My processor (6700K) and the processor before it (920) can execute 8 threads simultaneously as well. That's what SMT means, after all.

Thuban has 6 dedicated L2 caches because it is a legitimate 6-core processor.

The TechARP x264 benchmark comparing FX-8150 and 1055T would have ran 8 threads on the former and six threads on the latter. FX-8150 did not pull ahead. FX-8150 would likely do better in single threaded simply because of the 800 MHz clockspeed advantage.

Bulldozer was not ahead of its time considering the older Thuban architecture can best it in some scenarios and Intel bests it in most scenarios.

Bulldozer's design does look more like SPARC (huge ALU performance, little FPU performance) than a desktop CPU should. Even so, Bulldozer isn't exactly competitive with comparative Xeons.

FordGT90ConceptMy processor (6700K) and the processor before it (920) can execute 8 threads simultaneously as well. That's what SMT means, after all.

Thuban has 6 dedicated L2 caches because it is a legitimate 6-core processor.

The TechARP x264 benchmark comparing FX-8150 and 1055T would have ran 8 threads on the former and six threads on the latter. FX-8150 did not pull ahead. FX-8150 would likely do better in single threaded simply because of the 800 MHz clockspeed advantage.

Bulldozer was not ahead of its time considering the older Thuban architecture can best it in some scenarios and Intel bests it in most scenarios.

Bulldozer's design does look more like SPARC (huge ALU performance, little FPU performance) than a desktop CPU should. Even so, Bulldozer isn't exactly competitive with comparative Xeons.

You try hard and will lose forever.True =/= Performance, You're using Intel's Definition that's why you say 4 core.
From anandtech :

By anandtechArchitecturally Bulldozer is a significant departure from anything we've ever seen before. We'll go into greater detail later on in this piece, but the building block in AMD's latest architecture is the Bulldozer module. Each module features two integer cores and a shared floating point core. FP hardware is larger and used less frequently in desktop (and server workloads), so AMD decided to share it between every two cores rather than offer a 1:1 ratio between int/fp cores on Bulldozer. AMD advertises Bulldozer based FX parts based on the number of integer cores. Thus a two module Bulldozer CPU, has four integer cores (and 2 FP cores) and is thus sold as a quad-core CPU. A four module Bulldozer part with eight integer cores is called an eight-core CPU. There are obvious implications from a performance standpoint, but we'll get to those shortly.

FordGT90ConceptMy processor (6700K) and the processor before it (920) can execute 8 threads simultaneously as well. That's what SMT means, after all.

Thuban has 6 dedicated L2 caches because it is a legitimate 6-core processor.

The TechARP x264 benchmark comparing FX-8150 and 1055T would have ran 8 threads on the former and six threads on the latter. FX-8150 did not pull ahead. FX-8150 would likely do better in single threaded simply because of the 800 MHz clockspeed advantage.

Bulldozer was not ahead of its time considering the older Thuban architecture can best it in some scenarios and Intel bests it in most scenarios.

Bulldozer's design does look more like SPARC (huge ALU performance, little FPU performance) than a desktop CPU should. Even so, Bulldozer isn't exactly competitive with comparative Xeons.

HT is not SMT, your CPU multithreads only on 4 cores, the rest is what Windows sees. That's why it's called Hyperthreading.
Though I guess thats another debate in itself.....;)

Ahead of it's time as in AMD gambled code was going be optimized for multicores but it didn't happen. Ironically Vishera has improves with age..... never thougfht I'd say that about a CPU...

Not looking forward to Fallout 4 chugging along at 45fps though lol





Imo it boils down to 2 points,

a) what is a "core" exactly?

and

b) 8 threads/ 8 cores? isn't that the same thing?.......Does it matter?

the CPU is 8 cores, just some resources are shared between 2 cores

I don't see what the fuss is all about, it's not like Fiji wasn't an "overclocker's dream".

Jeez people, lets use some fucking block diagrams here and explain exactly what arguments exist for there being and not being real "cores".

1. One instruction/data fetch unit per two Integer cores and one floating point core (with two 128-bit FMAC units.)
   1. I personally don't buy this one, mainly because how much can be fetched per cycle can vary depending on the CPU. It's possible that it's a limiting factor but I doubt it.
2. Decoder, the initial Bulldozer had only one uOp decoder per module.
   1. There could be some argument here as AMD went from the Phenom II being able to decode 3 per cycle as opposed to Bulldozer which could only do 4 per module. AMD revised this in Steamroller (as they should have, damn it!) and a reasonable performance gain came out of it. Obviously nothing to truly counteract the size of the pipeline.
3. Dedicated schedulers and and L1. Nothing "shared" about that.
4. Floating point unit... We've discussed this and I still think that it's laughable that this and only this can be a measure of how many "cores" a CPU has.
5. L2 cache. Lets remember that the Core 2 Duo had a shared L2. Not going to go further into that one.

In all seriousness, consider Xen coming up. A lot of hardware with the SMT integer core (which is bigger than your run of the mill integer core in a bulldozer module,) is still shared with the FPU. The flaw with Bulldozer through excavator is the length of the pipeline, the IPC is a dead simple indicator for this. It's the very reason why Intel moved to a 14-stage architecture from a >30 stage one with Netburst, just as AMD's current lineup is now.

I want to say this again, Bulldozer doesn't suck because of shared resources, it sucks because the pipeline is too damn long.

The problem is when a hazard is encountered, it is much harder for a longer pipeline to recover from the stall it generates and as a result, IPC suffers and higher clock speeds are required to overcome it, (just like the Pentium 4.)

Stop whining about the FPU and focus on the damn pipeline.

behrouzYou try hard and will lose forever.True =/= Performance, You're using Intel's Definition that's why you say 4 core.
From anandtech :.......AMD advertises Bulldozer based FX parts based on the number of integer cores

I don't agree with the whole lawsuit thing - as should be readily apparent from the tone of my previous posts, but your argument in this instance is a reach. While AMD added a ton of footnotes to the PPS slide deck, the retail selling mentions nothing of shared resources. I just dug out(and ripped apart in the name of barely interested fact finding) an old BD retail box ( I don't use them but I've built a few systems with them), and nowhere is any mention of integer cores or FPU contingency



You might also have a hard time picking up the same information from any ODM/OEM's advertising

AquinusI want to say this again, Bulldozer doesn't suck because of shared resources, it sucks because the pipeline is too damn long.

That was brought up at the time. The branch misprediction penalty effectively scuppered the architecture.

behrouzFrom anandtech :

AMD says...so they can rope a dope...which is why AMD is getting sued.

Pill MonsterHT is not SMT

Yes, it is. Two threads on one core at the same time is the very definition of symmetrical multithreading.

Pill MonsterAhead of it's time as in AMD gambled code was going be optimized for multicores but it didn't happen. Ironically Vishera has improves with age..... never thougfht I'd say that about a CPU...

AMD knew it wouldn't be but went ahead with it anyway.

Pill Monstera) what is a "core" exactly?

Your typical x86-64 core has a instruction decoder, some type of scheduler/dispatcher, ALUs, FPUs, L1 cache for instructions and data, and an L2 cache. Under that may or may not lie an L3 cache, memory controller, and some kind of interface to communicate with the rest of the system. In short, the core can function in its entirety on its own (excluding the "under that" portion anyway).

Pill Monster8 threads/ 8 cores? isn't that the same thing?.......Does it matter?

No, 8 symmetrical threads indicates 8 logical processors. If this was a SPARC T5, that would mean there's only one underlying core processing it. In the case of Core i7-6700K, that means quad-core with Hyper-Threading enabled. Both scenarios get you 8 logical processors but with widely different technology under it. Bulldozer has 8 logical processors but, using the definition above, it only contains four cores with symmetrical multithreading. The L1 instruction cache, instruction decoder, dispatcher, FPU and L2 cache in each core handle all threads that pass through the core.

It matters because when AMD markets something as having "8 cores" when it largely has the guts of a 4 core, performance underwhelms. Intel i7-5960X is an example of 8-core processor. Compare the relative performance of that compared to FX-8###. They are leagues apart--especially in multithreading. It's like comparing a 4790K (4 cores) to 5960X (8 cores) in heavily multithreaded benchmarks: 5960X runs away. There's a huge difference between an actual 8-core processor compared to a 4 core with SMT.

Aquinus...

The suit is about misleading consumers. Those of us in the know see through AMD's BS (divide core count by two) but those not in the know have to learn the hard way.

By the way, this slide better shows how much is shared:

Pretty much everything except the actual number crunching.

HumanSmoke

I actually have an FX-6300 box right next to me. You're missing the sticker (can see the residue) which is the only place that says what is in the box.

FX 6300
AMD FX 6-core.

OEMs are repeating AMD's lie.



Zen omits the nonsensical second "integer core" as well as fattening up the ALUs and FPUs:

FordGT90ConceptBy the way, this slide better shows how much is shared....

This one might offer a better indication of workflow

FordGT90ConceptI actually have an FX-6300 box right next to me. You're missing the sticker (can see the residue) which is the only place that says what is in the box.

The sticker on mine was still attached to the top of the box - just mentions the SKU, cache, frequency, serial and part numbers, socket, and QR code. Nothing earth shattering in the way of architectural revelations.

R-T-BYou however don't have to look any farther than the modern era of arm cores and even some server archictecutres (SPARC comes to mind) to find modern "octacore" CPUs with only one FPU.

This is nonsense. It's a core if it can math at all. Heck earlier CPUs lacked a MULTIPLICATION FUNCTION. They were still considered a core.

Pretty much this - I can run 8 processes doing integer based tasks (even some FP too) completely independently of each other on an FX-8xxx. There's definitely more independence. I sometimes play a game while streaming it while also encoding movies in the background ;)

FordGT90Concept, you have a huge chip on your shoulder.

HumanSmokeThis one might offer a better indication of workflow

I posted that one several pages back. Yes, it's probably the best.

Related note: block diagrams for Intel processors seem scarce. :(

HumanSmokeThe sticker on mine was still attached to the top of the box - just mentions the SKU, cache, frequency, serial and part numbers, socket, and QR code. Nothing earth shattering in the way of architectural revelations.

Top has...
-QR code, fancy AMD logo that changes under light, and AMD logo
-Model number
-Model description
-Clockspeed, # MB Total Cache
-Serial # Barcode
-Serial # Text

---tear to open the box----

Back has...
-Black Edition
-Socket AM3+, Includes Heat Sink Fan
-Part #
-UPC barcode
-UPC printed

FordGT90ConceptAMD says...so they can rope a dope...which is why AMD is getting sued.


Yes, it is. Two threads on one core at the same time is the very definition of symmetrical multithreading.

No, it isn't. And nowhere on the page you linked to does it mention SMT. (Similtaneous btw, not Symmetrical.)

Intel simply state mutiple threads can run on one core, and not similtaneously.









I'm with Aquinous in so far as this debate is getting a little repetitious.


Much ado about nothing imo...

Pill MonsterNo, it isn't. And nowhere on the page you linked to does it mention SMT. (Similtaneous btw, not Symmetrical.)

Intel simply state mutiple threads can run on one core, and not similtaneously.

During the same clock cycle, which might as well be simultaneously in practical terms.

@FordGT90Concept, I'm going to make this clear. Your argument is stupid, because to make it you have to find a person who is both a genius and an idiot.

If you are to argue that performance figures are what the plaintiff is using (which as per my earlier links, they are), then you've got to argue against some standard. Intel is not a direct competitor, and thus isn't a standard. If you're arguing Thuban as a comparison, then you've got to explain monetary discrepancy and an architecture change. Neither of these things is grounds for a lie, or Netburst should have had two lawsuits filed against it. As AMD published information well in advance of the release of Bulldozer, there is no reasonable assertion that they lied about the core count. Heck, I could make my own CPU, wherein each processor is single bit and have a 20 core processor. To argue that AMD lied about core count, when they previously clearly defined what a core was, is to acquiesce to being a moron. I don't think the lawyers are that stupid, because the case would be immediately dismissed by the judge.

If you argue that AMD lied, then prove it. They didn't release factually wrong benchmarks, they just cherry picked the best results. That's been considered fair game for decades.

If your argument is that the removed components are necessary, you need to be an idiot savant. You have to completely understand processor architecture, have future knowledge about how coding will use what you are developing, and you have to be so moronic as to not read the technical information put out by the company releasing the product (per the 2009 Anandtech article). Find me that idiot savant, and I'll find you the person who can single hand design the successor to Zen.

FordGT90Concept...
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?"...

This type of idiot does not need to be protected by the legal system. The clerk behind the counter is culpable for recommending that they buy a processor. The consumer is responsible for not educating themselves on the purchase. AMD has made the information they require to make an informed decision publicly available for literally years, yet they decided not to inform themselves. Our legal system does not exist to help those with retarded mental processes; it exists to mete out reparations for those who have done things which the law forbids, to mete punishment for those which haven't done what the law requires, and most importantly determine when one is guilty of either of these things.


What you're arguing is that you feel bad. I agree, I feel that the marketing was atrocious and misleading. At the same time all of the relevant data was widely available, and AMD published their data well in advance of the Bulldozer launch. Your argument for culpability on AMD's part is an argument made via emotion. Your staunch defense of said points, despite ample proof that AMD never lied, exemplifies this denial. Saying that you know it'll be thrown out, despite wanting it to happen, is asking for a massive waste of resources to no real end.



Let me be fair though. Looking at @HumanSmoke's pictures, I can't find a single mention of core count. I'm now looking at the box for an Intel processor (4790k). That box proudly states "4 Cores / 8-Way Multitask Processing." You've spent the better part of a page arguing out the core count crap, but haven't even tried to justify your point. The core count isn't listed on the processor box of AMD. The core count on Intel's box is only 4 (despite HT). Neither of which define what a core is. Neither of which promise a numeric performance level. Most problematically, the Intel processor lists 4 cores despite having 8 logical cores with HT. Neither of the companies have demonstrably lied on their packaging. The only chance this suit has is if the judge decides to rule on the advertising material...Oh wait, they can't do that. The FTC rules on fairness in advertising.

Sorry, but your entire argument is based upon the false premise that this is a fact and logic based argument. It isn't. This is some idiot trying to cash out because they think that everyone complaining about its performance on the internets just haven't decided to cash in yet.

That particular bit of anger comes courtesy of my distaste for the law firm handling this. Seriously, if you do any research into them at all you'd see that they are the next incarnation of copyright trolls. All of their cases are arguing about high end technologies, where no precedence is set, and their track records is...spotty. Basically, like any slimy lawyer they are willing to sue anybody and represent anyone willing to cough up cash. These are the kind of leeches who give lawyers a bad name, whenever public defenders (also a type of lawyer) do so much good that it isn't funny. This is why our legal system is a joke, and it takes years just to get something to happen if you're wronged.

Should AMD be held accountable for misleading advertising; maybe. Should this be in court; absolutely not. Should this have been filed before this year; if it was actually in the public interest it should have been filed in 2011. The argument that there exists any grounds for this, given the information presented by the plaintiff, is a joke. You can argue technicalities all night, but the plaintiff must prove damages and lies (that's the point of innocent until proven guilty). Every argument you make has a simple counter. I think you're in the wrong here, because your heart is leading your head.


Edit:
Added quote and framed it.

lilhasselhofferLet me be fair though. Looking at @HumanSmoke's pictures, I can't find a single mention of core count. I'm now looking at the box for an Intel processor (4790k). That box proudly states "4 Cores / 8-Way Multitask Processing."

Just to clarify, the packaging for all the processors is generic (as Ford mentioned), any relevant info is on the seal/sticker. The FX-8120 I have here states 8-Core (albeit the actual sticker is a bit munched up), and the info layout is the same as this.

All seems a little storm in a teacup to my way of thinking. If AMD had provided a white paper with every processor and emblazoned every package with a screed of info about shared resources and a sea of asterisks, I doubt it would have altered the buying habits of most people, any more than the GTX 970 kerfuffle seems to have deterred its uptake.

HumanSmokeJust to clarify, the packaging for all the processors is generic (as Ford mentioned), any relevant info is on the seal/sticker. The FX-8120 I have here states 8-Core (albeit the actual sticker is a bit munched up), and the info layout is the same as this.

All seems a little storm in a teacup to my way of thinking. If AMD had provided a white paper with every processor and emblazoned every package with a screed of info about shared resources and a sea of asterisks, I doubt it would have altered the buying habits of most people, any more than the GTX 970 kerfuffle seems to have deterred its uptake.

Fair, but what I said isn't incorrect. An SKU can have whatever information you'd like on it, and it isn't advertising. It's a stock keeping unit.

Look at some of the barcodes out there, and tell me they're lying. There's a wealth of them which shorten "assorted" to "ass," and the fun ensues.


Intel and AMD don't advertise their barcodes. They label product with barcodes to identify it. This suit is about advertising, not barcodes.


Edit:
For reference, the 4790 SKU and barcode have no mention of core count. They only refer to frequency (max turbo of 4.0 GHz), cache (8MB), and socket.

AquinusJeez people, lets use some fucking block diagrams here and explain exactly what arguments exist for there being and not being real "cores".


I want to say this again, Bulldozer doesn't suck because of shared resources, it sucks because the pipeline is too damn long.

The problem is when a hazard is encountered, it is much harder for a longer pipeline to recover from the stall it generates and as a result, IPC suffers and higher clock speeds are required to overcome it, (just like the Pentium 4.)

Stop whining about the FPU and focus on the damn pipeline.

Well actually it's both, shared resouces and pipeline.. depends on the workload.

I'll leave it there .....had enough debates for today. ;)






OT anyone know why spell ckeck wouldn't be working in Fiirefox?

lilhasselhofferFair, but what I said isn't incorrect. An SKU can have whatever information you'd like on it, and it isn't advertising. It's a stock keeping unit.
Look at some of the barcodes out there, and tell me they're lying. There's a wealth of them which shorten "assorted" to "ass," and the fun ensues.
Intel and AMD don't advertise their barcodes. They label product with barcodes to identify it. This suit is about advertising, not barcodes.
Edit:
For reference, the 4790 SKU and barcode have no mention of core count. They only refer to frequency (max turbo of 4.0 GHz), cache (8MB), and socket.

As I mentioned, I just added the information for clarification - not to contend any point.

I bet the guy wins,and rightfully so.
I don't like that it's kicking AMD when they're down but..they brought it on themselves.
Victim of Corporate shenanigans.

I'll just let the die shot do the talking. The FX-8000 series has 4 modules, each module contains 2 cores with shared resource. You can see cores in the die shot so "physically" it's there. On the other hand, you can't see hyper threading on the die shot so it doesn't count as a core. You may argue on the definition of what a core is. Me on the other I'd stick on what can be seen.

The main problem lies in what is the definition of a core?? I'll be surprised if AMD looses this case... AMD is a quite fragile position atm and what I am worried about is the bad publicity that this generates...

www.cpu-monkey.com/en/cpu-intel_core_i7_2600k-6
www.cpu-monkey.com/en/cpu-amd_fx_8350-7
www.cpu-monkey.com/en/cpu-intel_core_i7_5820k-440

Cinebench R11.5, 64bit (Multi-Core)
Intel core i7 2600k = 6.83
AMD FX-8350 = 6.94
Intel core i7 5820 = 11.05

-------------------------------------
Cinebench R11.5, 64bit (Single-Core)
Intel core i7 2600k = 1.66
AMD FX-8350 = 1.11
Intel core i7 5820 = 1.73
-------------------------------------
Multi thread/Single thread
Intel core i7 2600k = 4.1145 (6.83/1.66)
AMD FX-8350 = 6.2522 (6.94/1.11)
Intel core i7 5820 = 6.387 (11.05/1.73)

Multi thread doesn't mean scalar liner , but my calc shows that AMD FX-8350 acts as 8 Core with very poor IPC.If AMD's IPC was 1.66 , Number of Cinebench R11.5, 64bit (Multi-Core) would be 10.378 , almost 52% faster than Core i7 2600K.

Edit:
AMD lawsuit over false Bulldozer chip marketing is bogus

FordGT90Conceptblock diagrams for Intel processors seem scarce. :(

Behold

from the xbitlabs article www.xbitlabs.com/articles/cpu/display/sandy-bridge-microarchitecture_3.html
Also important for intel architectures since nehalem is ring interconnect bus for l3 cache

from the same article www.xbitlabs.com/articles/cpu/display/sandy-bridge-microarchitecture_4.html

Pill MonsterWell actually it's both, shared resouces and pipeline.. depends on the workload.

I'll leave it there .....had enough debates for today. ;)






OT anyone know why spell ckeck wouldn't be working in Fiirefox?

Yeah but that's not because the L2 is shared, that's because AMD sucks at making fast SRAM cache stores. The Core 2 chips had a shared L2 between two full cores and they didn't suck. :p

Well either way at the end of the day...AMD will still have the first consumer 8 core once Zen comes out. Unless Intel decides to jump on it too. They did finally bring in a 6 core. Can't expect Intel to sit quietly as AMD unleashes an 8 core 16 thread monster on them.

Actually, I hope Intel does because right now I suspect AMD wouldn't price it competitively enough unless Intel had something for them to undercut.