Friday, November 6th 2015

AMD Dragged to Court over Core Count on "Bulldozer"

This had to happen eventually. AMD has been dragged to court over misrepresentation of its CPU core count in its "Bulldozer" architecture. Tony Dickey, representing himself in the U.S. District Court for the Northern District of California, accused AMD of falsely advertising the core count in its latest CPUs, and contended that because of they way they're physically structured, AMD's 8-core "Bulldozer" chips really only have four cores.

The lawsuit alleges that Bulldozer processors were designed by stripping away components from two cores and combining what was left to make a single "module." In doing so, however, the cores no longer work independently. Due to this, AMD Bulldozer cannot perform eight instructions simultaneously and independently as claimed, or the way a true 8-core CPU would. Dickey is suing for damages, including statutory and punitive damages, litigation expenses, pre- and post-judgment interest, as well as other injunctive and declaratory relief as is deemed reasonable.
Source: LegalNewsOnline
Add your own comment

511 Comments on AMD Dragged to Court over Core Count on "Bulldozer"

#76
boogerlad
FordGT90ConceptPretty sure he's going to win. I don't think there's any nomenclature to properly describe Bulldozer's design and even if it had existed, AMD wasn't using it.


x264 HD Benchmark runs on GPU and AMD undeniably has a stronger GPU in FX-8150 than Intel has in i7-2600K. The problem stems from floating point operations executed on the CPU. If you heavily load the FPUs in one core, the FPU performance of both cores will effectively half.
You're hilarious. x264 only uses opencl for frame look ahead and even then, the boost is very small. There is no gpu in the fx-8150 anyways.
Posted on Reply
#77
FordGT90Concept
"I go fast!1!11!1!"
lilhasselhoffer3) How do you prove it?
Block diagrams. K8 versus Bulldozer.
lilhasselhoffer4) What experts support your claims?
Any "expert" that can interpret the block diagrams before the court. It likely doesn't require much explaining that Bulldozer is missing a lot of parts in the module to constitute two discreet logic processors.
lilhasselhoffer5) How are damages being calculated?
Dickeys likely gave a number using his own formula. Court will have to decide if that formula is fair or not.
lilhasselhoffer6) Does the judge have an unbiased opinion?
Judges aren't supposed to be biased. If they have a bias, they're supposed to recuse themselves.
boogerladYou're hilarious. x264 only uses opencl for frame look ahead and even then, the boost is very small. There is no gpu in the fx-8150 anyways.
Tech ARP x264 HD Benchmark, from what I was able to research, doesn't explain its process methods. All that is abundantly clear is that it benefits from more cores (highest scores go to Xeons with many cores). Judging by the benchmarks, it appears that it is heavily ALU oriented which plays to Bulldozer's benefit.
Posted on Reply
#78
moproblems99
FordGT90ConceptTech ARP x264 HD Benchmark, from what I was able to research, doesn't explain its process methods. All that is abundantly clear is that it benefits from more cores (highest scores go to Xeons with many cores). Judging by the benchmarks, it appears that it is heavily ALU oriented which plays to Bulldozer's benefit.
So doesn't that just corroborate that a Bulldozer has 8 cores? The more cores the higher the scores?
Posted on Reply
#80
m0nt3
By this same metric, shouldn't he take Intel to court over the 286/386 math co-processor. Which was a floating point unit......
Posted on Reply
#81
moproblems99
FordGT90ConceptNo because a) it chokes when you feed it floating points and b) a two-way 4-core K8-based Opteron workstation will beat the "8-core" bulldozer in almost every way.
Will it beat it in Tech ARP x264 HD Benchmark that you stated above benefited from more cores? I'm not trolling, or my intent anyway. Just trying to figure out your thought process.
Posted on Reply
#82
FordGT90Concept
"I go fast!1!11!1!"
I edited and the answer is yes. It only takes six K10 cores to beat eight Bulldozer "cores" even at significantly lower clocks (2.8 GHz versus 3.6 GHz).

I see it coming: "oh, but Bulldozer is technically only 4-core so a 6-core should be it!" My point, exactly.
m0nt3By this same metric, shouldn't he take Intel to court over the 286/386 math co-processor. Which was a floating point unit......
FPUs were spotty around 1990s simply because it was brand new technology. You could argue Bulldozer was brand new technology too but, at that point, the definition of "core" was pretty well established for 6 years prior to that. The use of the word "core" where it isn't appropriate is why this lawsuit has merit.
Posted on Reply
#83
moproblems99
FordGT90ConceptI edited and the answer is yes. It only takes six K10 cores to beat eight Bulldozer "cores" even at significantly lower clocks (2.8 GHz versus 3.6 GHz).

I see it coming: "oh, but Bulldozer is technically only 4-core so a 6-core should be it!" My point, exactly.



FPUs were spotty around 1990s simply because it was brand new technology. You could argue Bulldozer was brand new technology too but, at that point, the definition of "core" was pretty well established for 6 years prior to that. The use of the word "core" where it isn't appropriate is why this lawsuit has merit.
I see your point but honestly, if someone buy things solely on advertising, they deserve what they get.
Posted on Reply
#84
dorsetknob
"YOUR RMA REQUEST IS CON-REFUSED"
FordGT90Concept6 real cores are better than 4 cores masquerading as 8
And that's why i went Xeon rather than I7

Lawsuit in America = I Farted now i'm being sued in a class Action Because i polluted breathing Air:)
Posted on Reply
#85
lilhasselhoffer
FordGT90Concept...
Any "expert" that can interpret the block diagrams before the court. It likely doesn't require much explaining that Bulldozer is missing a lot of parts in the module to constitute two discreet logic processors.


Dickeys likely gave a number using his own formula. Court will have to decide if that formula is fair or not.


Judges aren't supposed to be biased. If they have a bias, they're supposed to recuse themselves....
The counter arguments are simple. Who defined that a CPU required certain things? Your own earlier statements comparing various architectures prove that the point being made is invalid. You can't justify that a component is necessary, unless you can prove it directly influences end results, which they can't reliably do if even one instance proves the contrary. The plaintiff accuses AMD of removing a critical component, yet demonstrably it is not critical. Kinda hard to have an argument when the basis for said argument is impossible to justify.

Dickey is full of crap here, based on claims. This is a civil suit, and filed based upon a California law which doesn't have many parallels universally recognized throughout the country. As others have stated, this guy is basically taking what may be a couple of hundred dollars of processor and suing AMD for it, magically lost time, legal bills, and everything else. I'm sorry, but if this was actually about lost performance that is being claimed they'd have something more than that. I understand that a judge will only consider the plaintiff's request, but there's a difference between bargaining like this is a used car lot and asking for fair reparations. Whenever somebody asks for $5, and the cost of the original product was $1 they've got to either have an exceptional case or exceptional proof. Their "proof" as yet is a bunch of technical data sheets and block diagrams. www.pacermonitor.com/public/case/9674725/Dickey_v_Advanced_Micro_Devices,_Inc Hell, the filing fee in this court is $400, which could have bought a new system with an Intel quad core. This isn't about helping consumers, and the money proves it.

Really? I understand wanting to believe that judges have no bias, but what sort of world do you live in? The one I live in has people being named as judges. These people have motivations, such as seeing the best thing done for their community, and delivering their own form of "justice." To the former, suing an ailing company into the ground will have a negative impact on locals. If this were MS, Samsung, or Seagate I'd be less concerned with impartiality. To the later, you have to weigh timing. This person made no effort to get refunded, waited until years after official marketing material was released, hired lawyers from Chicago to represent them in Oakland, and has yet to show any desire or interest in the public good. I'm sorry, but with all that easily demonstrable, it's impossible for a technological hermit to not have an underlying bias when dealing with someone. Judges are human, above all other things.



One last point here. Intel had Pentium 4, and the nutburst...ahem...netburst issues. They got sued, so theoretically you can use that as a basis for the AMD suit. Except, you can't. The reason Intel lost that suit was they manipulated benchmarks to sell their product. They LIED to customers:www.zacks.com/stock/news/153085/intel-settles-pentium-4-lawsuit-by-paying-15-to-customers AMD didn't lie. They may have been optimistic to think that changing the architecture around would allow performance to universally be better, but they released benchmarks which were confirmed by outside sources. Yes, calling them octo-cores is sleazy, but it isn't a lie or marketing altering the truth. AMD's already paid for Bulldozer being a turd with years of poor sales, this is an opportunist trying to make money because AMD is likely to settle and make this go away. Zen is too big of a component of AMD's future to allow a pending lawsuit to tarnish the name. The Chicago lawyers know that, and they're using it to get functionally free money.

Again, read through the lawyer's own page. If you don't want to punch them in the face afterwards you're a far more tolerant person than I.


Edit:
FordGT90ConceptI edited and the answer is yes. It only takes six K10 cores to beat eight Bulldozer "cores" even at significantly lower clocks (2.8 GHz versus 3.6 GHz).

I see it coming: "oh, but Bulldozer is technically only 4-core so a 6-core should be it!" My point, exactly.



FPUs were spotty around 1990s simply because it was brand new technology. You could argue Bulldozer was brand new technology too but, at that point, the definition of "core" was pretty well established for 6 years prior to that. The use of the word "core" where it isn't appropriate is why this lawsuit has merit.
So let me get this straight.

On one hand the plaintiff is smart enough to know what components a core entails, based upon the CPU architecture.

On the other hand, the plaintiff is not responsible enough to seek out any information on what is advertised as a completely new architecture. They are assumed to never have seen any information from 2009 to 2015 (look back to the Anandtech link I posted).


This person exists in such a narrow bubble of knowledge and ignorance that they can't possibly exist. It's be like saying a person has eaten hamburgers their entire life, and because of the name they assumed that they were made out of pork. They are now suing McDonalds because they were in fact a unique branch of Hindu, and killing pigs was acceptable but killing cows wasn't.

To say that preposterous statement hurts my cognitive faculties. They want me to drive a rusty spoon through my brain and scoop out my frontal lobe. The US is full of stupid lawsuits, but that doesn't mean we need to find the few examples of when they're true.
Posted on Reply
#86
Pill Monster
Bulldozer does have 8 cores but only 4 L2 cache chips so they are arranged in pairs (4 modules).

Which if anyone is intersted is the reason why MS released a hotfix for the scheduler. in W7


Default core scheduling in Windows is 1,2,3,4,5,6,7 but BD/PD ideal scheduling is 1,3,5,7,2,4,6,8 due to the shared cache. They are real cores btw, HT is purely logical.

Just some fyi. :p
Posted on Reply
#87
truth teller
Bulldozer cannot perform eight instructions simultaneously and independently as claimed, or the way a true 8-core CPU would
it will be dismissed once amd proves that even a "true dual core" processor is able to execute more than 8 instruction in parallel (instruction pipelining, out-of-order execution, etc.), still they deserve the bad rep for advertising modules as real cores (they are closer to real cores than HT modules, but nonetheless still not full cores)
Posted on Reply
#88
HumanSmoke
dorsetknobLawsuit in America = I Farted now i'm being sued in a class Action Because i polluted breathing Air:)
Unless you own up to it, you'll probably get taken to court by the EPA for violation of hazardous natural gas disposal.
Posted on Reply
#89
lilhasselhoffer
HumanSmokeUnless you own up to it, you'll probably get taken to court by the EPA for violation of hazardous natural gas disposal.
Good god. Your jokes are surprisingly close to the truth:
www3.epa.gov/airquality/sulfurdioxide/

75 ppb is the official emissions level which cannot be exceeded. This is why the EPA is a good idea, but so poorly implemented as to be a joke.

Edit:
Also @HumanSmoke, you should watch your jokes. From Wikipedia: Since New Zealand produces large amounts of agricultural products, it is in the unique position of having high methane emissions from livestock compared to other greenhouse gas sources. The New Zealand government is a signatory to the Kyoto Protocol and therefore attempts are being made to reduce greenhouse emissions. To achieve this, an agricultural emissions research levy was proposed, which promptly became known as a "fart tax" or "flatulence tax". It encountered opposition from farmers, farming lobby groups and opposition politicians

en.wikipedia.org/wiki/Flatulence

I'd be crying, if I wasn't laughing so hard.
Posted on Reply
#90
dorsetknob
"YOUR RMA REQUEST IS CON-REFUSED"
lilhasselhoffer75 ppb is the official emissions level which cannot be exceeded.
I suspect every fart contains more than 75 parts per billion of Methane
and so for you Americans i fully expect the EPA to prosecute (on behalf of Obunnya) every american for Air Pollution
lets just call this a living Tax to help clear the Deficit :p
Posted on Reply
#91
HumanSmoke
lilhasselhofferGood god. Your jokes are surprisingly close to the truth:
www3.epa.gov/airquality/sulfurdioxide/

75 ppb is the official emissions level which cannot be exceeded. This is why the EPA is a good idea, but so poorly implemented as to be a joke.

Edit:
Also @HumanSmoke, you should watch your jokes. From Wikipedia: Since New Zealand produces large amounts of agricultural products, it is in the unique position of having high methane emissions from livestock compared to other greenhouse gas sources. The New Zealand government is a signatory to the Kyoto Protocol and therefore attempts are being made to reduce greenhouse emissions. To achieve this, an agricultural emissions research levy was proposed, which promptly became known as a "fart tax" or "flatulence tax". It encountered opposition from farmers, farming lobby groups and opposition politicians

en.wikipedia.org/wiki/Flatulence

I'd be crying, if I wasn't laughing so hard.
Yep, this country has had love/hate relationship with its primary producers for years.
Anyhow, I think a discussion on farts makes more sense than whether Mr. Dickey thinks a CPU core has to intrinsically execute one floating point operation per cycle. Can't say I've ever seen that as a prerequisite of a CPU core....which is why I haven't been taking this thread at all seriously.
Posted on Reply
#92
Steevo
So some jackass thinks he will win a technical lawsuit against the worlds second largest (see what I did there?) X86-64 CPU producer for his own feels?


That's a bold move cotton, lets see how it plays out!
Posted on Reply
#93
FordGT90Concept
"I go fast!1!11!1!"
lilhasselhofferThe plaintiff accuses AMD of removing a critical component, yet demonstrably it is not critical.
It is critical to differentiate between uni-core and dual-core. Bulldozer is the only processor I know of that shares compute resources among cores. The rest generally only share memory.
lilhasselhofferThey LIED to customers:www.zacks.com/stock/news/153085/intel-settles-pentium-4-lawsuit-by-paying-15-to-customers AMD didn't lie.
How is Bulldozer "8-core" processors not lying? They're quad cores, with SMT, and an extra integer cluster. Literally the only difference between HyperThreading and Bulldozer is the addition of extra integer cluster. We don't call any processors that feature SMT by the thread count so why does Bulldozer get a pass?

Picture John Doe walking into [insert computer store here] and tells the clerk I want an 8-core processor. The clerk hooks John Doe up with a Bulldozer. He gets home and starts encoding videos on it. He quickly discovers it is no faster than his old Phenom II X6 1055T and starts looking for the reason. He stumbles upon threads like this, block diagrams of Bulldozer, reviews saying Bulldozer underperforms, benchmarks proving the poor performance, and--most importantly--he discovers Intel Core i7-5960X which thoroughly trounces his Bulldozer "8-core." How does John Doe not feel that he was mislead by the clerk, whom was mislead by AMD calling their processors "8-core?"
lilhasselhofferThey may have been optimistic to think that changing the architecture around would allow performance to universally be better, but they released benchmarks which were confirmed by outside sources.
AMD was thinking DirectCompute would negate the need for FPUs. AMD had a sense of euphoria after buying out ATI thinking that it will drastically change how computing is done. They couldn't have had it more wrong.
lilhasselhofferOn one hand the plaintiff is smart enough to know what components a core entails, based upon the CPU architecture.
Plaintiffs don't walk into lawsuits not doing their research. There is plenty of failure analyses all over the internet explaining why Bulldozer is a steaming pile of shit.

I'm not going to discuss (rather, attack) the plaintiff. Like I said, there is merit to the complaint and I'm shocked it wasn't done much sooner.
Pill MonsterBulldozer does have 8 cores but only 4 L2 cache chips so they are arranged in pairs (4 modules).
There's only 4 L2 caches because there is only 4 cores. The two threads running on the same core require access to all of the L2 because the required data can exist anywhere in there.
Pill MonsterThey are real cores btw, HT is purely logical.
HT has hardware just like Bulldozer. The only major difference between HD and Bulldozer is AMD added some hardware to the SMT implementation so that integer performance does not suffer. It really shows their lack of knowledge of SMT; hence the horrible implementation. Jim Keller, whom knows a thing or three about SMT came in to set AMD straight with Zen. More cores with SMT is better than cores with SMT that has extra hardware attached.
SteevoSo some jackass thinks he will win a technical lawsuit against the worlds second largest (see what I did there?) X86-64 CPU producer for his own feels?


That's a bold move cotton, lets see how it plays out!
Again, I cite the hard drive lawsuit. Seagate had the technical win (correct use of "GB") but still lost on the surface (Windows doesn't show what Seagate claims). I think Dickey has the technical (>50% of the core is shared between two threads causing bottlenecks) and surface win (nothing suggests Bulldozer "8-core" is really an 8-core).

AMD will try to use "8 integer cores" as a defense. It won't stick because outside of technical documents, "integer" is left out.
Posted on Reply
#94
RealNeil
dorsetknobLawsuit in America = I Farted now i'm being sued in a class Action Because i polluted breathing Air:)
HumanSmokeUnless you own up to it, you'll probably get taken to court by the EPA for violation of hazardous natural gas disposal.
Or they'll make you walk around with a tiny flame next to your ass to burn it off at once. :eek:
Posted on Reply
#95
Steevo
FordGT90ConceptIt is critical to differentiate between uni-core and dual-core. Bulldozer is the only processor I know of that shares compute resources among cores. The rest generally only share memory.


How is Bulldozer "8-core" processors not lying? They're quad cores, with SMT, and an extra integer cluster. Literally the only difference between HyperThreading and Bulldozer is the addition of extra integer cluster. We don't call any processors that feature SMT by the thread count so why does Bulldozer get a pass?

Picture John Doe walking into [insert computer store here] and tells the clerk I want an 8-core processor. The clerk hooks John Doe up with a Bulldozer. He gets home and starts encoding videos on it. He quickly discovers it is no faster than his old Phenom II X6 1055T and starts looking for the reason. He stumbles upon threads like this, block diagrams of Bulldozer, reviews saying Bulldozer underperforms, benchmarks proving the poor performance, and--most importantly--he discovers Intel Core i7-5960X which thoroughly trounces his Bulldozer "8-core." How does John Doe not feel that he was mislead by the clerk, whom was mislead by AMD calling their processors "8-core?"


AMD was thinking DirectCompute would negate the need for FPUs. AMD had a sense of euphoria after buying out ATI thinking that it will drastically change how computing is done. They couldn't have had it more wrong.



Plaintiffs don't walk into lawsuits not doing their research. There is plenty of failure analyses all over the internet explaining why Bulldozer is a steaming pile of shit.

I'm not going to discuss (rather, attack) the plaintiff. Like I said, there is merit to the complaint and I'm shocked it wasn't done much sooner.


There's only 4 L2 caches because there is only 4 cores. The two threads running on the same core require access to all of the L2 because the required data can exist anywhere in there.


HT has hardware just like Bulldozer. The only major difference between HD and Bulldozer is AMD added some hardware to the SMT implementation so that integer performance does not suffer. It really shows their lack of knowledge of SMT; hence the horrible implementation. Jim Keller, whom knows a thing or three about SMT came in to set AMD straight with Zen. More cores with SMT is better than cores with SMT that has extra hardware attached.



Again, I cite the hard drive lawsuit. Seagate had the technical win (correct use of "GB") but still lost on the surface (Windows doesn't show what Seagate claims). I think Dickey has the technical (>50% of the core is shared between two threads causing bottlenecks) and surface win (nothing suggests Bulldozer "8-core" is really an 8-core).

AMD will try to use "8 integer cores" as a defense. It won't stick because outside of technical documents, "integer" is left out.
WIndows sees 8 cores, and launches 8 hardware accessible threads, so by the same token that Seagate lost......
Posted on Reply
#96
FordGT90Concept
"I go fast!1!11!1!"
They're "logical processors," not "cores." Windows, as far as I know, only uses "core" in relation to processor power management...and that setting is hidden via the system registry.
Posted on Reply
#97
the54thvoid
Intoxicated Moderator
FordGT90ConceptThey're "logical processors," not "cores." Windows, as far as I know, only uses "core" in relation to processor power management...and that setting is hidden via the system registry.
In the end it doesn't matter what any of us here say.

AMD say this:

There are two independent integer cores on a single Bulldozer module

Plaintiff says:

Not

I'm outta here.
Posted on Reply
#98
Uplink10
the54thvoidAMD say this:

There are two independent integer cores on a single Bulldozer module
FordGT90ConceptThere's only one question here and it is this: are the two logical cores of the Bulldozer design separable? If yes, they are "cores." If not, the combined unit is a "core" with SMT.
Uplink10If it is possible to disable 7/5/3/1 core(s) in a firmware then he should lose but if it is not then he should win.
Posted on Reply
#99
R-T-B
If it is possible to disable 7/5/3/1 core(s) in a firmware then he should lose but if it is not then he should win.
That's a good point. I'll agree with that conclusion.
Posted on Reply
#100
Pill Monster
FordGT90ConceptThere's only 4 L2 caches because there is only 4 cores. The two threads running on the same core require access to all of the L2 because the required data can exist anywhere in there.


HT has hardware just like Bulldozer. The only major difference between HD and Bulldozer is AMD added some hardware to the SMT implementation so that integer performance does not suffer. It really shows their lack of knowledge of SMT; hence the horrible implementation. Jim Keller, whom knows a thing or three about SMT came in to set AMD straight with Zen. More cores with SMT is better than cores with SMT that has extra hardware attached.
t.
Nah man there are 8 cores, 8 integer cores if want to break it down, but 8 physical units which can execute 8 threads similtaneeously.
HT isn't hardware multithreading HT is based around the OS scheduler which can schedule 2 threads to one core using spare cycles, something like that.. I don't knopw the exact science but it's done in software anyway..


The main differnce between BD and Thurban is Thurban has 6 dedicated L2 cache banks each with 1 fetch/decode unit, that's basically it. L2/L1 ram is freaking expensive, plus it takes up room on the die.

Btw the 2 way shared L2 is one of Piledrivers biggest handicaps, if not the biggest. Round trip time between CPU and L3 is super slow, about double that of Thurban or Deneb.

Piledriver has around 27ns L3 latencey, Phenom II is what, 8ns or something? I could check.....
Hence overclocking L3 does nothing to improve performance as it did on Phenom II..

If you put an X6 and an 8320 head to head I imagine the X6 will rape the 8320, as you pointed out earlier, but only up to 6 threads. After that PD is gonna pull ahead.

I wouldn't be surprised if x264 wasn't running 8 threads...? Did they say how many?
But hey that was 4-5 years ago, now with everything mutithreaded......things have changed the archetecture was way ahead of it's time......

Also corect me uif I'm wrong but wasn't BD originally designed as a server chip?




sorry about my spelling, spellcheck isn't working.. :p
Posted on Reply
Add your own comment
Apr 26th, 2024 20:52 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts