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

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
AMD stands to lose millions for advertising its last-generation, FX-series “Bulldozer” CPUs as an eight-core chip. While this is arguably correct on a technical level, the novel implementation of two discrete cores in a single module led to a lawsuit filed in 2015 that argued otherwise, claiming it really had four. A California judge has now determined the argument has merit and is allowing the class-action lawsuit to go forward.

The so-called "eight core" chips contain four Bulldozer modules, the lawsuit notes, and these "sub-processors" each contain a pair of instruction-executing CPU cores. So, four modules times two CPU cores equals, in AMD's mind, eight CPU cores. And here's the sticking point: these two CPU cores, within a single Bulldozer module, share caches, frontend circuitry, and a single floating point unit (FPU). These shared resources cause bottlenecks that can slow the processor, it is claimed.
 
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
 
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

I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs. Would this equate to something like NVIDIAs Vram issue where some of it doesn't exactly perform as good as the rest of the module?
 
I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs. Would this equate to something like NVIDIAs Vram issue where some of it doesn't exactly perform as good as the rest of the module?

you don't get to define what the designers determine is a core, amd took a gamble and said 'this will be good' turns out it wasn't exactly all that good, it wasn't because they claimed it was a 8 core but infact it wasn't, it was because it was just a poor design, combined with lack of proper software, everything was being designed according to what intel determined was a proper cpu, it was encoded to work well on intel cpus...

again they are saying because of how amd designed it, it wasn't a 8 core....

I mean, I am not exactly educated on how all this goes, but from my understanding they are basicly claiming, a V8 engine isn't a v8 engine, if it can be beat by a 4 cylinder, or 6 cylinder....



da fuck?

oh and don't get me wrong, im not supporting the bulldozer, imo it was a absolute piece of shit..... I stayed clear away from amd since the duron/Athlon/athlonxp days...…. I only recently came back with the 2000 series ryzen cpus..


but im also not sitting here claiming it wasn't a 8 core cpu….
 
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
Let not mix the merit of the lawsuit with your misguided negative view of class action lawsuits.
As far as the lawsuit goes, the only thing it shows so far, is how uneducated, probably old yhe judge is/ was.
I thought 2x4 equals eight?
 
you don't get to define what the designers determine is a core, amd took a gamble and said 'this will be good' turns out it wasn't exactly all that good, it wasn't because they claimed it was a 8 core but infact it wasn't, it was because it was just a poor design, combined with lack of proper software, everything was being designed according to what intel determined was a proper cpu, it was encoded to work well on intel cpus...

again they are saying because of how amd designed it, it wasn't a 8 core....

I mean, I am not exactly educated on how all this goes, but from my understanding they are basicly claiming, a V8 engine isn't a v8 engine, if it can be beat by a 4 cylinder, or 6 cylinder....



da fuck?

oh and don't get me wrong, im not supporting the bulldozer, imo it was a absolute piece of shit..... I stayed clear away from amd since the duron/Athlon/athlonxp days...…. I only recently came back with the 2000 series ryzen cpus..


but im also not sitting here claiming it wasn't a 8 core cpu….

I didn't read the article, but it appears...
It's more like saying a V8 iant a V8 if put two pistons in a single cylinder.

Advertising and marketing representations matter. Especially when it's a technical product that may be poorly understood by the customer base. I do not see this as merit-less. I think it's important that companies find the edges of the boundaries from time-to-time.
 
I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs. Would this equate to something like NVIDIAs Vram issue where some of it doesn't exactly perform as good as the rest of the module?

Performance doesn't define what is a core. If you really want to go down that rabbit hole, if an Intel i5-2500k is defined as having 4 core performance, a Core2Quad will be defined as a dual core. Or ARM cores will only be considered as half cores, and an Intel i5-8600k would be an 8 core CPU. See how ridiculous it is to use performance as an indicator as what a core is?

Bulldozer had 8 integer units with 4 256-bit FPUs that can act as 8 independent FPUs under the then common 128-bit wide workloads. It was hampered by a front-end that can't feed the cores fast enough when both were being utilized, but it still underperformed in 4 core workloads as well when compared to Intel. The core design was inherently poor, and taking it out of the module design would not have fixed that.
 
I don't think the case has merit. It does have 8 cores, it's just that the cores share more resources than other designs, as AMD designed the chip around maximizing multi-threading performance and less of a reliance on FPU as more of that was being offloaded to the GPU. Their vision of software development over the next 5-10 years unfortunately didn't pan out and the chip didn't perform as well as expected. I don't feel like it was marketed intentionally as misleading, and there is no de facto standard for what constitutes a core, so I'm not sure how they can argue otherwise.
 
I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs.

Bulldozer had 8 separate integer cores and did well in multi-threaded workloads where they could actually be put to use. Their weak single-threaded performance was ultimately the trade-off for having eight small cores instead of four large cores on one die. There's no evidence that the shared FPU was a major bottleneck.

Beyond that, we also now know from the Spectre/Meltdown fiasco that the performance advantage of Intel products largely came at the expense of major security holes. The patches kill performance on older Intel machines, if they're even available for the customer's hardware.
 
The only thing I can say at this point is that AMD has already suffered enough. They have paid for their architectural mistakes. This lawsuit is a sham. It's shameful to see a class action lawsuit attempt penalize any company on the basis that some people feel butt hurt over having bought one of these chips. I built no less than two of these machines and they weren't bad, they just weren't even as fast as the Phenom II x6 I had.

So, if this suit is actually "won" does that mean that any other company that builds a CPU with a similar stripped core/shared resources architecture would have to advertise the processor as half the cores? So, AMD is moving to chiplets... Is some asshat going to come out of the woodwork and say that one chiplet is 1 CPU?

Better yet... Wait until someone starts arguing about the CCX modules on the Current Ryzen...

AMD has paid enough already.
 
Last edited:
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

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 didn't read the article, but it appears...
It's more like saying a V8 iant a V8 if put two pistons in a single cylinder.

Advertising and marketing representations matter. Especially when it's a technical product that may be poorly understood by the customer base. I do not see this as merit-less. I think it's important that companies find the edges of the boundaries from time-to-time.

a better analogy would be an auto manufacturer saying an engine is DOHC because it has a single cam shaft on each head. that is NOT DOHC by any ones stretch of the word and a dual overhead cam shaft is TWO actual cam shafts in each head. you can rework the meaning of DUAL in this all you want but its accepted practice that it is two actual cam shafts and in a v-whatever engine that would be a total of 4 cam shafts and in a straightline that would be 2 physical cam shafts.

all performance aside, its really going to come down to does the bulldozer qualify as 8 cores and if not, did AMD misrepresent that?

as for anyone saying you dont get to say what is or isnt a core- thats not the point. the point is that a company does not get to misrepresent what they are selling as something it is not. and these kind of lawsuits do more than just fund a lawyers bmw and cocaine funds (although lets be honest... thats where all the money goes and we all know it).
 
Core is not a performance specification. They are saying they can run 8-wide, and they can. I'd personally call it an 8 thread CPU, but that's not really a useful distinction to the consumer.

Any expectation of performance should be validated prior to purchase. I don't look to see how many threads/cores/shaders are in GPUs - I look at what performance comes out for expected workloads. We should treat CPUs the same way.
 
Not that I would Sue AMD for it

My point on this the fx 8 core assist cpus was 4modules with 2 threads per module

Same with so called 6 core fx cpus and 4 core type A and cpus

Even windows 10 lists fx cpus as 4 core 8 logical just to make sure different workload we're not put on like the same to cores on the same module as there was a 50% performance penalty for doing it

The bulldozer architecture is basically using real cores as smt/HT and should be sold based on there module count not there 50/50% core limited per module streaming and gaming shows this when total cpu load goes past 50% (it be like selling smt/HT cpus as actual cores)
 
Last edited:
I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs. Would this equate to something like NVIDIAs Vram issue where some of it doesn't exactly perform as good as the rest of the module?

In simple terms, AMD is counting 8 schedulers and 8 integer units as processors. There are people that disagree with this thinking that these 8 cores are incomplete because they don't feature dedicated resources such as L2 cache and they only have 4 FPUs between them. One for each pair of integer units. L2 cache and FPUs aren't intrinsic to the x86 or compatible designs. While you wouldn't imagine a CPU without them today, they weren't always part of the deal. L2 cache first showed up on motherboards and at one time you had to purchase x87 (MathCo, FPU's) separately. You can make a better argument that the decoders and fetch hardware / logic is shared between cores making for incomplete integer units. Again, you would have to clearly define what a CPU core is or isn't for that standard to apply. Right now, it doesn't.
 
a better analogy would be an auto manufacturer saying an engine is DOHC because it has a single cam shaft on each head. that is NOT DOHC by any ones stretch of the word and a dual overhead cam shaft is TWO actual cam shafts in each head. you can rework the meaning of DUAL in this all you want but its accepted practice that it is two actual cam shafts and in a v-whatever engine that would be a total of 4 cam shafts and in a straightline that would be 2 physical cam shafts.

all performance aside, its really going to come down to does the bulldozer qualify as 8 cores and if not, did AMD misrepresent that?

as for anyone saying you dont get to say what is or isnt a core- thats not the point. the point is that a company does not get to misrepresent what they are selling as something it is not. and these kind of lawsuits do more than just fund a lawyers bmw and cocaine funds (although lets be honest... thats where all the money goes and we all know it).

How is it not 8 cores though? There are 8 integer units.

The more accurate automotive example (though still inaccurate) is like getting a Ford 4.6 V8 and then wondering why it doesn't perform as well as a Chevy 6.2 V8 because all V8s are equal, right?

Not that I would Sue AMD for it

My point on this the fx 8 core assist cpus was 4modules with 2 threads per module

Same with so called 6 core fx cpus and 4 core type A and cpus

Even windows 10 lists fx cpus as 4 core 8 logical just to make sure different workload we're not put on like the same to cores on the same module as there was a 50% performance penalty for doing it

The bulldozer architecture is basically using real cores as smt/HT and should be sold based on there module count not there 50/50% core limited per module streaming and gaming shows this when total cpu load goes past 50% (it be like selling smt/HT cpus as actual cores)

That's simply lazy programming in Windows, and should not be used as proof that it isn't 8 "real" cores. Windows only did that to prevent overloading the front end. The hyperthreading design is having two front ends feed a core. Entirely different beasts.
 
AMD stands to lose millions for advertising its last-generation, FX-series “Bulldozer” CPUs as an eight-core chip. While this is arguably correct on a technical level, the novel implementation of two discrete cores in a single module led to a lawsuit filed in 2015 that argued otherwise, claiming it really had four. A California judge has now determined the argument has merit and is allowing the class-action lawsuit to go forward.

The so-called "eight core" chips contain four Bulldozer modules, the lawsuit notes, and these "sub-processors" each contain a pair of instruction-executing CPU cores. So, four modules times two CPU cores equals, in AMD's mind, eight CPU cores. And here's the sticking point: these two CPU cores, within a single Bulldozer module, share caches, frontend circuitry, and a single floating point unit (FPU). These shared resources cause bottlenecks that can slow the processor, it is claimed.

Technically it is 8 cores and there is no arguing out of that whatsoever. A core has to fetch, decode, execute and retire. If 2 cores as a pair share fetch and decode that doesn't matter they're still cores. There's no rule they have to have a floating point unit either. Or any caches, branch prediction or a memory controller. What technically constitutes a core is very, very little.
 
I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs. Would this equate to something like NVIDIAs Vram issue where some of it doesn't exactly perform as good as the rest of the module?

"I do admit, i dont know much about how CPUs are exactly designed and made"

you answered yourself.
 
How is it not 8 cores though? There are 8 integer units.

The more accurate automotive example (though still inaccurate) is like getting a Ford 4.6 V8 and then wondering why it doesn't perform as well as a Chevy 6.2 V8 because all V8s are equal, right?



That's simply lazy programming in Windows, and should not be used as proof that it isn't 8 "real" cores. Windows only did that to prevent overloading the front end. The hyperthreading design is having two front ends feed a core. Entirely different beasts.

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. Lots of adjustments have been made to Windows in order to work with various hardware over the years and by itself is proof of nothing.

Technically it is 8 cores and there is no arguing out of that whatsoever. A core has to fetch, decode, execute and retire. If 2 cores as a pair share fetch and decode that doesn't matter they're still cores. There's no rule they have to have a floating point unit either. Or any caches, branch prediction or a memory controller. What technically constitutes a core is very, very little.

Exactly.
 
I am sorry, maybe I'm a bit lacking in this area, but isn't this the same as saying the q6600 isn't a real quad core processor even though it technically has 4 cores in a 2x2 die config glued together and in the same fashion that Ryzen is also, but I can see how the technical interconnect limitations really limit the processors capability, but that doesn't mean in reality that it isn't an 8core processor as it has 8cores just hampered with a quad core interconnect. I mean a core is a core, grant it not all are the same, but will a courtesy determines it actually matters?
 
I'll bite, can you elaborate? I read the article and as far as i can tell it has some Merritt. I do admit, i dont know much about how CPUs are exactly designed and made, but from what I can tell in the article the CPU is bottle-necked by the design and doesn't have much of a performance increase over traditional 4 core CPUs. Would this equate to something like NVIDIAs Vram issue where some of it doesn't exactly perform as good as the rest of the module?
Have you ever gone into a computer-store and been curious how many CPUs it has so you hit control+shift+escape, open it's task manager and click on the processors tab telling you the processor load of each core? If you did that with a 4-core HT Intel processor, it's going to show 8 cores because 2 cores are simulated for hyperthreading for each real physical core. So it looks like 8 but it's actually '2' virtual cores being fed into '1' physical core with a little processing module designed to speed up processing via hyperthreading.

If Intel started to advertise 8 cores after hyperthreading, most people who know about computers would be aware that 8 virtual cores 'after' hyperthreading is the same as '4' physical cores before hyperthreading that perform slightly better. People who know nothing about computers might be confused and think they are getting a 8 physical cores. The problem is when you advertise "8 cores!" and don't really specify that clearly....virtual or physical...people get confused. The onus imo is kind of on the consumer to do a tiny bit of research and maybe look up a review. Every review on a processor that involves hyperthreading usually has a very brief synopsis explaining what I said above.

However, if the company says 8 cores and they meant virtual and you as a customer never bothered to check a review but you do technically get 8 virtual cores, you got 8 cores. It just wasn't what you were expecting. You expected to get an amazing, high-end(for the time) 8 core processor for cheap as 8 cores from Intel were like $1000+ for were believing amd would sell you something close for $200 and instead got a perfectly competent midrange processor for cheap ($300-$400 ish for an Intel equivlanet value) for $200. If something sounds that good to be true, and it turns out not to be true, it's akin the 'omg, my coffee was hot!' situation except in that case it was a pretty legitimate lawsuit. That coffee was heated far too hot and gave the lady severe third degree burns requiring skin graphs.
 
Technically it is 8 cores and there is no arguing out of that whatsoever. A core has to fetch, decode, execute and retire. If 2 cores as a pair share fetch and decode that doesn't matter they're still cores. There's no rule they have to have a floating point unit either. Or any caches, branch prediction or a memory controller. What technically constitutes a core is very, very little.

This is technically called a thread. The relationship between a thread and execution units varies from architecture to architecture.

Most people seem to want to call a cluster of execution dedicated to M number of threads (usually 1-2) a "core". That works until you get to the architectures I worked on most recently, where there is nothing dedicated to anything. It's M threads as far as the OS is concerned, but we have a huge (nondedicated) bank of integer, FP, specialized units behind that. Is that just one core? It's clearly not in any real way.

This is why I just want people to look at actual performance instead of trying to count things.
 
Seen a lot of people who don't really understand that they don't actually have an 8 core cpu (i7 with HT as they don't get what HT is or fx with core assist witch is more complicated to explain as it does have 8 cores but they are more like ht/smt threads)

explaining why there fx cpu is not really 8 core but really 4 core with core assist (or especially the A10 witch is more a duel core then 4 core) and why your game fps tanks is because cpu total utilisation is higher then 50% is slightly difficult (but they norm get it once I mention its a 4 module cpu with 2 mini cores per module and they share resources on the module and performs about 50% less when both are Been Used and why they getting inconsistent performance)

one good thing is you can on most motherboards was change the module to cpu count from 4m/8c to 4m/4c so all resources go into each one core and is not shared
 
How is it not 8 cores though? There are 8 integer units.

The more accurate automotive example (though still inaccurate) is like getting a Ford 4.6 V8 and then wondering why it doesn't perform as well as a Chevy 6.2 V8 because all V8s are equal, right?



That's simply lazy programming in Windows, and should not be used as proof that it isn't 8 "real" cores. Windows only did that to prevent overloading the front end. The hyperthreading design is having two front ends feed a core. Entirely different beasts.

If we're going to stick with flawed car analogies, you should compare a push-rod V8 with a carburetor to a DOHC V8 with sequential port injection.



Core means absolutely jack and squat. It only exists because in 2004 AMD needed a way to illustrate the cost and performance benefits of their new dual-core products against already established single processor and multiple processor products. They could have called Bulldozer a 12-core processor and it would have just as much meaning as calling it a single-core.
 
I can always tell there are a lot of gear heads on this forum because we all use car analogies :D
 
  • Like
Reactions: N4CR
like this
me the judge: I find in favor of the plaintiffs and award that in restitution AMD must deliver a competitive high-end gpu.
amd: what?
me the judge: what?
 
AMD suffering already is irrelevant to the lawsuit.

That said....

As a former 8750 owner user it worked just fine for everything I needed it to do.

A Pair of Pants has two legs and is one item.... I wonder if some inane logic like that is in play.
 
I am sorry, maybe I'm a bit lacking in this area, but isn't this the same as saying the q6600 isn't a real quad core processor even though it technically has 4 cores in a 2x2 die config glued together and in the same fashion that Ryzen is also, but I can see how the technical interconnect limitations really limit the processors capability, but that doesn't mean in reality that it isn't an 8core processor as it has 8cores just hampered with a quad core interconnect. I mean a core is a core, grant it not all are the same, but will a courtesy determines it actually matters?

AMD used to say that Intel's CPU's were a non-native design with that very distinction in mind. People were smart enough to know that didn't mean jack shit. The numbers spoke for themselves.
 
AMD is going to get their asses handed to them. Not because of the merit of the suit, no, it's going to be because the jury is made up of the same fool asses that would have bought into their marketing.
 
AMD used to say that Intel's CPU's were a non-native design with that very distinction in mind. People were smart enough to know that didn't mean jack shit. The numbers spoke for themselves.
Well I mean the Q6600 absolutely rocked as a great processor(3.0ghz with a thermaltake V1, on mine), Bulldozer was uhm, when I worked in a PC shop nobody had anything kind to say about them, at all. Even Vishera wasnt exactly looked at in a good light, on paper it looked good, but in physical product.......no one wanted it and alot jumped to Intel if they didn't already. We all gave Ryzen a try, and I'm impressed...,..for now, as are most of my friends.
 
I can always tell there are a lot of gear heads on this forum because we all use car analogies :D

Normally "car analogy" is a euphemism for "completely irrelevant anecdote".

As a delightful counterpoint, ryan_975 is spot on, as always.
 
Back
Top