workloads this soc will handle.

The last question was one AMD answered quite well, essentially why would you need eight relatively weak cores on a die instead of big x86 ones? It all comes down to workloads, specifically non-cachable workloads like, oh, search and social media. If a core only needs a small code block that operates on a large data set, a low IPC core can be just as efficient as a bigger high IPC core. Why? The work is dominated by cache misses and other types of stalls, loading from main memory is key. For this a simple, cheap, and power efficient core on an SoC with lots of RAM are just the ticket.

Lol Amd is already downplaying IPC again at least it took them some time on Bulldozer. Can care less what their excuses are. Its like sometimes when they say things i wonder why the person doesn't tell them about a company called Intel that is already extremely better in performance over IPC and perhaps after bulldozer they would of learned.

 

[juanrga]

4A Games' chief game developer talks about new consoles. Confirms my point that current gen games are not using all the hidden potential of the hardware and confirm that consoles will be able to match the performance of a ~4TFLOP gaming PC (R9-280X)

Digital Foundry: In our last interview you were excited by the possibilities of the next-gen consoles. Now you've shipped your first game(s) on both Xbox One and PlayStation 4. Are you still excited by the potential of these consoles?

Oles Shishkovstov: I think what we achieved with the new consoles was a really good job given the time we had with development kits in the studio - just four months hands-on experience with Xbox One and six months with the PlayStation 4 (I guess the problems we had getting kits to the Kiev office are well-known now).

But the fact is we haven't begun to fully utilise all the computing power we have. For example we have not utilised parallel compute contexts due to the lack of time and the 'alpha' state of support on those consoles at that time. That means that there is a lot of untapped performance that should translate into better visuals and gameplay as we get more familiar with the hardware.

Digital Foundry: Xbox 360 and PS3 were highly ambitious designs for the 2006/7 era. Xbox One and PS4 are much more budget conscious - have they got what it takes to last as long as their predecessors?

Oles Shishkovstov: Well obviously they aren't packing the bleeding edge hardware you can buy for PC (albeit for insane amounts of money) today. But they are relatively well-balanced pieces of hardware that are well above what most people have right now, performance-wise. And let's not forget that programming close to the metal will usually mean that we can get 2x performance gain over the equivalent PC spec. Practically achieving that performance takes some time, though!

But to answer the question - they could last as long. Just remember - back when PS3 first hit the stores - Nvidia G80 was released as well, and it was almost 2x faster than the RSX at the time...

 

[etayorius]

Can`t AMD just copy some of the design Intel has with their newer archs on Cache to improve their IPC?

Or is there some sort of license that stops AMD from doing so?

What type of License AMD has with Intel if they can´t even copy them? what if AMD dumps x86 completely and uses only x64? would that be possible? most apps are starting to move to x64 anyway.

 

[blackkstar]

Can`t AMD just copy some of the design Intel has with their newer archs on Cache to improve their IPC?

Or is there some sort of license that stops AMD from doing so?

What type of License AMD has with Intel if they can´t even copy them? what if AMD dumps x86 completely and uses only x64? would that be possible? most apps are starting to move to x64 anyway.

It is just as difficult to design something in hardware or software than it is to do so and avoid infringing on patents and end up paying big licensing fees or ending up in court.

I remember reading somewhere that Intel patented a lot of the ways they do caches, branch prediction, etc and that part of AMD's problem is designing something that avoids the patents Intel holds. Which might explain partially why they are off doing radically different things from Intel (modules, K12 and x86 sister core, semi-custom, etc)

I could find the source but I'm just checking in, cleaning the house right now. Waiting for 16GB of DDR2 and a motherboard, etc.

 

[de5_Roy]

So how are K12 and "Zen" going to be positioned?

K12 for wide and undemanding tasks, x86 for high demanding tasks that are the opposite of K12's tasks, and then use as much of each chip (uncore?) as possible in the other one?

amd will build a truly hybrid processor with both arm and x86 cores in 4+4 or 4+8 configuration with zen and k12 cores. project skybridge's early goal is to accomodate either cores on the same socket. so logically(!!) the next step would be to have both on the same die, under same power limit. additionally, zen cores on 14nm bulk finfet should be able to hit 3GHz and higher (according to an old semiwiki article on bulk finfet). zen seems to be aiming mostly at semicustom, embedded and server cpus and laptops, small desktops and high end tablets. i don't think zen will go higher than $190 cpus' performance level (e.g. per-core perf of a10-7850k, i5 4430). this is a very broad and nearly baseless assumption though (won't stop me from gloating if any of these come true ).

a lot of things can change. it has taken amd almost 3 years to out a vanilla a57-based soc, almost 2 years(3 counting the original timeline) to improve on piledriver with kaveri. it can take 2-3 years to launch zen and k12 counting from the time of first announcement. i think we should read into a little why amd announced k12 and not zen, just "k12's sister x86 uarch". that wasn't to downplay x86, rather to inform us the amount of effort it'll take them to design zen. with k12, they'll most definitely have arm's cooperation. imagine them asking for intel's help with zen (it might turn out to be true because it's so crazy! ).

Lol Amd is already downplaying IPC again at least it took them some time on Bulldozer. Can care less what their excuses are. Its like sometimes when they say things i wonder why the person doesn't tell them about a company called Intel that is already extremely better in performance over IPC and perhaps after bulldozer they would of learned.

seattle doesn't look like a performance chip despite how it was hyped here. it seems like each core serves a sata port and memory, simply outputs/processes to and from memory and storage, almost no amount of compute involved. ....kinda like a storage-oriented asic..? <-newbie math in action.

 

[gamerk316]

I'm also glad someone finally decided to do a test where they disable cores one by one and check if there's any detectable decrease in performance. BF4 scales to at least 8 cores, but as you said, dropping to 5 has no negative impact. That extra scaling isn't getting you any performance, yet carries additional overhead that has a global performance cost.

Last consoles devote six cores to games. We will see lots of new games optimized for six cores. BF4 is just one of them. Huddy has mentioned that six-eigth cores is the target for game developers.

Again, consoles are a special case, due to having exactly one set of HW to code against. You can do all sorts of low-level optimizations that you CAN NOT DO ON A GENERAL PURPOSE PC. This is highlighted by previous generations of games that didn't scale to that many cores on a PC.

Look, I've worked on systems where you actually code against the HW. For example, an internal switch takes x ms to flip and settle, and a NOOP takes y ms, so I have to put in z NOOP instructions after I tell the switch to flip. Consoles are a lot like that, though at a much higher level all the same, but you can still do those low level code optimizations. Example: You can guarantee that data will be in specific locations in RAM, therefore it will take a constant amount of time to load that data into a CPU register, so you can do something else for exactly some amount of time before going back in finishing your calculation. By comparison, a PC will sit around doing nothing waiting for that data to get into a CPU register. That's the type of stuff you can get away with on any piece of specialized HW, and why a PS4 can play BF4 and a E4000 series C2D with a 7800 GTX can't.

Are you implying AAA games for PS4 and Xbone have significant amounts of hand written assembly?

I was giving an example for a platform I work on (yes, handwritten ASM is involved), but at least as far back as the PS2/Gamecube/Xbox days there were significant ASM code and manual memory management going on. There's likely a lot less of that going on now, but I'd be surprised if it went away entirely.

And there's more to low level then just handwritten assembly. I'm talking very low level optimizations you can do, because no other unexpected tasks are running, you can guarantee, at any point in time, the state of the entire platform. This allows for very fine control of program execution that you can't get on PCs.

Example: On a Core 2 Quad, CPU cores 0,2 and 1,3 share the L2 cache. On a low level system, you can ensure that tasks that need access to specific data run on only those two cores, to gain slight performance boosts due to not needing to copy data across the L2 cache. On a PC, doing that is dangerous, since you can't guarantee what other tasks are running, and as a result, that specific CPU cores are not being used. That's a "low level" optimization, since you're bypassing the OS and manually assigning threads to cores. You can't (safety) do this on a PC, but on a platform that no other tasks are running on, you can.

 

[juanrga]

Spoiler

Translation: Its not much worse, but cheaper.

It is acceptable for basic academic use, but outdated for the industry.

No core based in that ISA will be a relevant competitor for AMD's K12/A57 based products.

In the ultra low power field? RISC and MIPS CPUs are king, because even ARM is too power hungry. Heck, there's a reason why the Z80, 68000k, and i386 are still used a lot in the embedded market: Because they are cheap, low power, and get the job done.

As mentioned above, their comparison to a Cortex-A5 was invalid because this core has to support lots of extras that were avoided in the Rocket design. A more accurate comparison would be a Cortex-M3 plus MMU and cache.

AMD will be aiming (check above slide) the "ultra low-power client" with Cortex-A57 cores. Those new cores have improved efficiency.

I found what AMD exactly means by "ultra low-power client": They mean sub 2 watt client.

 

[juanrga]

Can`t AMD just copy some of the design Intel has with their newer archs on Cache to improve their IPC?

Or is there some sort of license that stops AMD from doing so?

What type of License AMD has with Intel if they can´t even copy them? what if AMD dumps x86 completely and uses only x64? would that be possible? most apps are starting to move to x64 anyway.

Intle spends lots of people and money on R&D. It would be unfair if competing companies (AMD or other) could 'stole' their advances and use them freely. In fact, those costly advances are protected by patents and laws about infringment of patents.

But you wouldn't care about caches. Since Keller returned to AMD, he has been working in caches and AMD has filled about a dozen of patents on cache technologies, including patenting a 'new' type of cache subsystem (stacked cache).

If the data at hand is rigth. Keller new cache subsystem provides (i) a reduction of cache misses and (ii) latency reduction when compared to a more tradittional cache thechnology. I only can guess that this new cache could be to AMD future K12 what the uop cache was to Sandy Bridge:

The uop cache is one of the most promising features in Sandy Bridge because it both decreases power and improves performance. It avoids power hungry x86 decoding, which spans several pipline stages and requires fairly expensive hardware to handle the irregular instruction set. For a hit in the uop cache, Sandy Bridge’s pipeline (as measured by the mispredict penalty) is several cycles shorter than Nehalem’s, although in the case of a uop cache miss, the pipeline is about 2 stages longer. The uop cache increases performance by more consistently delivering uops to the back-end and eliminating various bubbles in the fetch and decode process. For example, the 16B fetch, length changing prefixes and the decoding restrictions all limit the traditional front-end, whereas the uop cache ignores those issues and can achieve higher performance.

As you know Steamroller is that was designed before Keller arrived at AMD. However, there are persistent rumours that Keller got his hands on Excavator and part of his work on patents will be used in Excavator.

Don't expect miracles for Excavator however; the whole of the new cache techniques is left for K12.

AMD has a cross-licensing agreement with Intel regarding x86. Why? Because AMD developed the 64bit ISA as an extension on top of x86-32 ISA and thus needed of the latter. With the cross-licensing, Intel gave to AMD the right to use x86-32 and AMD gave to Intel the right to use AMD64 (aka x86-64). AMD cannot rip x86-32 and use only 64bit because is not a separate, independent ISA.

This contrast with what ARM engineers did. ARM64 is a separate ISA. In fact ARM64 is a clean ISA where all the bizzarre aspects of ARM32 were eliminated. AMD could implement only ARM64 in its future K12 core, which would bring cost, performance and efficiency advantages over a full ARM64/ARM32 implementation. Howeever, there is no significant 32bit legacy code in ARM servers, thus the cost of avoiding legacy is free.

Finally, the AMD-Intel x86 cross-licensing agreement is very entertainmet. It is full of funny clausules, such as if AMD goes bankrupcy then Intel gets for free the right to use AMD64 or if AMD market share pass above certain limit, AMD loses rights about AMD64...

 

[Ags1]

The title thread is "AMD CPU speculation... and expert conjecture". It is a bad title, in the first place because APUs are more important to AMD than CPUs, and second because we also want discuss about GPUs, SSDs, and SoCs.

Maybe... "AMD speculation... and expert conjecture". I'd also drop the "expert" bit, it's unnecessary.

 

[8350rocks]

Lol Amd is already downplaying IPC again at least it took them some time on Bulldozer. Can care less what their excuses are. Its like sometimes when they say things i wonder why the person doesn't tell them about a company called Intel that is already extremely better in performance over IPC and perhaps after bulldozer they would of learned.

Only on the ARM cores...

 

[juanrga]

The title thread is "AMD CPU speculation... and expert conjecture". It is a bad title, in the first place because APUs are more important to AMD than CPUs, and second because we also want discuss about GPUs, SSDs, and SoCs.

Maybe... "AMD speculation... and expert conjecture". I'd also drop the "expert" bit, it's unnecessary.

Agree.

 

[juanrga]

This is another occassion where AMD said something completely different to what some few pretend here...

What AMD said is that there exists a concrete class of server workloads (e.g. I/O bound) that don't benefit from bringing big cores with higher IPC. Using those big cores doesn't accelerate the workloads, but add to both the initial cost and the power consumption. For those workloads smaller, simpler, cheaper, cores are enough. This is the microserver market.

This is the market for the old AMD Opteron servers based in jaguar cores. This is the market for Intel Avoton servers based in Silvermont cores.

AMD is replacing Opteron-X by Opteron-A because the ARM A57 cores are faster and efficient that the x86 jaguar cores. The Opteron X2150 scores 28.1 @1.9GHz in SPECint_rate 2006. The Opteron A1100 scores 80 @2.0GHz. The A57 core is 43% faster than the jaguar core but consumes about half the power.

Intel Avoton (8 core version) scores about 96 @2.6GHz. Applied Micro X-Gene is expected to score 110 @2.4GHz.

Ordinary big servers (requiring high compute) will continue using big cores with higher IPC and lots of power consumption. A pair of posts ago, I gave an updated roadmap of AMD products for each category, from ultra-low power up to big server, but it was ignored. As shown in the roadmap, the A57 core (a small core) will be used in low power products such as microservers, whereas the future AMD K12 core (a big core) will be used in high-power products such as high-performance servers.

AMD K12 will compete against Intel Broadwell-EP and Skylake-EP, against Applied Micro X-Gene 3, and against Broadcom Vulcan (a Haswell-like core). I gave my early estimations of K12 performance in a previous post. If someone did pay attention he or she will see that I am expecting AMD K12 single thread performance to be ~80% of Broadwell in general workloads.

 

[de5_Roy]

amd introduced 3 new 28nm athlon cpus, now the specs are on their website:
http://www.amd.com/en-us/products/processors/desktop/athlon-cpu
cpu-world has them in table form:
http://www.cpu-world.com/news_2014/2014082801_AMD_reveals_details_of_new_Athlon_X2_and_X4_CPUs.html
now we wait for these to show up on store shelves.

Samsung's DDR4 modules for servers have quadruple-stacked memory dies
http://techreport.com/news/26985/samsung-ddr4-modules-for-servers-have-quadruple-stacked-memory-dies

xbitlabs claims tonga has updated gcn 1.2 uarch
http://www.xbitlabs.com/news/graphics/display/20140826114104_AMD_Quietly_Reveals_Third_Iteration_of_GCN_Architecture_with_Tonga_GPU.html

The GCN 1.2-based GPUs and APUs also feature universal video decoder 6.0 (UVD 6), video encoder engine 3.1 (VCE 3.1) and a new scaler for better and more energy-efficient video playback.

if that last bit means no more jump in power use during video playback, good news for dgpus and great for the future igpus.

 

[Reynod]

Semiaccurate sounds very butthurt that they enable simple things like this because how their business model fails. It is actually quite funny. Anyone with any knowledge would know that PI's sucessor would not be in that time frame and the name faraway island doesn't even make any sense. Tech news has always been dodgy with all the rumors, just gotta avoid sites like wccftech and everything is pretty good. Anything that first gets published on wccftech is probably wrong.

That was the point of the fake tags to see how quickly the "news" would be "leaked".

The whole history is about more than a simple leak. Charlie/Semiaccurate accuse them of false attribution as well.

That's not quite what Charlie said but close. Just be aware here we don't slag off other sites ... including the one you write for, and S|A.

A lot of trogs (Litho Engineers) post on Charlie's site as well as here, as does Copper.

On a related note please do not mention John Fruehe again as he is also a friend, and has moved jobs twice since being VP Servers at AMD ... currently he is an analyst in IT.

We were lucky in that JFAMD paved the way for reps (other than wriiters like Crashman et al.,) from OEM's to interact on toms and we value them here. Should any reps post the mod team generally takes a dim view on users who troll or abuse them. Most of them post in their own time and share little tidbits of stuff that is invaluable, so we like to encourage them and make them feel welcome.

John did not misrepresent AMD re: Bulldozer, he stated facts according to the test data he had. He certainly cherrypicked very well though ... and you can't blame him for that.

John is a reare breed of individual who understands technology and what it means to the enduser at the corporate level. Charlie is a rare breed of writer cut from the Mike Magee mould at :The Inquirer" who is cheeky and has the odd weird rant at Jhen-sun just to stir the pot, or is brave enough to dress up in a bunny suit (wrong bunny suit) after losing a bet wit hthe ex-CEO at Intel.

Personally I like the characters we have in the tech world.

May you all find a free Titan-Z in your letterbox !!

 

[juanrga]

Semiaccurate sounds very butthurt that they enable simple things like this because how their business model fails. It is actually quite funny. Anyone with any knowledge would know that PI's sucessor would not be in that time frame and the name faraway island doesn't even make any sense. Tech news has always been dodgy with all the rumors, just gotta avoid sites like wccftech and everything is pretty good. Anything that first gets published on wccftech is probably wrong.

That was the point of the fake tags to see how quickly the "news" would be "leaked".

The whole history is about more than a simple leak. Charlie/Semiaccurate accuse them of false attribution as well.

That's not quite what Charlie said but close.

Hum, From Charlie email (bold from mine):

It would be one thing if a site said that we put it in the tags and speculated that it could be X, Y, and Z. It is quite another thing when CHW says that we claimed FI is the successor to PI. Not just that, it seems we said it would have ‘unknown’ architectural changes but is definitely a 20nm product made at GF. Not just that, it will have full support for DX12 and crush PI in perf/watt. But wait, it gets better, this architecture, according to me according to CHW, will be incorporated in both future x86 and future ARM SoCs, high performance ones of course. This includes Cheetah again according to them according to me. Wow, I didn’t know I knew that! I really didn’t know I knew I said that it would be an Opteron too, wow some days I surprise even myself. In short CHW saw joke of a tag line and made up a snotload of info on it

From the SA article (bold from mine):

Let me be very clear here. SemiAccurate never said one bit of what was said to come from us in the CHW article. It is pure fabrication on their part. Worse yet there are three real victims, you, SemiAccurate, and the truth. The truth died at the first CHW story, they literally made it up, complete lies. WCCFTech added more lies to it and didn’t credit, the rest were just a mish-mash of bad behavior that would get anyone bounced out of journalism school or a real job faster than you can say, “bye”.

Thus both sites are also accused of false attribution.

 

[h2323]

Finally, the AMD-Intel x86 cross-licensing agreement is very entertainmet. It is full of funny clausules, such as if AMD goes bankrupcy then Intel gets for free the right to use AMD64 or if AMD market share pass above certain limit, AMD loses rights about AMD64...

You are misinterpreting the license agreement. Actually you are way off.

If Intel's share of x86 market drops below 65% for 4 Q's the license agreement is renegotiated. AMD retains AMD64. It does not say Intel gets everything.

Agreement ends and renegotiation begin if AMD has a change of control as well.

Same if they go bankrupt, Intel does not just get AMD64, they can use it, they don't own it.

AMD cannot spin off x86, but they can be taken over but a renegotiation of terms would have to take place as contract would be void.

It's like this. The contract ends 2019, is then renegotiated. If either one of them break the terms then its just like when the contract ends, an all new negotiation begins. Intel does not get anything if AMD breaks contract, it's bad for them actually, thus the agreement. It's not all stipulations on AMD, Intel has several things that they have to stay in line with as well. Intel is heavily limited technically and through there marketing and sales practices. They can't lie and be sleezy.

What is bad for AMD if the contract ends....there is nothing legally speaking that stops Intel from its old practices of market manipulation and third party product impairment.

The renegotiation would actually take place after the dispute and litigation processes in the contract.

There is nothing really silly or entertaining it makes sense really.

 

[jdwii]

After looking at some benchmarks on Intel's latest(really 390$ for that unlocked 6 core) starting to think i'm running on primitive technology. Either way i do not feel its right for Amd to have to let Intel use their technology if they go under. Its like Nintendo going under and having to let others use their games.

Here is a nice link comparing Amd and Nvidia with driver progress
http://www.eteknix.com/examining-nvidias-driver-progress-since-launch-drivers-gtx-780-ti-gtx-680/

 

[juanrga]

Finally, the AMD-Intel x86 cross-licensing agreement is very entertainmet. It is full of funny clausules, such as if AMD goes bankrupcy then Intel gets for free the right to use AMD64 or if AMD market share pass above certain limit, AMD loses rights about AMD64...

You are misinterpreting the license agreement. Actually you are way off.

If Intel's share of x86 market drops below 65% for 4 Q's the license agreement is renegotiated. AMD retains AMD64. It does not say Intel gets everything.

Agreement ends and renegotiation begin if AMD has a change of control as well.

Same if they go bankrupt, Intel does not just get AMD64, they can use it, they don't own it.

Well I have to admit that I am not well versed in legal issues, but where you got that it is "renegotiated"? Clausule 6.2 says:

[limitations] shall terminate in its entirety, without the need for any further action by Intel, and shall be of no further force or effect if:

a. Intel’s microprocessor unit share of the Worldwide PC Market Segment, which includes desktop, server and mobile platforms, drops below sixty-five percent (65%) for four consecutive quarters as reported by Mercury Research in its published report entitled “PC Processors and Chip Sets Updated Edition Market Strategy and Forecast Report.” In the event that Mercury Research ceases publishing such share data during the term of this Agreement, the parties will agree in good faith on another publication to reference for such share data.

b. AMD attempts or purports to assign or transfer to any Person any right or obligation in violation of Section 9.7.

So far as I understand that, Intel could chose a renegotiation or not.

AMD cannot spin off x86, but they can be taken over but a renegotiation of terms would have to take place as contract would be void.

It's like this. The contract ends 2019, is then renegotiated. If either one of them break the terms then its just like when the contract ends, an all new negotiation begins. Intel does not get anything if AMD breaks contract, it's bad for them actually, thus the agreement. It's not all stipulations on AMD, Intel has several things that they have to stay in line with as well. Intel is heavily limited technically and through there marketing and sales practices. They can't lie and be sleezy.

I was discussing x86 licenses. The settlement claims that the cross-licensing finished in five-years:

AMD and Intel obtained patent rights from a new 5-year cross license agreement, and AMD and Intel relinquished any claims of breach from the previous license agreement. Intel also entered into a license agreement with Global Foundries, Inc., a manufacturing entity formed by AMD and Advanced Technology Investment Company. The parties agreed that the settlement was intended solely as a compromise of disputed claims, and was not to be understood as a concession or determination that either party has engaged in any wrongdoing.

It was signed in 2009, Nov 12. The cross-licensing agreement finishes this year not in 2019.

What is bad for AMD if the contract ends....there is nothing legally speaking that stops Intel from its old practices of market manipulation and third party product impairment.

The same happens if Intel market share drops below 65%, if AMD transfer rigths to third parties or if AMD is purchased by another company.

It is worth mentioning that the current cross-licensing agreement was signed some months after Intel accused AMD of violating the former x86 license agreement and AMD cross-acused Intel of the same

http://www.tomshardware.com/news/amd-intel-x86-cpu,7285.html

http://arstechnica.com/gadgets/2009/03/amd-intel-engaged-in-x86-licensing-staredown/

 

[h2323]

Juan


Yes it is a 5-year patent cross license agreement. sorry... wait, I found where I saw 10 years look at section 6.0 as it pertains to section 2.0 which is BUSINESS PRACTICES. So that's IP 5 years and business practices 10.

6.1 Terms: Section 2.0 of this Agreement shall remain in force and effect for a period of ten (10) years from and after the
Effective Date, and shall thereafter terminate, and be of no further force and effect, of its own accord without any further action by any Party, except that the requirements in Section 2.0 shall terminate before the expiration of ten (10) years in any of the circumstances set forth in Section 6.2, below. The remaining provisions of this Agreement will survive any termination or expiration of this Agreement and the termination or expiration of the requirements in Section 2.0.

9.7 Except in the event of a Change of Control as set forth in Section 6.3, this Agreement is personal to the Parties, and neither the Agreement nor any right or obligation under this Agreement is assignable, whether in conjunction with a change in ownership, merger, acquisition or the sale or transfer of all, substantially all or any part of a Party’s business or assets or otherwise, either voluntarily, by operation of law or otherwise, without the prior written consent of the other Party, which consent may be withheld at the sole discretion of such other Party. Any such purported assignment or transfer shall be deemed a breach of this Agreement and shall be null and void. This Agreement shall be binding upon and inure to the benefit of the Parties and their permitted successors and assigns

6.3 Change of Control . In the event of a Change of Control of AMD, the definition of AMD Microprocessor as defined in Section 1.5 shall be limited to those devices that fell within Section 1.5 on the date of the Change of Control and shall further be limited to x86 AMD Microprocessors for use in a Personal Computer.

1.5 AMD Microprocessor ” shall have the same meaning as the term “AMD Processor” in the New Patent Cross License.

Had a look over it again. No matter what.... Intel does not gain control over AMD IP, no matter what AMD does. The vast AMD x86 IP was not put on the chopping block in the terms of the agreement.

Intel can say again that your x86 license expires, and their ability to use 64 bit, integrated graphics and multi-core design would be in trouble...It's still a mess. I could see them both just extending the agreement. Legally I see no reason why AMD could not be sold as a whole unit. Intel would fight it but their position is weak.

This is an AMD favorable license. AMD stock skyrocketed when this agreement became formal.


In other news, look at that 8350 age with grace as Intel pumps them out

http://techreport.com/review/26977/intel-core-i7-5960x-processor-reviewed/12

 

[noob2222]

Technically the license isn't for x86 as that patent expired. Its for all the extras that go into it, sse, avx, fma, amd-64, etc. This is why nvidia was toying with the idea of x86 cpus but without being compatible with the software, not much point.

Originally Intel had AMD paying royalties for use of the extensions then went and sodomised AMD by disabling those extensions via software.

Intel is such an honorable company ...

"Edited for effect" Reynod

 

[gamerk316]

Well, nice to see Intel's approach of making a big core then making more of them is working out. 5960x looks beastly outside of gaming.

Seriously, look at those BF4 results. We all admit BF4 is the most well threaded game out right now, so obviously, the 8 cores of the 5960x will be a huge gain for Intel, right?

i7 5690x: 3.0 GHz, 8 Physical Cores, 16 Logical Cores
i5 4590: 3.3 GHz, 4 Physical Cores

The i5 4590 wins in our most threaded title out right now. Why? Because the CPU isn't the problem. Even in our most threaded game right now, pure per-core power wins the day over more, slower cores. Which I've been saying for 5 years now...Thief, another example of a "well threaded" game, is the same way, favoring the ever so slightly more power quad, despite the fact it scales well.

 

[gamerk316]

Technically the license isn't for x86 as that patent expired. Its for all the extras that go into it, sse, avx, fma, amd-64, etc. This is why nvidia was toying with the idea of x86 cpus but without being compatible with the software, not much point.

Originally Intel had AMD paying royalties for use of the extensions then went and fked AMD in the arse by disabling those extensions via software.

Intel is such an honorable company ...

You can't "disable" HW extensions in SW.

 

[de5_Roy]

As AMD And Origin PC Bury The Hatchet, A Personal Story About AMD's Efforts To Win Back Trust
http://www.forbes.com/sites/jasonevangelho/2014/08/28/as-amd-and-origin-pc-bury-the-hatchet-a-personal-story-about-amds-efforts-to-win-back-trust/
A New AMD Catalyst Linux Driver Unofficially Surfaces
http://www.phoronix.com/scan.php?page=news_item&px=MTc3NjU

No BS Podcast #231: AMD and Origin PC Settle Past Dispute on the Show
Plus: AMD's commitment to high-end CPUs, DDR4, 5-way GPU support, 20nm GPUs, and more!
http://www.maximumpc.com/no_bs_podcast_231_amd_and_origin_pc_settle_past_dispute_show
mostly "can't comment on future products" and "we're looking into it". btw, at present, fx brand includes mobile kaveri apus.

 

[juanrga]

The key here is what one means by x86. Does "x86" only means the original instruction sets used in the 8086 or further extensions are also part of it?

When we refer to a modern CPU we don't refer to it as a x86+x87+MMX+SSE+3DNow!+x86-64+AVX+... CPU, true? we refer to it collectively as a x86 CPU. These ISAs, including AMD's x86-64, are extensions to the original x86 set, they are not independent ISAs that can be implemented alone.

This contrast with what ARM did. The new 64 bit ISA is a clean, designed from scratch, independent ISA. The new 64 bit ARM is not an extension to ARM32. You can implement both AArch32 and AArch64 in a processor or you can implement only AArch64, simplifying the design and increasing efficiency. Broadcom will implement full AArch64, but only the user-mode of AArch32. As consequence, Broadcom Vulcan will be able to run legacy 32bit applications but not 32bit operative systems. I would expect something similar from AMD for the K12 and, in a close future (K13? K14?), I expect pure AArch64 processors: this is the reason why ARM designed 64bit as an independent ISA in the first place.

In my opinion the term "x86" in the AMD/Intel "x86 cross-licensing agreement" doesn't refer to the original instruction set in the 8086 but denotes the basic original set plus the needed extensions to build a modern CPU. Wikipedia article on x86 agrees on that only a subset of x86 is open but rest requires licensing from both AMD and Intel:

Open
Partly. For some advanced features, x86 may require license from Intel; x86-64 may require an additional license from AMD. The 80486 processor has been on the market for more than 20 years[1] and so cannot be subject to patent claims. The pre-586 subset of the x86 architecture is therefore fully open.

According to that, Nvidia has the rigth to do a 80486 chip, but needs a x86 license to make 80586 and posterior chips. It is interesting that the 80486 processors include both the original x86 set and the x87 extension. i.e. it was the first processor to broke the early distinction 8086/8087, 80286/80287, 80386/80387. There is no 80487 processors because the x87 unit was integrated in the 80486.

Charlie from SA also agrees (bold from mine):

Everyone contacted by SemiAccurate for the past several years has stated unequivocally that Nvidia does not have the legal right to make an x86 CPU. The problem is not the instruction set behind the chips, aka x86, but the underlying patents.

It seems Intel, AMD and VIA have some very strong patents that cover the basic functionality necessary to implement an x86 CPU. The patents are about how the instruction is carried out, not what the instruction is called. The right to use these patents are often referred to as an ‘x86 license’.

Intel and AMD have come to terms with those, the latest round settled late in 2009. VIA beat Intel with the patent stick as well, and the result was that VIA has the right to make x86 CPUs. While the rest of the agreement was confidential, the FTC just extended it for several more years.

Speaking of the FTC, it is strongly rumored that the one thing Nvidia wanted out of the FTC settlement was the right to make an x86 CPU. Sources at Nvidia tell SemiAccurate that this was the reason they complained to the FTC in the first place, however, we have not been able to independently confirm that claim.

In any case, Nvidia did not achieve that goal, and do not appear to have any right to make a CPU that uses certain of Intel’s patents. Even if they managed to negotiate use of those patents, there is still AMD and VIA to contend with. Anyone want to bet that Intel, AMD, and VIA aren’t keen to let a new player into the niche that they occupy?

One big clue about how badly Nvidia lost is in the FTC settlement, under Section I. F., Other Definitions. It defines, “Compatible x86 Microprocessor” as, in part iii. as “that is substantially binary compatible with an Intel x86 Microprocessor without using non-native execution such as emulation”. Our bold. Hear that? It is the sound of the door to Nvidia’s dreams slamming shut.

This won’t stop Nvidia though, their one managerial tactic when confronted with something they’re not allowed to do is to simply do it anyway then figure out how to make it legal later. This is what happened with Rambus, the recent 2% DDR royalty rate seems to be quite a bit higher than more tactful companies pay, and that still only covers items shipping from this point on. Things already shipped are still in dispute with Rambus and in court.

Intel is unlikely to let something like an Nvidia x86 chip hit the market, much less stay on the market for a few years, if they feel patents are violated. Nvidia will point out that Transmeta was never sued by Intel, so their approach is ‘legal’. Unfortunately for Nvidia, Intel does not share either opinion; that Transmeta was ‘legal’, or that Nvidia using the same approach would be ‘legal’ either.

From what we are told, Intel had a strong case against Transmeta, but the company went under before it was filed, if it was worth filing at all. Patent litigation has a lot of risks involved, the largest one is invalidation of those patents by a jury of laymen. It is a minefield, and losing is not the only down side.

If Nvidia uses the same techniques, and we are told that that is exactly what they are doing, Intel likely has the case against them already built. In fact, all Chipzilla needs to do is dust it off, search and replace a few names and dates, and off they go.

One other tactic Nvidia IR and PR uses to ‘sell’ their spin to investors is that since Intel didn’t go after Transmeta, they won’t or, if the analyst is gullible enough, can’t go after Nvidia. Nvidia made this fight personal, and Intel isn’t about to miss any opportunity to put a boot in.

More importantly, any litigation will be patent based, not copyright based. Copyrights and trademarks can be invalidated if they are not defended, but patents can’t. If Nvidia is actually trying to produce and then market CPUs based on these theories, they are in for a painful surprise.

 

[noob2222]

@gamerk

If cpu = AMD then SSE = null.

Intel did it and its a known fact.

And are you still not understanding single player campaign != multithreaded, single player campaign = scripted.