AMD Can't Shake Off US Class-Action Lawsuit over Bulldozer "8-Core" Advertising

[Nobu]

Well it’s news to me that we’re moving away from GHz and that “core” is now a new metric.

This lawsuit is absurd

Also, a core may perform many ops in a single clock cycle, depending on what they are...so saying a core can only perform one task calculation at a time is dubious at best. And that's ignoring multithreading and other similar techniques.

Edited to match terminology: if they had said task, they wouldn't really be that wrong. Context switches are pretty expensive after all. But a cpu core may perform many calculations in a single cycle, and some ops even modify two pieces of data at a time (while performing the calculation).

 

[OutOfPhase]

Also, a core may perform many ops in a single clock cycle, depending on what they are...so saying a core can only perform one task at a time is dubious at best. And that's ignoring multithreading and other similar techniques.

Right. All superscalar processors do this, thus their (the lawyers) own definition means Bulldozer is a 20 core chip.

 

[g-money]

To be fair the coffee lady had a good case. The coffee was somehow about 30 degrees hotter than it should have been served.

I actually knew a lawyer involved in representing her. Her burns were... Life Altering. We're talking skin grafts on her VJ region and massive re constructive surgery. You are quite correct.

So interestingly - that case appeared to come down to three things (ref: WikiPedia which of course has its own sources listed):

1. The court & jury found McDonalds 80% at fault and the lady 20%
2. It was only argued (and supported by some evidence from other chains) that the coffee was found to be too hot (180/90 vs 160)
3. The actual fault primarily lied in the warning label (not being big enough / visible enough) and McDonald's past history of having settled similar claims for up to $500,000 over several years.

Further, McD's and other chains serve coffee just as hot, or hotter today. My take is that this is simply a very unfortunate accident and that at the time McD's needed to improve their warnings & cup material to maximize safety - but that they should have settled before this ever reached a jury.

this reaks of "I didn't know coffee was hot befor I spilled it on me"

edit:

and also people who don't know shit about technology, being in a position to make judgements about said technology.....

but hey, congrats to everyone who is a part of this class action, I hope you spend your 8 dollars well when you get it....
meanwhile the lawyers will be riding around in new bentlys

For the class action against AMD - I don't really see the similarity in this comparison. Knowing that hot coffee is hot and may damage you if you spill it isn't in any way like buying an '8 core' CPU and getting less performance out of it based on thinking that each core is physical, produces the same performance, scales, and doesn't have resource contention or bottlenecking issues with the architecture. One definitely IS common sense (yes I have spilled coffee on myself) and the other takes a good bit of technical understanding. This lawsuit will probably, and unfortunately, come down to perception of marketing. Actual damages on the other hand may be hard to prove so the final figures (or settlements) may end up being relatively minor.

Freely admit, I may be wrong but afterall, just my opinion --- (and I find coffee related things very interesting)

 

[ManofGod]

So interestingly - that case appeared to come down to three things (ref: WikiPedia which of course has its own sources listed):

1. The court & jury found McDonalds 80% at fault and the lady 20%
2. It was only argued (and supported by some evidence from other chains) that the coffee was found to be too hot (180/90 vs 160)
3. The actual fault primarily lied in the warning label (not being big enough / visible enough) and McDonald's past history of having settled similar claims for up to $500,000 over several years.

Further, McD's and other chains serve coffee just as hot, or hotter today. My take is that this is simply a very unfortunate accident and that at the time McD's needed to improve their warnings & cup material to maximize safety - but that they should have settled before this ever reached a jury.



For the class action against AMD - I don't really see the similarity in this comparison. Knowing that hot coffee is hot and may damage you if you spill it isn't in any way like buying an '8 core' CPU and getting less performance out of it based on thinking that each core is physical, produces the same performance, scales, and doesn't have resource contention or bottlenecking issues with the architecture. One definitely IS common sense (yes I have spilled coffee on myself) and the other takes a good bit of technical understanding. This lawsuit will probably, and unfortunately, come down to perception of marketing. Actual damages on the other hand may be hard to prove so the final figures (or settlements) may end up being relatively minor.

Freely admit, I may be wrong but afterall, just my opinion --- (and I find coffee related things very interesting)

Wow, an 8 core cup of coffee? Does that mean it cannot be decaf?

 

[noko]

AMD laid out what they considered cores over and over again. There is no clear cut defined definition of a CPU core, it has varied significantly over the years with external caches, FPU or no FPU at all on the core, memory controllers etc. AMD showed what they had, benchmarks were readily available for the performance you would get. Someone making up a new definition of what a core is will probably contradict numerous designs already made. Will be interesting - think AMD should win this one. Even though Bulldozered kinda sucked in the scheme of things.

 

[Legendary Gamer]

So interestingly - that case appeared to come down to three things (ref: WikiPedia which of course has its own sources listed):

1. The court & jury found McDonalds 80% at fault and the lady 20%
2. It was only argued (and supported by some evidence from other chains) that the coffee was found to be too hot (180/90 vs 160)
3. The actual fault primarily lied in the warning label (not being big enough / visible enough) and McDonald's past history of having settled similar claims for up to $500,000 over several years.

Further, McD's and other chains serve coffee just as hot, or hotter today. My take is that this is simply a very unfortunate accident and that at the time McD's needed to improve their warnings & cup material to maximize safety - but that they should have settled before this ever reached a jury.



For the class action against AMD - I don't really see the similarity in this comparison. Knowing that hot coffee is hot and may damage you if you spill it isn't in any way like buying an '8 core' CPU and getting less performance out of it based on thinking that each core is physical, produces the same performance, scales, and doesn't have resource contention or bottlenecking issues with the architecture. One definitely IS common sense (yes I have spilled coffee on myself) and the other takes a good bit of technical understanding. This lawsuit will probably, and unfortunately, come down to perception of marketing. Actual damages on the other hand may be hard to prove so the final figures (or settlements) may end up being relatively minor.

Freely admit, I may be wrong but afterall, just my opinion --- (and I find coffee related things very interesting)

I had a conversation with an attorney a while ago about it. I was much in the position of others at the time about dismissing it as personal idiocy. I got schooled in that conversation. I don't recall everything about that conversation, however, it wasn't pretty what happened to the woman.

I have spilled coffee on myself before in the car, it was a Venti Starbucks. Didn't kill me, didn't give me any sorts of burns at all. Was incredibly unpleasant when the lid popped off and my paper cup crushed... What this woman went through was something entirely different. WikiPedia is great and all, however, it's entered by people that aren't necessarily in the possession of all the details and facts. So, while I don't recall all the details of my conversation... What I did remember were specific details that changed my perspective on the woman. Wikipedia didn't likely speak to an attorney that had anything to do with the case and if they had, they probably couldn't list it. My conversation was a personal one, that's a public website. I take everything with a healthy dose of salt when I'm doing Internet research.

You are free to draw whatever conclusion you want to. I will firmly stand by my, altered, perspective. It's all too easy to dismiss things in life when it's not happening to you.

 

[mope54]

The problem is less about what does or doesn't constitute a core, but more about how these architectural changes are marketed and how they impact the general consumer.

For decades now, CPU (and GPU to a lesser extent) advertising has been smoke and mirrors. The kind of claims that many of us roll our eyes at and witnessed as abrupt shifts to get ahead of the market at various times (coming next: 5GIG processors!; coming next: dual core processors!, etc.). We all just played along with the marketing and consumed the products regardless of how well the advertising and technical merits synced. Many here would have explained the disjuncture away as marketing teams vs. engineering teams not really speaking the same language. But it's to the point where consumers are being harmed. This lawsuit and attempt by consumers and the law to define computing in more concrete terms is indicative that we are moving out of the wild west era of computing and into a more solidified formulation.

 

[OutOfPhase]

The problem is less about what does or doesn't constitute a core, but more about how these architectural changes are marketed and how they impact the general consumer.

For decades now, CPU (and GPU to a lesser extent) advertising has been smoke and mirrors. The kind of claims that many of us roll our eyes at and witnessed as abrupt shifts to get ahead of the market at various times (coming next: 5GIG processors!; coming next: dual core processors!, etc.). We all just played along with the marketing and consumed the products regardless of how well the advertising and technical merits synced. Many here would have explained the disjuncture away as marketing teams vs. engineering teams not really speaking the same language. But it's to the point where consumers are being harmed. This lawsuit and attempt by consumers and the law to define computing in more concrete terms is indicative that we are moving out of the wild west era of computing and into a more solidified formulation.

I'm very uncomfortable with this path, as it leads to weird situations where engineering is told they need to design something in a compromised way Because Lawyers.

There is a very wide variety of topologies possible to design, with varying allocations to each part of the processor. Each will be better at some tasks, worse at others. When we're designing something, I really want to focus on hitting the design goals for a given use case, and not be worrying over whether or not such a design will get my company sued. It is actually true that most consumer workloads need far more integer processing than floating point. I can save you money by leaving out silicon which won't be used effectively, or applying that same space towards features which would actually help.

As someone posted earlier, there are octo-core chips which will be outperformed by 2 modern intel skylake cores in most workloads. Core count is not a meaningful metric, except comparing within the same architectural family.

While it may come across as a bit cavalier and Laissez-Faire, the onus really is on the consumer to determine what they are buying. There is no other solution which will result in a positive outcome for the consumer.

 

[mope54]

I'm not advocating for it to happen, but you're describing what my last sentence is driving at: less innovation and clamped down design fitting squarely within the contours of what marketing and consumers can understand in definitive terms.

Society had to go through this on the software side, as well. Laws eventually defined OS, kernels, APIs, and licenses, for example. Decades ago, people were just toiling away over the technical distinctions without much regard for the legal or perceptive differences. While it's true that context allowed for a lot of innovation, it's also true the distinctions created by the major players that emerged from that milieu were ginned up to sell product. When we moved away from the gigahertz wars, tech savvy people were relieved from my perspective ("thank goodness that crap is over) but now we've been in a cores war, which again people will be glad when it is finally put to rest.

So I get the dangers of locking down operational terms, but at the same time without guardrails the industry just steamrolls right over the average consumer.

 

[OutOfPhase]

I'm not advocating for it to happen, but you're describing what my last sentence is driving at: less innovation and clamped down design fitting squarely within the contours of what marketing and consumers can understand in definitive terms.

Society had to go through this on the software side, as well. Laws eventually defined OS, kernels, APIs, and licenses, for example. Decades ago, people were just toiling away over the technical distinctions without much regard for the legal or perceptive differences. While it's true that context allowed for a lot of innovation, it's also true the distinctions created by the major players that emerged from that milieu were ginned up to sell product. When we moved away from the gigahertz wars, tech savvy people were relieved from my perspective ("thank goodness that crap is over) but now we've been in a cores war, which again people will be glad when it is finally put to rest.

So I get the dangers of locking down operational terms, but at the same time without guardrails the industry just steamrolls right over the average consumer.

Laws did not and do not define what an OS is, or a kernel, or API. Legal folks have come to better understand those terms, but they did not decide or define them. And here's a fun exercise - ask 10 lawyers if Linux is a kernel or OS.

The industry is in no way steamrolling any consumer on this topic. There's this giant ubiquitous source of searchable information they can access at any time, for everyone. If a consumer chooses to not make use of that, then they are of course very likely to buy something which may not be what they want. This is no different than a car, coffemaker, vacuum, vacation rental, musical instrument, dog food, etc.

Now if AMD said you'd get performance X in application Y and you did not, then there's perhaps a case. But saying it has 8 cores, when it does by any technical definition, is not wrong. The fact that those 8 cores kinda sucked is not even the foundation of the lawsuit!

 

[mope54]

Laws did not and do not define what an OS is, or a kernel, or API. Legal folks have come to better understand those terms, but they did not decide or define them. And here's a fun exercise - ask 10 lawyers if Linux is a kernel or OS.

They came to "better understand those terms" through court cases.

That's the history of it; you may not have lived through it but I did and I'm not just quoting from wiki but whatever. I don't understand how you think those terms came to be defined otherwise. They aren't organic terms and have been contested over time.

 

[Nobu]

They came to "better understand those terms" through court cases.

That's the history of it; you may not have lived through it but I did and I'm not just quoting from wiki but whatever. I don't understand how you think those terms came to be defined otherwise. They aren't organic terms and have been contested over time.

If it is defined in this court case, I can't see it being defined as anything other than "a group of transistors in a processor responsible for performing calculations and mathematical operations in a cpu" and maybe "...and supporting cache/memory where applicable".

 

[mope54]

I really hope you two are simply playing devil's advocate and not as detached from the civic process as your posts imply.

These aren't some downtrodden companies just trying to innovate--this is a problem of their own making. They deliberately obscured the difference between processor speeds, numbers, and then types in order to outsell the competition; and those are just the big bullet points.

This case won't define cores. Law doesn't operate like that. Law, like technology, is an iterative process. During this case, different interpretations, operational terms, and practical implications will be discussed as both sides try to untangle the web of what both Intel and AMD try to sell the public on what a "core" is and how that impacts computing experiences.

The sides will present their evidence for their position of what constitutes a core and why it does or doesn't matter, the judge will preside over whether that evidence is presented according to the rules, and the jury will determine how they understand the evidence. From there, another case will arise and another and another until we have a sufficient body of law to adequately and clearly define the points in contention.

That's how law and society interact with one another while jury members act as surrogates for the rest of us. It's a sloppy mechanism but it's the one we've got and some say it's the best there is or ever was.

 

[Nobu]

I really hope you two are simply playing devil's advocate and not as detached from the civic process as your posts imply.

These aren't some downtrodden companies just trying to innovate--this is a problem of their own making. They deliberately obscured the difference between processor speeds, numbers, and then types in order to outsell the competition; and those are just the big bullet points.

This case won't define cores. Law doesn't operate like that. Law, like technology, is an iterative process. During this case, different interpretations, operational terms, and practical implications will be discussed as both sides try to untangle the web of what both Intel and AMD try to sell the public on what a "core" is and how that impacts computing experiences.

The sides will present their evidence for their position of what constitutes a core and why it does or doesn't matter, the judge will preside over whether that evidence is presented according to the rules, and the jury will determine how they understand the evidence. From there, another case will arise and another and another until we have a sufficient body of law to adequately and clearly define the points in contention.

That's how law and society interact with one another while jury members act as surrogates for the rest of us. It's a sloppy mechanism but it's the one we've got and some say it's the best there is or ever was.

I just don't see a case unless they can define a core. AMD described the architecture as completely as possible without giving out trade secrets, they provided examples of performance you could expect, and listed the specifications afa cache, frequency, and number of cores. The processors were reviewed...what case do they really have? If it wasn't negligence, then how is AMD providing as much information as they can somehow misleading? Do they need to provide a disclaimer saying "hey, you know, this cpu may not perform as well as you expect it to in other workloads, because those other workloads utilize a different part of the CPU?" Should Intel be required to do that because it's inferior in certain workloads? How about VIA?

 

[Nobu]

I'm not saying this case is meaningless, just that I don't see any circumstance where AMD could be held liable here for any sort of damages.

 

[Tsumi]

I really hope you two are simply playing devil's advocate and not as detached from the civic process as your posts imply.

These aren't some downtrodden companies just trying to innovate--this is a problem of their own making. They deliberately obscured the difference between processor speeds, numbers, and then types in order to outsell the competition; and those are just the big bullet points.

This case won't define cores. Law doesn't operate like that. Law, like technology, is an iterative process. During this case, different interpretations, operational terms, and practical implications will be discussed as both sides try to untangle the web of what both Intel and AMD try to sell the public on what a "core" is and how that impacts computing experiences.

The sides will present their evidence for their position of what constitutes a core and why it does or doesn't matter, the judge will preside over whether that evidence is presented according to the rules, and the jury will determine how they understand the evidence. From there, another case will arise and another and another until we have a sufficient body of law to adequately and clearly define the points in contention.

That's how law and society interact with one another while jury members act as surrogates for the rest of us. It's a sloppy mechanism but it's the one we've got and some say it's the best there is or ever was.

Do you expect AMD and Intel to supply a chart of expected performance on literally every application in existence with every CPU? Because that is the only way they can do what you are asking for.

 

[Dick Johnson]

Windows should never be used as proof of anything regarding your hardware configuration. It was modified to show 32bit editions of Windows as having 4GB of RAM even though that wasn't available.

Windows reports your hardware configuration just fine. What you are talking about is something completely different.

The amount of RAM you have installed (actual physical RAM) and how much of it is actually usable by you, is not the same. The OS has to use some of it in order to function. Unfortunately, a lot of stupid people don't understand this, so, Microsoft got lots of complaints from people saying "Where's the rest of my RAM? I have 4GB installed but Windows says I only have 3.5GB".

Microsoft wasn't doing anything wrong, they were actually being honest for a change, and reporting how much RAM was available for you to actually use. But, after a while, Microsoft got tired of dealing with stupid people who don't understand how computers work, so, instead of showing you how much usable RAM you have, they changed it to just show you how much total RAM is installed.

 

[mope54]

I just don't see a case unless they can define a core. AMD described the architecture as completely as possible without giving out trade secrets, they provided examples of performance you could expect, and listed the specifications afa cache, frequency, and number of cores. The processors were reviewed...what case do they really have? If it wasn't negligence, then how is AMD providing as much information as they can somehow misleading? Do they need to provide a disclaimer saying "hey, you know, this cpu may not perform as well as you expect it to in other workloads, because those other workloads utilize a different part of the CPU?" Should Intel be required to do that because it's inferior in certain workloads? How about VIA?

My comments aren't about liability. I'm discussing how marketplace definitions arise in our society--at the intersection between law and society.

I'm not certain what or how disclaimers should be provided, but the notion you're introducing (in order to undermine corporate responsibility, strangely) has been the "what we know to be true" among industry insiders for the better part of decades now. It's time for the major players and the law to step in and determine what, if anything, needs to be disclosed and how it needs to be done. It's not enough any longer to simply say well cores don't matter (except when we market that they do) and speed ratings don't matter (except when we market that they do).

You keep making the same argument as, "well everyone already knows castor oil doesn't cure cancer. Why do we need a law telling people that they can't make that claim? If anyone is confused, just go read some medical journals about cancer and known cures, etc." That's not how our marketplace works. We don't live in a society where people selling things can just say whatever they want regardless of how true it is or regardless if the industry can't even say definitively what it actually is.

This whole conversation is bizarre. I suspect that everyone participating in this thread already knows that core speed and core count aren't directly relevant to system performance.

The shortest answer to your question are not endless caveats, but rather if you can't define a core (which seems to be where your argument landed) then you can't use it as a marketing strategy.

 

[Tsumi]

My comments aren't about liability. I'm discussing how marketplace definitions arise in our society--at the intersection between law and society.

I'm not certain what or how disclaimers should be provided, but the notion you're introducing (in order to undermine corporate responsibility, strangely) has been the "what we know to be true" among industry insiders for the better part of decades now. It's time for the major players and the law to step in and determine what, if anything, needs to be disclosed and how it needs to be done. It's not enough any longer to simply say well cores don't matter (except when we market that they do) and speed ratings don't matter (except when we market that they do).

You keep making the same argument as, "well everyone already knows castor oil doesn't cure cancer. Why do we need a law telling people that they can't make that claim? If anyone is confused, just go read some medical journals about cancer and known cures, etc." That's not how our marketplace works. We don't live in a society where people selling things can just say whatever they want regardless of how true it is or regardless if the industry can't even say definitively what it actually is.

This whole conversation is bizarre. I suspect that everyone participating in this thread already knows that core speed and core count aren't directly relevant to system performance.

The shortest answer to your question are not endless caveats, but rather if you can't define a core (which seems to be where your argument landed) then you can't use it as a marketing strategy.

If someone marketed castor oil as a cure for cancer, it is against the law because it is false advertisement. AMD's marketing of Bulldozer as an 8 core is not false advertisement, it technically has 8 cores. AMD did not guarantee a performance level when marketing it as 8 cores, and therefore cannot be held to any kind of false marketing claim.

To make another flawed car analogy, it's like Lexus marketing the RX as a competitor to the Mercedes ML. Both are luxury crossover SUVs. However, you have to do your due diligence to find out both are vastly different cars.

 

[mope54]

The public understands "cores" to mean something that is more defined than the industry uses the term. This is largely because the industry marketed "cores" to mean more than they do by promoting them as more defined than they are and promoting them as having more bearing on performance than they do.

The fact that the industry can define "cores" more loosely than the public doesn't absolve them from any guilt in promulgating the myth that "cores" matter in computing.

The reason there are laws against false advertising is because of cases like this. That's why castor oil can't be marketed like that--not because it just existed a priori! People used to advertise it as cures all the time...until cases like this occurred.

We went through the same thing with response time on monitors. We went through it with capacity of media. We went through it with internet speeds.


If you guys want a car analogy that works it's this: the same type of lawsuits were directed at the auto industry over horsepower claims, torque claims, efficiency claims, and safety claims. We're much better off for those lawsuits than previously when they could claim anything that approximated something in reality.

 

[Zarathustra[H]]

The public understands "cores" to mean something that is more defined than the industry uses the term. This is largely because the industry marketed "cores" to mean more than they do by promoting them as more defined than they are and promoting them as having more bearing on performance than they do.

The fact that the industry can define "cores" more loosely than the public doesn't absolve them from any guilt in promulgating the myth that "cores" matter in computing.

The reason there are laws against false advertising is because of cases like this. That's why castor oil can't be marketed like that--not because it just existed a priori! People used to advertise it as cures all the time...until cases like this occurred.

We went through the same thing with response time on monitors. We went through it with capacity of media. We went through it with internet speeds.


If you guys want a car analogy that works it's this: the same type of lawsuits were directed at the auto industry over horsepower claims, torque claims, efficiency claims, and safety claims. We're much better off for those lawsuits than previously when they could claim anything that approximated something in reality.

No.

A core is a very well defined technical term that refers to the integer portion of the CPU.

Absolutely everyone who bought an AMD FX CPU got exactly the core count they paid for.

They weren't the best performing cores, but they were real integer cores.

This is a completely frivolous lawsuit. I wasn't happy about how the Bulldozer and on architecture performed either, but there were launch day reviews, so no one should haveb been surprised at what they got.

This is the very example of a lawsuit which should be thrown out in preliminary hearings.

 

[Tsumi]

The public understands "cores" to mean something that is more defined than the industry uses the term. This is largely because the industry marketed "cores" to mean more than they do by promoting them as more defined than they are and promoting them as having more bearing on performance than they do.

The fact that the industry can define "cores" more loosely than the public doesn't absolve them from any guilt in promulgating the myth that "cores" matter in computing.

The reason there are laws against false advertising is because of cases like this. That's why castor oil can't be marketed like that--not because it just existed a priori! People used to advertise it as cures all the time...until cases like this occurred.

We went through the same thing with response time on monitors. We went through it with capacity of media. We went through it with internet speeds.


If you guys want a car analogy that works it's this: the same type of lawsuits were directed at the auto industry over horsepower claims, torque claims, efficiency claims, and safety claims. We're much better off for those lawsuits than previously when they could claim anything that approximated something in reality.

Except those lawsuits had objective and tangible targets and standards. A V8 is a V8 regardless of whether it's an anemic 240 HP 4.6 or a monster 480 HP 6.4. The 4.6 is well within its rights to be advertised as a V8. Similarly 8 cores is 8 cores. It will perform better in multithreaded tasks than a 6 core from the same architecture, but it isn't a given that it will perform better than 6 cores of a different architecture. With how varied workloads can be, it is near impossible to establish "performance indicators" like the automotive industry has. However, even the automotive industry does not publish torque and hp curves most of the time, and that is usually left to independent guys. Similarly, we have independent hardware reviewers to pick apart hardware performance.

 

[mope54]

No.

A core is a very well defined technical term that refers to the integer portion of the CPU.

Absolutely everyone who bought an AMD FX CPU got exactly the core count they paid for.

They weren't the best performing cores, but they were real integer cores.

This is a completely frivolous lawsuit. I wasn't happy about how the Bulldozer and on architecture performed either, but there were launch day reviews, so no one should haveb been surprised at what they got.

This is the very example of a lawsuit which should be thrown out in preliminary hearings.

"Cores" can't be well-defined if you think they mean one thing yet everyone else in this thread thinks they mean something else.

 

[Nobu]

My comments aren't about liability. I'm discussing how marketplace definitions arise in our society--at the intersection between law and society.

I'm not certain what or how disclaimers should be provided, but the notion you're introducing (in order to undermine corporate responsibility, strangely) has been the "what we know to be true" among industry insiders for the better part of decades now. It's time for the major players and the law to step in and determine what, if anything, needs to be disclosed and how it needs to be done. It's not enough any longer to simply say well cores don't matter (except when we market that they do) and speed ratings don't matter (except when we market that they do).

You keep making the same argument as, "well everyone already knows castor oil doesn't cure cancer. Why do we need a law telling people that they can't make that claim? If anyone is confused, just go read some medical journals about cancer and known cures, etc." That's not how our marketplace works. We don't live in a society where people selling things can just say whatever they want regardless of how true it is or regardless if the industry can't even say definitively what it actually is.

This whole conversation is bizarre. I suspect that everyone participating in this thread already knows that core speed and core count aren't directly relevant to system performance.

The shortest answer to your question are not endless caveats, but rather if you can't define a core (which seems to be where your argument landed) then you can't use it as a marketing strategy.

If that was my argument, I wouldn't have provided my own (simple, concise) definition. The issue is that cores are different between architectures, even within families of architectures. They include more or fewer instruction sets (x86 extensions: avx, 3dnow!, sse, etc.), fpus, cache...their layout differs, with some having cores further from certain caches (which affects cache access latency), among other differences. A core from 'a' company's 'x' processor cannot be compared directly to a core from 'b' company's 'y' processor without considering the design of those two processors individually, and how they perform in specific tasks, even if they come from the same company, sometimes even when they use the same architecture.

Now, I think they would have a case if amd had released an updated processor of the same architecture with more cores, but these cores shared resources whereas the previous had not, and AMD hadn't made that clear. But AMD did clearly layout the specifications of these processors, and it was a new architecture not based on a previous one. (Rather, not an update to a previous one, like zen+ to zen, or even zen2 to zen+)

 

[mope54]

Except those lawsuits had objective and tangible targets and standards. A V8 is a V8 regardless of whether it's an anemic 240 HP 4.6 or a monster 480 HP 6.4. The 4.6 is well within its rights to be advertised as a V8. Similarly 8 cores is 8 cores. It will perform better in multithreaded tasks than a 6 core from the same architecture, but it isn't a given that it will perform better than 6 cores of a different architecture. With how varied workloads can be, it is near impossible to establish "performance indicators" like the automotive industry has. However, even the automotive industry does not publish torque and hp curves most of the time, and that is usually left to independent guys. Similarly, we have independent hardware reviewers to pick apart hardware performance.

The objective and tangible targets came about because of the lawsuits. It's very strange how you continue to post about a thing that exists as if it always existed without acknowledging how it came to be.

I don't know which country you are in, but in the United States every single car is sold with it's HP and Torque claims loudly and proudly marketed. In the past car companies used to claim ridiculous HP and Torque until they were sued and now have to publish *when* that power occurs (lb-ft@rpms).

The same way with response time. Companies used to shift all kinds of claims about what mattered when it came to response time in a monitor until they had to step into court and codify what they meant. That led to the now standard of reporting "Grey to Grey." Industry didn't just decide to be less vague and more accurate with their marketing claims because people wanted them to, they did it because people made them through the court system.

Whether this particular case prevails isn't particularly relevant to the process I'm describing to you. The conversations that occur in the court will become part of the body of our law that will direct future court proceedings. Eventually that body of law will build up into a formal understanding of the thing ("core" in this case). That's different from informal understandings of a thing, which is where were are now.

 

[mope54]

The issue is that cores are different between architectures, even within families of architectures. They include more or fewer instruction sets (x86 extensions: avx, 3dnow!, sse, etc.), fpus, cache...their layout differs, with some having cores further from certain caches (which affects cache access latency), among other differences. A core from 'a' company's 'x' processor cannot be compared directly to a core from 'b' company's 'y' processor without considering the design of those two processors individually, and how they perform in specific tasks, even if they come from the same company, sometimes even when they use the same architecture.

The reason this is true is because industry has been allowed to market all kinds of definitions of what a "core" is and whether or to how much it matters.

All the stuff you listed that matters is because you have taken the time over the years to decode the technology from the marketing speak. Then an industry grew up around deciphering "performance" from these objective factors. An entire 3rd party industry exists for the sole purpose of keeping enthusiasts informed separate from what the tech companies themselves claim...marinade on that for a bit.

There is no good reason to cheer AMD in this. To the extent that people have a misperception about what a core actually is or what it does, that's solely due to AMD and Intel advertising those things and creating that imagine in people's minds. You are arguing for the continuance of an industry to play fast and loose with technological aspects and features of a quick moving technology.

 

[Nobu]

The reason this is true is because industry has been allowed to market all kinds of definitions of what a "core" is and whether or to how much it matters.

All the stuff you listed that matters is because you have taken the time over the years to decode the technology from the marketing speak. Then an industry grew up around deciphering "performance" from these objective factors. An entire 3rd party industry exists for the sole purpose of keeping enthusiasts informed separate from what the tech companies themselves claim...marinade on that for a bit.

There is no good reason to cheer AMD in this. To the extent that people have a misperception about what a core actually is or what it does, that's solely due to AMD and Intel advertising those things and creating that imagine in people's minds. You are arguing for the continuance of an industry to play fast and loose with technological aspects and features of a quick moving technology.

So what you think will happen is there will be an industry defined set of benchmarks which will be required in specifications for any computer processor? They don't put in avx and sse, etc just to pull a fast one. They have tangible performance benefits in the programs which make use of those operations. In the same way, the number of FPUs and ALUs, and their layout inside a core, as well as their access to cache all have affects on performance. When you use a cache heavy benchmark it will have a different performance level to one that uses little cache, when you are accessing data on disc it will again perform differently – and not just because of the CPU. It has to interoperate with all the other components of the system as well. You cannot broadly define a core because small changes can have a large impact on the entire system. You cannot have a standard set of benchmarks because they won't be applicable to my use case.

You can provide a simple definition of a core, which I and others have in this thread, but their case cannot be supported based on such a definition (or any definition, imho).

If you know of a way to wriggle out a clear definition of a complete core based on AMD or Intel or VIA or any other CPU design, please share.

 

[BinarySynapse]

To define what a "core" is, you have to define what a CPU is, good luck with that one. Is a product with 3 integer units and an FPU a CPU, what if there isnt an FPU. What if theres only one integer unit that can execute all integer instructions and another that can only execute a subset of them (e.g. the original Pentium). How what about caches, busses, branch predictors, registers, memory controllers, and voltage regulator.

Horsepower is a well defined term. So are torque, response time, Hz, miles and gallons. What was codified was the requirement to disclose HOW these things were measured, not what they meant. "Core" is as meaningless a term for performance or features as calories is for health. Which is healthier: a 220cal candy bar or a 300cal steak?

 

[mope54]

The case is built on the claim of definitive cores. The case is about what is inside the box when the outside of the box says, "8 cores."

The law does not state, "Cores means x, y, and z." Even AMD did not argue this. They argued that the industry understands "cores means x, y, and z" and sometimes Intel relies on x to sell cores and sometimes AMD relies on z to sell cores but they all are within the definition of the industry's use of "cores."

The jury will decide whether AMD was or was not close enough to a reasonable legal standard of what a "core" is and is not. Cases like this also produce written standards for the industry and consumers to draw upon when labels get placed on the box. It's not cut and dry or the case wouldn't have made it past the first post.

All those things you guys keep saying are well-defined happened because of cases like this. I listed them for a reason. It's kinda silly for you to just repost them and say nope without looking into it. If you don't know it, you should read the legal history.

If your argument is that you can't define cores, then you can't sell "cores."

Interesting argument about "calories." There's a reason we have food labels now...cases like this.

 

[Nobu]

Why should they (not just AMD) be required to conform to a specific architecture (that's what you're basically leading to) going forward? The reason they change the architecture in the first place is (1) to get better performance and (2) to improve efficiency. If you define an architecture which they must conform to then (1) processors will be much less diverse, and (2) it will be harder for them to improve those metrics (it's arguably hard enough as it is). It may even make it impossible to use a specific node, because having x and y feature next to each other makes it cost prohibitive, or because it's impossible to make a certain feature on x node so it has to be made elsewhere on a different node.

 

[mope54]

Why should they (not just AMD) be required to conform to a specific architecture (that's what you're basically leading to) going forward? The reason they change the architecture in the first place is (1) to get better performance and (2) to improve efficiency. If you define an architecture which they must conform to then (1) processors will be much less diverse, and (2) it will be harder for them to improve those metrics (it's arguably hard enough as it is). It may even make it impossible to use a specific node, because having x and y feature next to each other makes it cost prohibitive, or because it's impossible to make a certain feature on x node so it has to be made elsewhere on a different node.

No one is talking about defining architectures and nothing about this is relevant to R&D. This is about marketing. The marketing of cores was completely detached from its relevance in pretty much everything except the public understanding that if one is good, more are better.

Like I said, if you can't define what a core is, then it can't be marketed adequately.
How is that not common sense to you?

 

[BinarySynapse]

If your argument is that you can't define cores, then you can't sell "cores."

That I'll agree with, if only because I've grown tired of explaining why a "dual-core" i5 6200U is curb-stompong a "quad-core" A10-8700B.

 

[Nobu]

The case is built on the claim of definitive cores. The case is about what is inside the box when the outside of the box says, "8 cores."

The law does not state, "Cores means x, y, and z." Even AMD did not argue this. They argued that the industry understands "cores means x, y, and z" and sometimes Intel relies on x to sell cores and sometimes AMD relies on z to sell cores but they all are within the definition of the industry's use of "cores."

The jury will decide whether AMD was or was not close enough to a reasonable legal standard of what a "core" is and is not. Cases like this also produce written standards for the industry and consumers to draw upon when labels get placed on the box. It's not cut and dry or the case wouldn't have made it past the first post.

All those things you guys keep saying are well-defined happened because of cases like this. I listed them for a reason. It's kinda silly for you to just repost them and say nope without looking into it. If you don't know it, you should read the legal history.

If your argument is that you can't define cores, then you can't sell "cores."

In that case, it's pretty clear what will happen. It will be held that "cores" cannot be used to describe a part of a cpu, end of story. Likely they'll start using threads instead. Would that satisfy you?

Edit:accidentally cut out part of the quote that was relevant...

 

[OutOfPhase]

No one is talking about defining architectures and nothing about this is relevant to R&D. This is about marketing. The marketing of cores was completely detached from its relevance in pretty much everything except the public understanding that if one is good, more are better.

Like I said, if you can't define what a core is, then it can't be marketed adequately.
How is that not common sense to you?

Except when the same company sells a processor with 6 cores, and another in the same architecture which has 8 right? Then they can say 6 or 8?

 

[mope54]

In that case, it's pretty clear what will happen. It will be held that "cores" cannot be used to describe a part of a cpu, end of story. Likely they'll start using threads instead. Would that satisfy you?

This doesn't have anything to do with my satisfaction. It's up to juries and eventually legislatures, as surrogates for all of us within civil society. That's how law and society interacts in the US, which is what I've been describing to you.

I don't pay attention to any marketing claims. But I rely on that 3rd party industry to learn what I need to about architecture and performance. I don't understand why anyone thinks that it's ok for our industry to be exempt from what other industries having to be honest with their customers.

 

[Zarathustra[H]]

No one is talking about defining architectures and nothing about this is relevant to R&D. This is about marketing. The marketing of cores was completely detached from its relevance in pretty much everything except the public understanding that if one is good, more are better.

Like I said, if you can't define what a core is, then it can't be marketed adequately.
How is that not common sense to you?

There was nothing wrong with AMD's marketing, but even if there were, marketing is completely irrelevant in an industry where everyone buys based on performance test results reported by reviews.

You could print pretty much anything on the box, and it is irrelevant.

 

[Nobu]

This doesn't have anything to do with my satisfaction. It's up to juries and eventually legislatures, as surrogates for all of us within civil society. That's how law and society interacts in the US, which is what I've been describing to you.

I don't pay attention to any marketing claims. But I rely on that 3rd party industry to learn what I need to about architecture and performance. I don't understand why anyone thinks that it's ok for our industry to be exempt from what other industries having to be honest with their customers.

I don't see the dishonesty, is my issue. It does have x cores. They are x bulldozer cores, and VIA's Nano has x Isaiah cores, and Ryzen has x zen cores. In each architecture, that number has significance. It may not be 1:1, but more cores does in fact mean more performance in tasks which utilize them.

 

[Nobu]

My other issue is you arguing like none of the facts that have been brought up here will be brought up in court. Instead of arguing why they won't be or why they are irrelevant, you say we aren't there so it doesn't matter what we think. Instead of presenting possible outcomes or providing your own definition, you say "the court will decide so it doesn't matter." So then, why are you in this thread at all? It's not even clear to me what issue you have with AMD marketing (in this case specifically) to begin with...

Nevermind, I remember now. The posts were flying pretty fast and it slipped my mind.

 

[mope54]

you say "the court will decide so it doesn't matter."

I say the courts will help decide over time. These cases are part of the building blocks that constitute what will eventually be a codified definition of a core--something both industry and public can agree upon and rely upon when transacting. That provides safety during purchasing to the public and it provides safety to the industry that they won't be sued as long they operate within the well-defined boundaries that exist.

I never said your opinion doesn't matter. I said that in the system we have your opinion is being represented by jury members and elected officials. So if you want your opinion to have impact, then you need to make it known to your proxies. I think your opinion matters, but I don't think it's safe to assume that everyone in the marketplace shares the same opinion so if it's time to scrutinize CPU marketing terminology to ensure it's adequately presenting the technology then so be it.

 

[Nobu]

I say the courts will help decide over time. These cases are part of the building blocks that constitute what will eventually be a codified definition of a core--something both industry and public can agree upon and rely upon when transacting. That provides safety during purchasing to the public and it provides safety to the industry that they won't be sued as long they operate within the well-defined boundaries that exist.

I never said your opinion doesn't matter. I said that in the system we have your opinion is being represented by jury members and elected officials. So if you want your opinion to have impact, then you need to make it known to your proxies. I think your opinion matters, but I don't think it's safe to assume that everyone in the marketplace shares the same opinion so if it's time to scrutinize CPU marketing terminology to ensure it's adequately presenting the technology then so be it.

And I have no issue with that, I understand it entirely. I don't believe they will be able to define a core (other than maybe a simple definition like the one I provided before), and such a definition will not alleviate the issue you have with the term. The only other option I see is to go to more technical specifications like ALUs and FPUs, or threads. The issue with threads is they may not perform the same, depending on whether it is a "hyperthread" (or similar) or a "real" thread, though you could list them both I suppose. They would still be susceptible to the issue bulldozer had, regardless, as threads are executed on cores, and cores vary by architecture. The issue I have with just listing ALUs and FPUs is that doesn't tell you anything about the actual performance you'll get either. That will depend on how the architecture leverages those units, as well as cache, etc. That leaves you with just frequency, cache, etc.. Well, you have to convey the number of threads a cpu can process concurrently, or else you have all these skus with no discernible difference other than clock and cache, and for some reason the lower clocked skus perform better in these multithread workloads and consume a lot more power, plus they're more expensive....