Bulldozer Core-Count Debate Comes Back to Haunt AMD

[TheGuruStud]

LOL, the problems started long before Phenoms. Management didnt lead anything, chip design team got hung out to dry, and the PR dept. was doing more harm than good, then came Phenoms...

Phenom II was good, just slow to arrive. 4 ghz OC from the cheapo chip waa spanking intel.

 

[bug]

Meh, the design of Bulldozer was known well before its launch, I'm not sure why anyone would fret about how to count the cores.
What's next, suing because the L3 cache is shared, so overall, a "core" does not get the fully dedicated resource?

 

[silentbogo]

Meh, the design of Bulldozer was known well before its launch, I'm not sure why anyone would fret about how to count the cores.

To tech-educated audience, not to an average consumer.

What's next, suing because the L3 cache is shared, so overall, a "core" does not get the fully dedicated resource?

Some CPUs don't have L3, but this does not make them "not CPUs". The issue is not whether it "counts" or "does not count" as a CPU core, but the fact that AMD failed to convey this information to consumers. E.g. if you write "8 cores" on the box, the consumer will buy the CPU and be disappointed later when he/she finds out that it may not be 8 cores at all. If you write 4 modules [8 cores] it won't make it significantly less appealing, but it will make the potential buyer curious about these "modules" and at least do a quick google-fu to see what this is all about.
When you have such a mess, it's always open for interpretation. Some may say "yes, it's 8 cores with shared components", some (like me) will say that this is 4 cores with extra stuffin'.
Sun did use similar approach in the past to make their CPUs more multithreading-capable, and while each core had 2 integer units and 1 floating point unit it was still being called one CPU core (capable of running 8 threads).

 

[londiste]

Even Execution Unit or Integer Unit does not necessarily mean the same thing across different architectures. For example, Bulldozer Integer Unit had 2 ALUs (Arithmetic Logic Unit) and 2 AGUs (Address Generation Unit) while Zen's Integer Unit has 4 ALUs (even though with a bit more resticted set of operations if I remember correctly) and 2 AGUs. Zen core has effectively the same integer calculation capacity in its Integer Unit that Bulldozer has.

 

[noel_fs]

I bet the price people paid for those "8 core" processor wasnt 800$ so im willing to say that its pretty fair. Nowdays motherboards companies call anything a phase so it is how it is.

 

[newtekie1]

The plaintiffs and defendants each have a key technical argument. The plaintiffs could point out operating systems treating 8-core "Bulldozer" parts as 4-core/8-thread (i.e. each module as a core and each "core" as a logical processor);

The problem with this argument is that even the operating systems couldn't keep it straight, and it was very clear that the OS wasn't detecting the CPU this way but instead was manually configured to show 4c/8t. Even Windows 10 still detects some Bulldozer architecture chips as 4c/4c.

 

[bug]

To tech-educated audience, not to an average consumer.

Some CPUs don't have L3, but this does not make them "not CPUs". The issue is not whether it "counts" or "does not count" as a CPU core, but the fact that AMD failed to convey this information to consumers. E.g. if you write "8 cores" on the box, the consumer will buy the CPU and be disappointed later when he/she finds out that it may not be 8 cores at all. If you write 4 modules [8 cores] it won't make it significantly less appealing, but it will make the potential buyer curious about these "modules" and at least do a quick google-fu to see what this is all about.
When you have such a mess, it's always open for interpretation. Some may say "yes, it's 8 cores with shared components", some (like me) will say that this is 4 cores with extra stuffin'.
Sun did use similar approach in the past to make their CPUs more multithreading-capable, and while each core had 2 integer units and 1 floating point unit it was still being called one CPU core (capable of running 8 threads).

That's all nice and everything, but I think it's clear to everyone this is just about lawyers noticing "core" hasn't been defined in a court of law before and grabbing their opportunity to squeeze some money out of it.

 

[Vayra86]

To tech-educated audience, not to an average consumer.

Some CPUs don't have L3, but this does not make them "not CPUs". The issue is not whether it "counts" or "does not count" as a CPU core, but the fact that AMD failed to convey this information to consumers. E.g. if you write "8 cores" on the box, the consumer will buy the CPU and be disappointed later when he/she finds out that it may not be 8 cores at all. If you write 4 modules [8 cores] it won't make it significantly less appealing, but it will make the potential buyer curious about these "modules" and at least do a quick google-fu to see what this is all about.
When you have such a mess, it's always open for interpretation. Some may say "yes, it's 8 cores with shared components", some (like me) will say that this is 4 cores with extra stuffin'.
Sun did use similar approach in the past to make their CPUs more multithreading-capable, and while each core had 2 integer units and 1 floating point unit it was still being called one CPU core (capable of running 8 threads).

I still struggle to see how an 'average consumer' DOES know that he's been ripped off with half a core count and doesn't know about other shared resources in chips. I also struggle how you would even know what core count is while not knowing AMD was always up front about this architecture and its design choices.

This is the same consumer that still buys into 'dual core' Intel chips that have i7 plastered on top, you know, that name that tends to represent a quad core in the desktop world since forever

Baseless case is baseless, IMO. AMD should certainly win this and if they don't, faith in humanity -100

That's all nice and everything, but I think it's clear to everyone this is just about lawyers noticing "core" hasn't been defined in a court of law before and grabbing their opportunity to squeeze some money out of it.

Amen.

 

[londiste]

AMD was always up front about this architecture and its design choices.

It depends. AMD was quite clear about the architecture. Unfortunately, not many people know or care about that and at the same time AMD very aggressively pushed the 8-core story.

 

[Vayra86]

It depends. AMD was quite clear about the architecture. Unfortunately, not many people know or care about that and at the same time AMD very aggressively pushed the 8-core story.

Well if the information was publicly available and even *presented* publicly, isn't that just customer due diligence territory? Though I see what you're getting at, there should have been a couple of powerpoint slides with disclaimers concerning the 8 core count, along with every outing of it. 'May not provide full performance of a single core' etc.

I think the only real approach here is the direct comparison to SMT and how AMD can explain that one is a 'core' and the other is SMT. They can bring their Ryzen die shots for that.

 

[bug]

Well if the information was publicly available and even *presented* publicly, isn't that just customer due diligence territory? Though I see what you're getting at, there should have been a couple of powerpoint slides with disclaimers concerning the 8 core count, along with every outing of it. 'May not provide full performance of a single core' etc.

I think the only real approach here is the direct comparison to SMT and how AMD can explain that one is a 'core' and the other is SMT. They can bring their Ryzen die shots for that.

The info was more than public. AMD repeatedly told us they went the way they did because in their opinion, floating point was not used enough to justify the die space a FPU per core needs. It was more like "we looked at this and here's what we think a clever solution we found". Even if almost everyone realized it was mostly a gamble.

Going even further back, people bought AMD's 5x86 and later K6 (and variants), knowing fully well they didn't measure up to Intel in FP performance. But since they were offered at a better price, they were good enough for plenty of people.

 

[neatfeatguy]

If something like this can take hold in courts, then maybe people should turn their attention towards the International Astronomical Union and file a lawsuit about how they demoted Pluto.

 

[XiGMAKiD]

Here you guys go;
-snip-
IEEE has to review the above to even allow it.

I'm more intrigued by why do you label the cores counter clockwise than by the IEEE review

 

[silentbogo]

That's all nice and everything, but I think it's clear to everyone this is just about lawyers noticing "core" hasn't been defined in a court of law before and grabbing their opportunity to squeeze some money out of it.

The reason this lawsuit went through is not "because there is no definition of the CPU core" and lawyers found this loophole, but because AMD gave consumers the reason to doubt their definition. Lawyers only pick up class action lawsuits if there is substance to it, and high chance of winning or settle.

This is the same consumer that still buys into 'dual core' Intel chips that have i7 plastered on top

Those i7 chips have clearly written 2c/4t in specs. AMD chips only have "core" count on the official spec sheet.
I still agree that it's missleading, but at least you get the info on your first google hit.

 

[londiste]

AMD chips only have "core" count on the official spec sheet.

They have both 8 cores and 8 threads in their specs.
FX-8150, for example: https://www.amd.com/en/products/cpu/fx-8150

 

[bug]

The reason this lawsuit went through is not "because there is no definition of the CPU core" and lawyers found this loophole, but because AMD gave consumers the reason to doubt their definition. Lawyers only pick up class action lawsuits if there is substance to it, and high chance of winning or settle.


Those i7 chips have clearly written 2c/4t in specs. AMD chips only have "core" count on the official spec sheet.
I still agree that it's missleading, but at least you get the info on your first google hit.

So, basically this about the architecture not being explained in the spec sheet? Because the net is full of block diagrams for Bulldozer (wikipedia included) that clearly show the shared FPU.

 

[silentbogo]

So, basically this about the architecture not being explained in the spec sheet? Because the net is full of block diagrams for Bulldozer (wikipedia included) that clearly show the shared FPU.

As I said earlier, a simple "4 modules / 8 cores" or "2 modules / 4 threads" would've been enough to provide "food for thought". Link to Bulldozer architecture brief would've been nice too.
Also, Wikipedia is not amd.com. If third-party sites have more info about your product than an official product page and it's presented in a more usable manner, it's definitely a f%$^k up on AMD's part.

 

[bug]

As I said earlier, a simple "4 modules / 8 cores" or "2 modules / 4 threads" would've been enough to provide "food for thought". Link to Bulldozer architecture brief would've been nice too.
Also, Wikipedia is not amd.com. If third-party sites have more info about your product than an official product page and it's presented in a more usable manner, it's definitely a f%$^k up on AMD's part.

Are you seriously suggesting something that was all over the place at the time was actually missing on amd.com? I don't have time to dig up the official specs right now, but wow!

 

[mouacyk]

Customer faith is shaken when an OS is reporting core count differently than what was marketed, regardless of technical details. One would think that a business pushing (even more) core count as its primary strategy would do everything to make sure that its concept isn't misrepresented to the customer. Either someone dropped the ball in marketing defense, Bulldozer's CPU core doesn't stack up, or all the operating systems are out to do AMD in.

 

[silentbogo]

I don't have time to dig up the official specs right now, but wow!

You don't have time to dig up specs, now put yourself in shoes of the average Joe/Jane who wanted to build a PC between 2012 and 2015, for example.

 

[bug]

You don't have time to dig up specs, now put yourself in shoes of the average Joe/Jane who wanted to build a PC between 2012 and 2015, for example.

Wtf, I'm at work trying to debug some stuff dude. Plus, between then and now, those specs are most likely archived.

 

[nickbaldwin86]

Regardless of the merits of the case, these class action lawsuits are a sham. The lawyer fees are disproportionately high, so any payout initially goes to the firms representing the plaintiff. All these tech suits are carried out in spite of consumer demand, not because of it. I hope AMD win, not because they were right but because the consumers won't win this, the lawyers will.

If AMD wins then they have not learn what they did isn't ok. (If that fact be true). the money just goes to the lawyers and AMD doesn't get to keep money that they wrongfully took ie falsely sold a product.

it is a loose loose really but either way hopefully AMD has learned something from this

 

[Vya Domus]

Never in the history of computing CPU cores have been equated to execution units. This is a erroneous association that sparked out of people's desire to slap a badge on things they don't understand. By the way has anyone filed a lawsuit yet on every GPU maker out there for marketing their products as having thousands of "cores" ? This will never go anywhere not because this is a grey area but rather because there is nothing to debate. That's an x86 compatible CPU and that's it's architecture, end of story. If you want to have the most purebred CPU core out there then go ahead and buy one of these :

No FPUs, no pipelining, no duplicate of nothing. But good luck because neither AMD nor Intel have made CPU cores that look like that for more than 3 decades now.

since AMD failed to convey that to the customer

Details about the architecture that they used were made public and several publications made some pretty in depth analyses on it. Clearly that was never the case.

Try explaining a customer why his A4-5300 shows up as a single core.

But somehow AMD was expected to spoon feed complex computer architecture concepts to them ? Try explain to a customer why CPUs from both AMD and Intel with duplicate front ends show up as having more cores in various programs. Come on.

some (like me) will say that this is 4 cores with extra stuffin'.

Then unfortunately you don't know enough about this subject as was the case with many.

 

[Vayra86]

Customer faith is shaken when an OS is reporting core count differently than what was marketed, regardless of technical details. One would think that a business pushing (even more) core count as its primary strategy would do everything to make sure that its concept isn't misrepresented to the customer. Either someone dropped the ball in marketing defense, Bulldozer's CPU core doesn't stack up, or all the operating systems are out to do AMD in.

Or its none of those things, and the core/thread count was just a way for a less conventional architecture to 'fit in' with an OS.

You don't have time to dig up specs, now put yourself in shoes of the average Joe/Jane who wanted to build a PC between 2012YOu and 2015, for example.

What about them? FX processors were reviewed and those reviews ALL pointed out something about its architecture being different, specifically in terms of core setup. AMD released countless powerpoint slides detailing the same thing. They also, without exception, noted the performance could be stellar or abysmal depending on workload.

You knew exactly what you were getting as a customer and if you didn't its entirely your own fault. Average Joe and Jane don't build PCs. They buy a laptop, let a shop build one, or let us build one.

Stop trying to find a good argument here because there is none. This case exists because perhaps there's some money to be had. Not because some customers were left out in the cold (which is what they should be used for).

at least you get the info on your first google hit.

FX Bulldozer, let's try (see thumbnail below)

There is even a nice picture with schematic/die shot, clearly specifying two integer and one FP unit and shared cache.

And here comes the kicker. Let's do i7 Core as well? You tell me what's more informative for a first google hit...

 

[HD64G]

Since I own a CPU of this arch for months now (FX8350) and knowing well what to expect from it for years before I got it, I can definitely say (after testing many different use cases on it) that it is an 8-thread CPU that acts some times as an 8-core one and others as a 4-core one due to the OS schedulers design. Most apps that need mutliple threading power use it completely at its 100% of its threads (i.e. browsers) and others use just the first thread per module until (or if not) more power is needed. And then, it begins using the rest of the threads. This type of behaviour shows that it is a dinstictive arch from what we knew until back then, but it doesn't make it a 4-core at all. Failing to fully understand something doesn't make it non-existent after all.