AMD Agrees To Pay Out $35 Per Chip Over FX Marketing Lawsuit

[mr moose]

11 hours ago, leadeater said:

 

The section you came back in on, didn't think I needed to specify that. Spiel or not at no point did I say it would end development only that it would impact designs going forward and restrict what designers may do. You don't need to misrepresent what was said either, you're falsely advertising what I said .

Whats the difference between impact designs and restricting designers and ending development?   One might sound absolute I'll grant, but in the end I still see not mechanism by which a definition would have any impact on development or design.  There was a legal definition for mail and that did not impact the development of email, I am sure there are plenty of things in the digital world that legal definitions exist for that have wero impact on their design and development.

 

11 hours ago, leadeater said:

The point was about needing to know what the product actually is in the terms existing operating systems and software use. This so called cold hard definition you speak of. Like I said the proposal put to me was to just start calling cores what ever the company pleases, as exampled, which is silly. That is the post of mine in reply to someone else you decided to quote a section of.

Because it doesn't sound right that an engineer of any description would have trouble with marketing semantics,  and what a company calls a core in marketing is often different to what they call it in the spec sheet.

[leadeater]

2 hours ago, mr moose said:

Whats the difference between impact designs and restricting designers and ending development?   One might sound absolute I'll grant, but in the end I still see not mechanism by which a definition would have any impact on development or design.  There was a legal definition for mail and that did not impact the development of email, I am sure there are plenty of things in the digital world that legal definitions exist for that have wero impact on their design and development.

Mail and email isn't really that related here, it would be more like saying an envelope isn't a envelope because it has 2 opening flaps, one on each end or sides. You could make something actually different, like email, but it would no longer be an x86 processor that operating systems and software understand. Mail and email is to what x86 is to PowerPC. So sure you could make something new and different but you'll have all the problems I talked about.

 

2 hours ago, mr moose said:

Because it doesn't sound right that an engineer of any description would have trouble with marketing semantics,  and what a company calls a core in marketing is often different to what they call it in the spec sheet.

They wouldn't have trouble with marketing semantics I'm just saying they may not do something that would necessitate changing it's name in marketing. If I do this these things that to me are two cores and function as two cores would have to be marketed as one core.

[leadeater]

2 hours ago, TechyBen said:

What would? Having twice the floats, or being honest to consumers? Yes... being honest does not require faster parts.

Because two single FPUs with one 128bit FMAC is the same performance as one FPU with 2 128bit FMACs in it. The number of FPUs doesn't tell you the performance unless you know the performance of the FPUs. Two FPUs being evaluated that have different performance, for example FPU A twice that of the other now referred to as FPU B, would have different performance when using the same number of them. 8 of FPU A is faster than 8 of FPU B, 4 of FPU A is the same as 8 of FPU B. All you need to do is look at Zen and Zen+, those FPUs in AVX2 have half the performance of Intel FPUs.

 

Honestly isn't the problem here, not when you're trying to say it is the number of FPUs or the sharing of them between cores. You want honesty in the marketing and performance claims, number of FPUs isn't going to do that.

[leadeater]

2 hours ago, TechyBen said:

AFAIK nope. You only need to show that the advertising was not aligned to the consumer expectation. The consumers might (wrongfully or rightfully) expect 5-10% increase year on year, if previous generations offered that "faster/new", for it to then be -5% of "equivalent" specs as labeled on the box, then consumers might go "huh, what happened?" and expect the box/seller to confirm the difference.

 

Take TVs going from square to widescreen but the shops using the  *same* metric that does not translate, diagonal size. So consumers rightfully felt "cheated" as a widescreen was "smaller" than the square versions horizontally. For you and me, who know pixel sizes and all that jazz, we are fine. The other 99% of consumers though felt rather sold out, as the companies did not instead use new metrics, or explain the new sizes ("15 inch but x high").

So exactly what I just said?

 

2 hours ago, TechyBen said:

The marketing may have said "faster", if it was not faster than the previous model, like for like packaging, then it needed to state "6 core faster than 4, or 8 core faster than 6" Not "faster" as the consumer would assume Phen 4 core was slower than the Bulldozer 6 core.

 

Yes, it may only be 10% or 5% of consumers who actually noticed/missed the white paper. But for consumers, they are not expected to be experts. It's not failure rates, or wear and tear, but out of the box specifications here, so it stands for potential for every customer.

You might want to read the paragraph again, you're agreeing with that I was saying but it doesn't sound like it?

[mr moose]

5 hours ago, leadeater said:

Mail and email isn't really that related here, it would be more like saying an envelope isn't a envelope because it has 2 opening flaps, one on each end or sides. You could make something actually different, like email, but it would no longer be an x86 processor that operating systems and software understand. Mail and email is to what x86 is to PowerPC. So sure you could make something new and different but you'll have all the problems I talked about.

What difference though, they are both mail just like the processors are both processors.   How would there be problems, there are no problems with calling email mail and there were no problems with developing it and promoting it and seeing it take over.  I just don't see how having a legal definition will impinge upon things from evolving and changing.

 

5 hours ago, leadeater said:

They wouldn't have trouble with marketing semantics I'm just saying they may not do something that would necessitate changing it's name in marketing. If I do this these things that to me are two cores and function as two cores would have to be marketed as one core.

 

So you think they would intentionally not improve a design because it would mean they couldn't call it a core and would have to call it something else? I think that is quite the stretch.

[leadeater]

44 minutes ago, mr moose said:

What difference though, they are both mail just like the processors are both processors.   How would there be problems, there are no problems with calling email mail and there were no problems with developing it and promoting it and seeing it take over.  I just don't see how having a legal definition will impinge upon things from evolving and changing.

Because the reality is this is far too much of an over simplification. If shared L2 cache between cores means marketing wise you have to say it is one core would you do it? Just calling it something else is no where near as simple as being tried to be made out. Is it still 1 core in the OS? How do you properly allocate work to something that actually is 2 cores but needs to be shown as 1? Treat it as SMT? Problem is there is a difference between a core and a thread to both an OS and software.

 

You can create all the names you like but they need to carry meaning and for it to carry meaning things need to understand what it is, which means developers need to build in that understanding, something already hard enough and poorly done on the Windows side already and that's just working with a single object of understanding i.e. a CPU core.

 

Edit:

I could create an envelope with 2 flaps and just call it a dualvalope but if this new design gets stuck in every mail sorting machine it's going to create huge problems isn't it?

 

44 minutes ago, mr moose said:

So you think they would intentionally not improve a design because it would mean they couldn't call it a core and would have to call it something else? I think that is quite the stretch.

I didn't say they wouldn't improve or be able to improve, there may be areas that things just aren't done or done in a certain way. There are many restrictions that exist in this world but that doesn't stop all progress, it also doesn't mean those have no impact either.

[mr moose]

25 minutes ago, leadeater said:

Because the reality is this is far too much of an over simplification. If shared L2 cache between cores means marketing wise you have to say it is one core would you do it?

Yes, I would call it a processor with X number of (insert specific attribute here) processing units, or I would call it 4 dual Int cores with shared L2 cache, or you could even call it 8 Int processors sharing 4 L2 cache units.  Because the reality is those who care about the tech specs will already know what it all means and those who rely on the marketing will simply have a more realistic and comparable description.   Because everyone would have to refer to each unit the same way and couldn't bend the truth a little to suit their marketing.

 

25 minutes ago, leadeater said:

Just calling it something else is no where near as simple as being tried to be made out. Is it still 1 core in the OS? How do you properly allocate work to something that actually is 2 cores but needs to be shown as 1? Treat it as SMT? Problem is there is a difference between a core and a thread to both an OS and software.

Ity's not hard to make a footnote in the advertising or link to product page that explains how many cores will be seen in the OS. 

25 minutes ago, leadeater said:

You can create all the names you like but they need to carry meaning and for it to carry meaning things need to understand what it is, which means developers need to build in that understanding, something already hard enough and poorly done on the Windows side already and that's just working with a single object of understanding i.e. a CPU core.

developers don;t have to do jack shit,  it's all about sales spiels and marketing,  nothing in this effects developers. they just keep developing better products and the marketer works out how to sell it without breaching consumer law.

25 minutes ago, leadeater said:

Edit:

I could create an envelope with 2 flaps and just call it a dualvalope but if this new design gets stuck in every mail sorting machine it's going to create huge problems isn't it?

Then don't sell it to the public claiming it is better.

25 minutes ago, leadeater said:

I didn't say they wouldn't improve or be able to improve, there may be areas that things just aren't done or done in a certain way.

Whats the difference?  If you are claiming that things won't get done or get done differently then you need to explain what they might be, because I fail to see what they would do differently just because marketing might be limited in what they can call it.

 

25 minutes ago, leadeater said:

There are many restrictions that exist in this world but that doesn't stop all progress, it also doesn't mean those have no impact either.

 

I have yet to see a legal definition of anything prevent progress.  I certainly can't think of any.  I don't know a company that decided not to make or improve something because it didn't fit a legal definition.

[leadeater]

25 minutes ago, mr moose said:

Yes, I would call it a processor with X number of (insert specific attribute here) processing units, or I would call it 4 dual Int cores with shared L2 cache, or you could even call it 8 Int processors sharing 4 L2 cache units.  Because the reality is those who care about the tech specs will already know what it all means and those who rely on the marketing will simply have a more realistic and comparable description.   Because everyone would have to refer to each unit the same way and couldn't bend the truth a little to suit their marketing.

And why would I not consider this misleading advertising? My OS shows me a different number [edit: or name] to whats on the box so what gives? What are you trying to hide with this new name that I don't understand?

 

If people are going to sue because they didn't understand what 'Shared Floating Point Accelerator' entailed or didn't see it or the felt it was improperly disclosed/explained and felt mislead then why wouldn't they sue if the product description doesn't match what their operating system and software is showing them? That isn't the full issue but it's a big part of the complaint as far as the supporting argument goes within it.

 

I don't necessarily object to product marketing such as that but it doesn't bring any more clarity and would increase confusion and for the people that actually care would want to know the legacy number of cores, which isn't a problem because you would just put that on the product spec sheet as these people would look for it. Also in a meeting with everyone being technically trained you'd just talk about it in the existing terms as is and ignore the new naming (not a problem).

 

In all it doesn't protect you from customer complaints and may actually make it more likely.

 

25 minutes ago, mr moose said:

I have yet to see a legal definition of anything prevent progress.  I certainly can't think of any.  I don't know a company that decided not to make or improve something because it didn't fit a legal definition.

If you are still saying something would not improve then you haven't at all understood the point. No popup headlights exist anymore because of car safety laws, there may be some very rare exceptions I don't know of but manufacturers aren't making any cars with them. Similarly they aren't making cars with chassis structures and hood heights below certain heights. It's not stopping manufactures from making improvements in designs but there are things they are not doing.

[Trixanity]

What's interesting is that the retail box itself holds no false advertisement so you would have to look elsewhere to be mislead. So you could go into a store and pick up an FX processor, be disappointed when you got home and still be a victim of your own doing.

 

On the topic of misleading advertisement you should look no further than Qualcomm:

 

It's very subtle but you'd get the impression that you're buying what would be essentially 8 A76 cores when in reality they're calling both A55 and A76 (four each) the same name: Kryo 485. The clock speed advertisement is saved by the 'up to' as only one core goes that high.

 

I've yet to see a class action lawsuit against Qualcomm despite the significant performance delta. I still see this as being an opportunistic money grab although the advertisement could be better and more accurate when it comes to some of the claims. However I still question how many buyers had actually seen much if any of the advertisement.

 

[mr moose]

21 minutes ago, leadeater said:

And why would I not consider this misleading advertising? My OS shows me a different number [edit: or name] to whats on the box so what gives? What are you trying to hide with this new name that I don't understand?

It doesn't have to be like that.  There is nothing in forcing comapnies to be honest in the advertising or defining what a core is that would result in that.  I even said this:

 

37 minutes ago, mr moose said:

Yes, I would call it a processor with X number of (insert specific attribute here) processing units, or I would call it 4 dual Int cores with shared L2 cache, or you could even call it 8 Int processors sharing 4 L2 cache units.  Because the reality is those who care about the tech specs will already know what it all means and those who rely on the marketing will simply have a more realistic and comparable description.   Because everyone would have to refer to each unit the same way and couldn't bend the truth a little to suit their marketing.

 

Ity's not hard to make a footnote in the advertising or link to product page that explains how many cores will be seen in the OS. 

 

 

 

Quote

If people are going to sue because they didn't understand what 'Share Floating Point Accelerator' entailed or didn't see it or the felt it was improperly disclosed/explained and felt mislead then why wouldn't they sue if the product description doesn't match what their operating system and software is showing them? That isn't the full issue but it's a big part of the complaint as far as the supporting argument goes within it.

No one sued because they didn't understand what a floating point accelerator is or even improperly explained, they are suing because the implied performance as a result of being the only one with said number of cores was misleading.

 

Quote

I don't necessarily object to product marketing such as that but it doesn't bring any more clarity and would increase confusion and for the people that actually care would want to know the legacy number of cores, which isn't a problem because you would just put that on the product spec sheet as these people would look for it. Also in a meeting with everyone being technically trained you'd just talk about it in the existing terms as is and ignore the new naming (not a problem).

 

In all it doesn't protect you from customer complaints and my actually make it more likely.

I really fail to see how forcing companies to be more precise with their claims would result in higher rates of confusion.  In fact the more we talk about it the more i would be in favor of defining cores etc so that you could directly compare Intel's claims to AMD's claims at a spec level.  At the moment you can;t do that because they have many different definitions for certain specs. TDP springs to mind (not that neither AMD nor Intel actually use that in their marketing and it is mainly tech nerds who get it wrong.

 

Quote

If you are still saying something would not improve then you haven't at all understood the point. No popup headlights exist anymore because of car safety laws,

Popup headlights were became rare because they were dangerous, not because of a legal definition that meant designers couldn't keep them legal and cheap to manufacture.  That is not the same as designers not making something because they can't call it a "core".

 

Quote

there may be some very rare exceptions I don't know of but manufacturers aren't making any cars with them. Similarly they aren't making cars with chassis structures and hood heights below certain heights. It's not stopping manufactures from making improvements in designs but there are things they are not doing.

Again because of safety concerns with the physical design when they hit pedestrians.  They have not stopped making them due to being forced to call them something different.  

 

 

[leadeater]

16 minutes ago, mr moose said:

No one sued because they didn't understand what a floating point accelerator is or even improperly explained, they are suing because the implied performance as a result of being the only one with said number of cores was misleading.

Then why is it central to the filed suit that the processors had 4 cores and not 8 cores due to the shared FPU and shared L2 cache? The plaintiffs were very adamant it was 4 cores not 8. Their given technical reasons and explanations were factually inaccurate as to why the performance was bad, AMD would obviously get their chance to counter argue but that doesn't mean it'll get sided with them (or should). The case is about the marketing so where is the need to go in to those specifics?

 

Edit:

And again why would I not sue over your proposed naming? It's not any less confusing or misleading if product marketing doesn't match what you actually see/get.

[TechyBen]

8 hours ago, leadeater said:

Because two single FPUs with one 128bit FMAC is the same performance as one FPU with 2 128bit FMACs in it. The number of FPUs doesn't tell you the performance unless you know the performance of the FPUs. Two FPUs being evaluated that have different performance, for example FPU A twice that of the other now referred to as FPU B, would have different performance when using the same number of them. 8 of FPU A is faster than 8 of FPU B, 4 of FPU A is the same as 8 of FPU B. All you need to do is look at Zen and Zen+, those FPUs in AVX2 have half the performance of Intel FPUs.

 

Honestly isn't the problem here, not when you're trying to say it is the number of FPUs or the sharing of them between cores. You want honesty in the marketing and performance claims, number of FPUs isn't going to do that.

I never asked for the performance. I asked for honesty.

 

A turbo charger does not make my car faster. If it is a small, low pressure one, and I reduce the number of cylinders and the fuel flow. A large exhaust does not make it faster, if I limit the airflow into the engine.

 

But guess what, legally, I expect the seller to tell me if the car is turbocharged or not.

 

I'll bow out of the rest of the argument. As this fact is clearly not understood.

 

[mr moose]

48 minutes ago, leadeater said:

Then why is it central to the filed suit that the processors had 4 cores and not 8 cores due to the shared FPU and shared L2 cache? The plaintiffs were very adamant it was 4 cores not 8.

You'll have to ask the lawyers that. it looks to me like it just furthered the case given that the marketing claimed heavily leaned on the number of cores being the inferred factor for performance.  As I said right back at the start though,  this isn't really a technical debate about cores or what constitutes a core, it is a marketing one.

48 minutes ago, leadeater said:

Their given technical reasons and explanations were factually inaccurate as to why the performance was bad, AMD would obviously get their chance to counter argue but that doesn't mean it'll get sided with them (or should). The case is about the marketing so where is the need to go in to those specifics?

Oooh I just said that in the above paragraph.  I guess we probably agree more than it appears.

48 minutes ago, leadeater said:

Edit:

And again why would I not sue over your proposed naming? It's not any less confusing or misleading if product marketing doesn't match what you actually see/get.

 

Because if they call it what it is then they can't be done for false advertising, The only problem AMD had in this whole affair was that they went too far with the marketing guff.  For the record I still believe it had 8 cores and they could have defended that in court, but they shot themselves in the foot with their claim it was due to having 8 cores that they were superior.

[leadeater]

2 minutes ago, TechyBen said:

I never asked for the performance. I asked for honesty.

From what I saw you where thinking just the number of FPUs were materially important information to the performance of the product. I think we can both agree that it's not as simple as that which was what I was trying to explain.

 

I wouldn't expect AMD to have said that Bulldozer was as outright bad as it was in such direct terms but I agree claims about superior performance was highly unrealistic, they were struggling to out perform their previous products with less cores.

[leadeater]

4 minutes ago, mr moose said:

Because if they call it what it is then they can't be done for false advertising

Then what was it? Would it also match what your OS is telling you? I don't think a simple change of naming is quite as simple and free from potential issues as you want it to be.

 

7 minutes ago, mr moose said:

You'll have to ask the lawyers that. it looks to me like it just furthered the case given that the marketing claimed heavily leaned on the number of cores being the inferred factor for performance.  As I said right back at the start though,  this isn't really a technical debate about cores or what constitutes a core, it is a marketing one.

I still don't see how it is necessary. Like I've said before you can prove the performance regression of the cores without doing so. Your solution to my issue with the case that would have gone to the court would be to let it happen without modification of the filing on the basis that it may not result in an outcome that would put legal definition to what isn't a core. My solution is to modify it so it doesn't require discussing so it couldn't possibly happen. For my issue I'll take mine, not possible is better, or at the very least far less likely.

[TechyBen]

10 minutes ago, leadeater said:

From what I saw you where thinking just the number of FPUs were materially important information to the performance of the product. I think we can both agree that it's not as simple as that which was what I was trying to explain.

 

I wouldn't expect AMD to have said that Bulldozer was as outright bad as it was in such direct terms but I agree claims about superior performance was highly unrealistic, they were struggling to out perform their previous products with less cores.

I understand that in *some* instances it was worse on the performance. "Materially" checking performance in your example depends on *how* we check it, not on an agreed metric. But to the customer expectations. They changed how a product worked. The consumer expected it worked the same. [edit] The consumer said "you did not tell me the number of FPUs", and while this may have no effect of performance, it is an example of, them not listing it on the box. This we agreed on, that it had changed, and was not listed as a significant change. Material "proof", though not necessarily the exact one affecting every customer. [/edit]

 

 

Intel releasing a poor product, ATOM chips, or Xe (I forget the name, those FPU thingies) never went to court. As they did not advertise them as "better", or just release the box for customers to buy in store, and find it was poor.

 

Even a chocolate company could be taken to court for refunds. If they change the recipie, and so tons of people walk in store, buy it, and it tastes horrible. They "sue" for "misleading", as previously, it tasted different, and were not warned/informed it had changed. Chocolate companies generally don't get this problem, because "taste guarantee, get a refund whenever you are unhappy" etc.

 

Often courts find extra info (the iPhone screen case, the news reported they were refurb screens, then later in court turns out they may have been actual counterfeits/excess Apple stock sold on the side), or they may argue the case incorrectly (McDonalds vs the poor old lady with burns was McDs arguing they were not liable in a legal definition, even though they were at fault for excessively hot food serving).

 

We may not agree with the consumer or AMD. But we certainly agree AMD were not totally honest in marketing. Not totally correct in their product lineup/segmentation. And so have seen consumer confusion and backlash because of it.

 

See the current naming schemes. AMD R73800 cpu. NVidia 1650. While stupid names, they are marketed and segmented into "5% or 10%" performance boosts. Releasing a "new £/$500 NVidia PXQ3900 our best ever GPU!" and it perform around a RTX2700... is not a problem of performance, technology, or "a bad product", but misleading "best ever" advertising.

 

[leadeater]

17 minutes ago, mr moose said:

Oooh I just said that in the above paragraph.  I guess we probably agree more than it appears.

I was talking about the plaintiffs, they are factually incorrect in their explanations. You think I'll ever be happy for such a thing to exist in a court case?

 

Edit:

Never mind I see the part you mean.

[mr moose]

3 minutes ago, leadeater said:

Then what was it? Would it also match what your OS is telling you? I don't think a simple change of naming is quite as simple and free from potential issues as you want it to be.

 

As I said before, it's not hard to note how many cores windows will see. If windows reports 4 cores then the ad fine print says "windows will report four cores due to the way it handles threads/cores/rotation of the earth". They already do it for many other specs that are effected by 3rd party software or other hardware.

 

3 minutes ago, leadeater said:

I still don't see how it is necessary. Like I've said before you can prove the performance regression of the cores without doing so. Your solution to my issue with the case that would have gone to the court would be to let it happen without modification of the filing on the basis that it may not result in an outcome that would put legal definition to what isn't a core. My solution is to modify it so it doesn't require discussing so it couldn't possibly happen. For my issue I'll take mine, not possible is better, or at the very least far less likely.

 

No, I would let it go to court on the basis that even if they did define what a core is for legal purposes it would have zero effect on anything except marketing and advertising.  The opposite of what you are saying is also very possible, had there been a legal definition in place and AMD abides by that, then this whole case would have evolved only around the claim that their cores were unmatched and what that meant to the general public.  In fact it is just as fair to say that with a legal definition of a core this case may not have even been a thing.

[leadeater]

9 minutes ago, mr moose said:

In fact it is just as fair to say that with a legal definition of a core this case may not have even been a thing.

If a definition is to exist that could be used in a legal setting then it should be done by an industry standards alliance comprised of industry stakeholders like other industry standards around products and protocols.

 

So yes the case would very likely not have happened had this been a thing but it wasn't, the court should be used when someone violates an agreed industry standard but the problem is there isn't one.