Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

[Phynaz]

for Intel one core gets all and the other gets also all but only if the "preferred" stalls.

Cool, I'm sure you could point me to the documentation on how this preferred thread thingy operates, right?

So cores = cores, but the amount of cores I have in the OS and I can really use and not the marketing naming thingy.

The amount of cores you have in the OS? As in displayed in Windows task manager?

THAT'S how you decide how many physical cores a CPU has??

Would you like to come work for me? I mean I had budgeted for 400 servers this year, but if all we need to do to increase our computing resources is make more CPUs show up in task manager then we will be stars!

We can just have my engineers keep provisioning more and more virtuals, all those extra cores will show up in the OS and be usable!

We'll save millions!

 

[Phynaz]

First, BD is designed with enough resources not to bottleneck.

That's BS John and you know it. If it were true then Bulldozer cores would scale at 100%.

It's okay to admit that trade-offs were made in Bulldozer's design. All chip design is about managing trade-offs.

 

[dmens]

Use a ridiculous amount of RAM and never page anything out?

It's not just page faults, there's cache misses, branch mispredicts, thread nukes for whatever reason, etc.

Yeah, you'll always be able to squeeze a little out of HT, but considering it's an extra $100 for the 2600K I'm not sure if that little bit is worth the extra dosh.

I'm not talking about the MSRP, I'm talking about the extra cost of die area to implement SMT which translates into dollar cost to the manufacturer.

If it's so much extra possible performance for so little transistor cost why not toss it in?

My guess would be design and validation cost.

 

[HW2050Plus]

First, BD is designed with enough resources not to bottleneck.

Second, you asked for a benchmark and I showed you the benchmark, so don't accuse me of cherry picking. Why did I choose SPEC int? Because that is a processor-only benchmark. If you want to argue any other benchmarks you start down the path of the platform and the processor only becomes a part of the equation.

The term "not to bottleneck" is a too strong term for what Bulldozer was designed for. This sounds like never ever which I doubt.

It was designed not to bottleneck in the very majority of workloads that is right as far as I can see from the design. However I could create workloads where Bulldozer would suffer from bottlenecks. I could get the decode to bottleneck using plenty SSE instructions which are all using 128 Bit constants (worst case scenario for decoder). Possibly I could even get the load/store units to become a bottleneck using only integer load and store instructions in parallel with SSE load and store instructions at both cores of a module at the same time until the L/S buffer is full. Yes that are rather artificial workloads but you used a too strong term there. Okay my claims that I could write some code which brings up some bottleneck in Bulldozer must be proven with an actual part but I think this never ever claim is questionable.

Sandy Bridge e.g. has a general bottleneck in the predecoder stage. Intel uses a trace cache to compensate this to some degree in case of loops (as long as the loop is small enough to fit into this trace cache, so e.g. loop unrolling optimizations create severe problems for Sandy Bridge). Bulldozer may not have a general bottleneck like this but for all processors you could create code snippets which would make them stall due to a bottleneck.

Taking SPEC(CPU2006) is not cherry picking. SPEC is the only fair real world CPU benchmark set out there and one which is fully transparent. All professionals use SPEC as a decision base for general CPU benchmarks. And I really don't know why SPEC benchmark results are not used at Anandtech and they prefer to use very specialized benchmarks.

Therefore, after the last few posts, it is possible to claim that in some cases, a hyperthreading CPU will only be able to run 1 thread, at any precise point in time, whereas a bulldozer module can run 2.

Sure it can switch between 2 threads quickly, but we are not discussing this.

Point made.

No, that is wrong.
In ALL cases, a hyperthreading CPU will only be able to run 1 thread, at any precise point in time, whereas a bulldozer module can run 2.
Then it is correct. So at any given clock cycle only one thread is running. And they splitted it into a priority thread (the fast one) and a "waiting" thread (the slow one).

But do not get fooled, because of processor stalls this still gives performance improvement in many cases. But this is the reason why Hyperthreading gives only -5 to 30% of performance boost while this Bulldozer "module technology" gives 80-100% of performance boost.

There are even scenarios possible where the waiting thread in Intel's HT will be never executed at all.

E.g. for chess programs the performance will even degrade if you enable HyperThreading, where Bulldozer's split core approach would perfectly shine. So e.g. chess programs would be a benchmark which will favor AMD Bulldozer and totally cannot profit from HT.

Generally speaking HT shines if the workload is extremly memory intense and the priority thread has often to wait for memory results to come in. If this is the case the waiting thread will execute. The waiting thread can also use other pipeline stalls of the priority thread (e.g. because of branch misprediction) to execute.

This gives the funny effect that if Intel improves branch predition and memory subsystem then the priority thread gives more performance while the waiting threads gives less performance. So overall there will be no performance improvement. For Bulldozer if branch prediction and/or memory subsytem is improved both threads will be faster so they double the overall performance gain from such improvements.

Likly that was a reason why AMD made significant improvements in the Bulldozer design for that (prefetcher, predictor with/deepth, double memory bandwith to caches, reduced latency of memory controller, increased load/store queue).

On the other hand this double half cores create more stress to the memory subsystem (not to the predictors and L1 since those are doubled).

 

[Phynaz]

If it's so much extra possible performance for so little transistor cost why not toss it in?

Everyone except AMD (and ARM?) has. IBM, Sparc, Fuju, etc..

I remember reading in the past the issue was AMD's CPU design at the time didn't really allow it to occur for them.

CMT is basically SMT with fewer shared resources anyway when you get down to it.

 

[HW2050Plus]

Cool, I'm sure you could point me to the documentation on how this preferred thread thingy operates, right?

Don't want to search Intel documents but you could use http://en.wikipedia.org/wiki/Hyperthreading and if you want to know more than follow the document links there.

I do not say that the roles of preferred and waiting threads cannot swap, but at any given point in time there are those roles. It works like this:

Start:
while (preferredIsAbleToExecute)
{
execute_preferred();
}
switchPreferredAndWaiting;
jump Start;

Please also do not forget that there are cases where both threads are stalled and no thread is executed as long as the shorter stall persists. (That also applies to Bulldozer cores of course).

 

[beginner99]

Use a ridiculous amount of RAM and never page anything out?

Yeah, you'll always be able to squeeze a little out of HT, but considering it's an extra $100 for the 2600K I'm not sure if that little bit is worth the extra dosh.

Would be interesting to see AMD add it to their chips in some capacity. I'm guessing that it's not a patent issue as the two companies have some cross licensing deals in place as I recall. If it's so much extra possible performance for so little transistor cost why not toss it in?

The 2600k also has 2 MB more cache (and a higher clock). So the 100$ is not only for HT.

Why AMD did no toss it in? No idea but considering Bulldozer is years late I would assume they never looked at it because they chose a different approach (=core/module) which turned out to be much more complicated -> huge delay.

HT may be a marketing gimmick for most users but in certain applications benefits are huge, like in seti@home.

I'm not going to claim I know how HT works but how I understood it, the 2 threads can execute simultaneously in 1 clock cycle. The idea is that most resources in an x86 cpu are unused even if the cpu is active. If you now have a second thread, you increase the usage.

 

[Accord99]

And to understand the marketing naming differences you should know why this is:
AMD Bulldozer provides a number of equally fast cores so they name them cores.
Intel provides a number of very fast cores + a number of very slow cores that is why they only call their fast cores as cores and the slow cores as (hyper) threads.

That's not how Hyperthreading works; when two threads are run through the core both threads get roughly equal access to execution resources.

From a marketing perspective this is clever because they can show that they have just more cores. But that is okay since the way AMD does core doubling ("module technology") is way superior to Intel's way of doubling cores ("Hyper Threading").

That's debatable; for AMD to match the throughput of a 6-core Gulftown with Hyperthreading at comparable power consumption, AMD needs 12 full cores. 4-core Sandy Bridge i7 has comparable throughput to 7 full current AMD cores at the same clock speed. And if the application(s) can't completely use every AMD core, the stronger Intel cores increasingly pull away from the AMD cores.


I would say Intel already has a Bulldozer type architecture, with the only difference that Intel cores can seamlessly switch between splitting execution resources between two threads and using all of them on one thread, leading to a well-balanced design that has excellent performance in either throughput or latency type applications.

 

[Mopetar]

That's BS John and you know it. If it were true then Bulldozer cores would scale at 100%.

It's okay to admit that trade-offs were made in Bulldozer's design. All chip design is about managing trade-offs.

It's probably not perfect, but no one in the marketing department will ever admit that their product is anything but great. If it scales at 95% or better per doubling of cores I think most people would accept that as good enough.

We don't have any official performance numbers so it's impossible to assess the validity of his statement. Just take it as, Bulldozer was carefully designed by people who know what they're doing.

 

[IntelUser2000]

No, that is wrong.
In ALL cases, a hyperthreading CPU will only be able to run 1 thread, at any precise point in time, whereas a bulldozer module can run 2.
Then it is correct. So at any given clock cycle only one thread is running. And they splitted it into a priority thread (the fast one) and a "waiting" thread (the slow one).

What are you blabbering about? I think you are confusing SMT with SoEMT, or even FMT.

 

[HW2050Plus]

The amount of cores you have in the OS? As in displayed in Windows task manager?

THAT'S how you decide how many physical cores a CPU has??

Would you like to come work for me? I mean I had budgeted for 400 servers this year, but if all we need to do to increase our computing resources is make more CPUs show up in task manager then we will be stars!

We can just have my engineers keep provisioning more and more virtuals, all those extra cores will show up in the OS and be usable!

We'll save millions!

What's up with you? If performance is all you want i wonder why you came up with that:

As I said in the first page of this thread:

Hey, an eight core CPU is faster than a four core CPU! Who woulda thought??

Really, would you compare that 2600K to an Athlon X2??

As I said you should compare all. And regarding best performance I compare the best consumer CPUs of both companies. If this is AMD Phenom II 1100T to i7-980 then I do this even when the i7-980 has 12 threads compared to the 6 of AMD Phenom II 1100T. And then I say, that Intel has the performance crown. Of course there are customers which buy a AMD Phenom II 1100T because of the lower price.

You are right with that:

Hey, an eight core CPU is faster than a four core CPU! Who woulda thought??

Same as I say though I use different words. However you could say it also like this: A 4 core Bulldozer CPU is faster than a 4 core Intel CPU because the AMD Core Multithreading (CMT) is superior to the Intel Hyperthreading (HT) technology of doubling threads.

It was this addition what brought up the discussion:

Really, would you compare that 2600K to an Athlon X2??

If you compare best performing of current Intel and AMD processors you would not select this Athlon X2. That was the wrong thing you brought in and which created some unnecessary discussion.

You see? You ask now for ultimate performance but brought in this low end Athlon X2.

So either you say I want a fair comparison. Fair would be from consumer perspective equal price and from technical engineering perspective equal die area. Or you opt for ultimate performance.

And regarding this:

The amount of cores you have in the OS? As in displayed in Windows task manager?

THAT'S how you decide how many physical cores a CPU has??

If you opt out for this phsical stuff. AMD Bulldozer Zambezi has 4 physical cores and 8 threads. Same as Intel 2600 and less as i7-980 which has 6 physical cores and 12 threads.

Yes the way of creating this additional threads for AMD with it's core multithreading (CMT) is way more effective than Intel's Hyperthreading (HT). It is so effective that AMD marketing advertizes this as "new" core count. As I said, don't get confused by marketing terms.

That is why I can use the amount of cores provided to the OS as a comparison indicator. You can use die size and prices as well which gives same results. Or you count physical cores (4 in both cases).

 

[HW2050Plus]

What are you blabbering about? I think you are confusing SMT with SoEMT, or even FMT.

I talk about Intel Hyperthreading and you?

 

[Phynaz]

You know what, forget it, I'm done with you too.

 

[jvroig]

However you could say it also like this: A 4 core Bulldozer CPU is faster than a 4 core Intel CPU because the AMD Core Multithreading (CMT) is superior to the Intel Hyperthreading (HT) technology of doubling threads.

Your last few posts are starting to be completely lost, and this one is completely bizarre. No wonder you previously said you are not bothered when they "leaked" a "quad-core" Zambezi fighting against a 4c/8t Sandy Bridge and winning hands down.

A 4-core Bulldozer can only do 4 threads. A 4 core Intel CPU can do 8 threads. Please do not invent your own terms. AMD and Intel has clearly stated what terms they use and what it is for.

 

[Phynaz]

AMD Bulldozer Zambezi has 4 physical cores and 8 threads.

It will be 8 cores, 8 threads.

Hmmmm.....Who to believe? The guy that works for AMD is my choice.



And BTW, it's Cluster Based MultiThreading, not Core MultiThreading.

 

[Phynaz]

It's probably not perfect, but no one in the marketing department will ever admit that their product is anything but great. If it scales at 95% or better per doubling of cores I think most people would accept that as good enough.

We don't have any official performance numbers so it's impossible to assess the validity of his statement. Just take it as, Bulldozer was carefully designed by people who know what they're doing.

Yeah I know, I'm just really picky. And I hate the marketing bullcrap. The vast majority of it is insulting.

 

[HW2050Plus]

That's not how Hyperthreading works; when two threads are run through the core both threads get roughly equal access to execution resources.

In most cases yes. What I described points out the differences. If you run less threads than they run faster on Intel Hyperthreading whereas I named this fast core. And if you add an additional thread this adds only little to performance whereas I named that slow core. With AMD CoreMultiThreading all threads run at equal (high) performance at all time. Therefore:

Intel HT (one thread): very high speed
AMD CMT (one thread): high speed
Intel HT (two thread): only one can run at that very high speed and the other on a slow speed (averaged out over a longer period two run on a mediocre speed if you like that more).
AMD CMT (two threads): both at high speed.

However again, if you look at a shorter time frame you see for Hyperthreading that there is one thread at full speed and one at low speed. And that is always the case! Only because of roles are changed between those two threads (which is not necessarily part of any Hyperthreading implementation) at longer times this as you said get roughly equal access to execution resources. This is not immanent but an additional design choice or function.

I would say Intel already has a Bulldozer type architecture, with the only difference that Intel cores can seamlessly switch between splitting execution resources between two threads and using all of them on one thread, leading to a well-balanced design that has excellent performance in either throughput or latency type applications.

Compared to current AMD CPUs yes because AMD does not have any threading technology. So it is clear that HT gives Intel an advantage that AMD does not have.

But this changes with Bulldozer where this CMT will be superior regarding throughput and the high frequency design was made to get superior as well in latency (plus some other improvements). Both together are the reason why AMD Bulldozer could get more than 100% faster than what they have now without taking additional core improvements into account which will come on top of that but are minor in effect (10%?).

So with all this technology AMD Bulldozer will be superior to e.g. Intel Sandy Bridge at throughput and latency at the same time.

And I disagree that Intel has already a Bulldozer type architecture, they have of course not. They have already a Threading technology but that is completly different. I mean AMD could even implement Intel's Hyperthreading Technology into Bulldozer resulting then in a 4C/16T CPU and squeeze out some more performance as Intel did it with their HT. And Intel could add the AMD CMT into their CPUs getting a large boost in performance.

As of now we will however see CPUs with either HT or CMT.
Whereas HT will give -5 to +30% performance (12,5% average?) and
CMT will give ~80% average additional performance.
That is why AMD Bulldozer will provide better performance until Intel either implements CMT or CMT + HT or something new we don't think of yet.

Intel long profited (for about 6 years?) from having a threading technolgy (HT) while AMD had none. Now AMD comes out with one that is significantly superior and we will have to see how long this advantage over Intel will last.

 

[HW2050Plus]

Hmmmm.....Who to believe? The guy that works for AMD is my choice.

As it is dependend on what you define this has nothing to do with believing. In the same way Intel could define that they have 8 cores and if you look in your Intel computer you really see that it provides 8 cores and could not say anything different. But Intel did not that is why I do not (neither for Intel nor AMD). So if you want to follow the new AMD terminology than feel free to do so. But then not to confuse you should use that on Intel CPUs as well.

I think you understand my point (you can use any terminology but you should use it consistently).

And BTW, it's Cluster Based MultiThreading, not Core MultiThreading.

Do you have a source for that, because I have a source for "Core MultiThreading":

http://ht4u.net/reviews/2010/amd_bulldozer_preview/index2.php

 

[Phynaz]

I mean AMD could even implement Intel's Hyperthreading Technology into Bulldozer resulting then in a 4C/16T CPU and squeeze out some more performance as Intel did it with their HT

Congrats. You've just proven you have no idea what you are talking about.

 

[PreferLinux]

No, that is wrong.
In ALL cases, a hyperthreading CPU will only be able to run 1 thread, at any precise point in time, whereas a bulldozer module can run 2.
Then it is correct. So at any given clock cycle only one thread is running. And they splitted it into a priority thread (the fast one) and a "waiting" thread (the slow one).

But do not get fooled, because of processor stalls this still gives performance improvement in many cases. But this is the reason why Hyperthreading gives only -5 to 30% of performance boost while this Bulldozer "module technology" gives 80-100% of performance boost.

There are even scenarios possible where the waiting thread in Intel's HT will be never executed at all.

E.g. for chess programs the performance will even degrade if you enable HyperThreading, where Bulldozer's split core approach would perfectly shine. So e.g. chess programs would be a benchmark which will favor AMD Bulldozer and totally cannot profit from HT.

Generally speaking HT shines if the workload is extremly memory intense and the priority thread has often to wait for memory results to come in. If this is the case the waiting thread will execute. The waiting thread can also use other pipeline stalls of the priority thread (e.g. because of branch misprediction) to execute.

This gives the funny effect that if Intel improves branch predition and memory subsystem then the priority thread gives more performance while the waiting threads gives less performance. So overall there will be no performance improvement. For Bulldozer if branch prediction and/or memory subsytem is improved both threads will be faster so they double the overall performance gain from such improvements.

Likly that was a reason why AMD made significant improvements in the Bulldozer design for that (prefetcher, predictor with/deepth, double memory bandwith to caches, reduced latency of memory controller, increased load/store queue).

On the other hand this double half cores create more stress to the memory subsystem (not to the predictors and L1 since those are doubled).

ftp://download.intel.com/technology/itj/2002/volume06issue01/vol6iss1_hyper_threading_technology.pdf (Page 7)

A second goal was to ensure that when one logical
processor is stalled the other logical processor could
continue to make forward progress. A logical processor
may be temporarily stalled for a variety of reasons,
including servicing cache misses, handling branch
mispredictions, or waiting for the results of previous
instructions. Independent forward progress was ensured
by managing buffering queues such that no logical
processor can use all the entries when two active
software threads2 were executing. This is accomplished
by either partitioning or limiting the number of active
entries each thread can have.

2 Active software threads include the operating system
idle loop because it runs a sequence of code that
continuously checks the work queue(s). The operating
system idle loop can consume considerable execution
resources.

In other words, you are exactly wrong.

 

[Accord99]

However again, if you look at a shorter time frame you see for Hyperthreading that there is one thread at full speed and one at low speed. And that is always the case! Only because of roles are changed between those two threads (which is not necessarily part of any Hyperthreading implementation) at longer times this as you said get roughly equal access to execution resources. This is not immanent but an additional design choice or function.

It sounds like you've never used a Hyperthreading processor, regardless of time frame, each thread gets roughly half and there is no fast thread or slow thread.

Compared to current AMD CPUs yes because AMD does not have any threading technology. So it is clear that HT gives Intel an advantage that AMD does not have.

That doesn't matter, a more powerful core with Hyperthreading allows Intel to compete and beat 1.5 FULL AMD cores at the same clock speed, or two FULL AMD cores at the same power consumption right now.

So with all this technology AMD Bulldozer will be superior to e.g. Intel Sandy Bridge at throughput and latency at the same time.

That's debatable, to match a 3.4GHz Sandy Bridge's single-threaded performance, the K10 core would need to be clocked at somewhere near 5 GHz. That's a big jump to make.

As of now we will however see CPUs with either HT or CMT.
Whereas HT will give -5 to +30% performance (12,5% average?) and
CMT will give ~80% average additional performance.

The thing that some keep missing is that HT is -5 to 30%, averaging around 20% of "X", while CMT is 80% of "Y". What if 1.2X > 1.8Y? 1.2X Westmere already beats 2.0Y of Magny Cours.

Intel long profited (for about 6 years?) from having a threading technolgy (HT) while AMD had none. Now AMD comes out with one that is significantly superior and we will have to see how long this advantage over Intel will last.

Intel already is capable of taking on more FULL AMD cores now. I'd expect a 4 core Sandy Bridge with HT to beat 8 core Bulldozer in desktop applications and a 6 core Gulftown to beat Bulldozer in multi-threading applications, and probably everything else except AVX.

 

[HW2050Plus]

Your last few posts are starting to be completely lost, and this one is completely bizarre. No wonder you previously said you are not bothered when they "leaked" a "quad-core" Zambezi fighting against a 4c/4t Sandy Bridge and winning hands down.

A 4-core Bulldozer can only do 4 threads. A 4 core Intel CPU can do 8 threads. Please do not invent your own terms. AMD and Intel has clearly stated what terms they use and what it is for.

I use my own terms and I did not invent them.

It is just AMD which brings up that confusion.
Before it was clear what a socket, a core and a thread means.

Now AMD renamed core to module and thread to core. They did this to put emphasis on the different technology to provide multiple threads within a core. And yes that is great marketing. With a wording change they can sell double core count. You can find my post in reply to JFAMD saying that this is very aggressive marketing.

And by no means that requires me to follow their terms especially when I want to explain the forum member Phynaz what renaming of genuine terms the AMD marketing fabricated.

Besides why you write "4c/4t Sandy Bridge"? I assume you mean "4C/8T Sandy Bridge" and that was a typo.

And yes that is why this 4 Core thing did not bother me.

Hate AMD for that renaming issue, but you may
either use AMD terms, or Intel terms or AMD terms for AMD only and Intel terms for Intel only. But and that was what Phynaz did: You may not use AMD terms for AMD only and Intel terms for Intel and then compare terms between them. That will lead to wrong conclusions.

My posts where just to explain that issue. Issue itself was "invented" by AMD.

 

[Jovec]

Wow, really getting adrift here.

This is really a two variable equation - how many core resources are shared versus duplicated, and what each architecture's IPC is.

With no shared resources, we get a full core like all current AMD CPUs and non-HT Intel ones. With mostly shared resources, we get the current virtual cores provided by HT. Bulldozer is reducing the number of shared resources in an attempt to strike a better compromise between a full core and a virtual core.

Until we can see how little or how much resource contention there is on a BD module for the apps that matter to us and what BD's IPC is, we can't compare it's performance.

BD's IPC, and not it's module design, is actually the bigger key here in my mind. I believe HT seems better than it otherwise would because of the rather large IPC advantage Intel currently has. We are dangerously close (if not there already with the SB arch) to where a 2c/4t Intel CPU beats a price comparable 4c/4t AMD CPU in multi-threaded apps.

At a basic level, a BD module is nothing more than a continuation of AMD's "more cores for less" strategy, and if it proves too effective, Intel can always counter with more than 4 real cores in the mainstream market. Without enough IPC improvement, AMD will need to counter with more modules and more duplicated hardware per module.

We can reason that it is going to be unlikely for AMD to match SB's IPC. Assuming they don't fall too far behind, it's likely that the status quo will remain. Intel will retain the single-thread advantage, and AMD will offer better multi-threaded performance for most situations.

edit: vv clockspeed speed too as corrected below, but that can be lumped together with IPC in the sense of IPC & clocks = how fast things can be done while cores/modules/ht = how many things can be done at once.

 

[itsmydamnation]

im not sure if im correct here, but isn't a bulldozer modules front end bigger then a SB core because the FP unit has its own scheduler where as SB doesn't?

remember IPC X clock speed. both IPC and clock speed by themselves are meaningless.

 

[jvroig]

Besides why you write "4c/4t Sandy Bridge"? I assume you mean "4C/8T Sandy Bridge" and that was a typo.

Yes, thanks for catching that typo. I have corrected it in the post.

Now AMD renamed core to module and thread to core. They did this to put emphasis on the different technology to provide multiple threads within a core. And yes that is great marketing. With a wording change they can sell double core count. You can find my post in reply to JFAMD saying that this is very aggressive marketing.

I understand where you are coming from. But that is not up to us. It is up to them. And considering the difference between their design and hyperthreading, it is not really such a surprise.

When you persist in using your own terms, instead of the terms accepted (as defined by those responsible - in this case AMD with regards to their core counts, and Intel with regards to theirs), it muddles the conversation. Maybe they are right to define their core as an int core (they probably are), or maybe they are wrong. Whatever the case, stick with the official terminology, so that everything is in the same context.

Hate AMD for that renaming issue, but you may either use AMD terms, or Intel terms or AMD terms for AMD only and Intel terms for Intel only. But and that was what Phynaz did: You may not use AMD terms for AMD only and Intel terms for Intel and then compare terms between them. That will lead to wrong conclusions.

I really can't see why. Their quad core handles 4 threads. Intel's quad-cores can handle eight (but nowhere near great scaling at the max # of threads)

What conclusions would be wrong? AMD chose to package their product that way, and Intel in another. You can compare which quad is better than another if that's the comparison you want to make. Or if prices were available, which products were priced similarly, even if it were an octo-core AMD vs quad Intel.

All is fair. If a quad core Zambezi (2 modules) comes out, and it is the same price as any quad Sandy Bridge product with HTT, then so be it. It will be quad vs quad, but with the Intel one having a hyperthreading advantage.

There is no sense trying to equalize "threads" count or whatever.

My posts where just to explain that issue. Issue itself was "invented" by AMD.

The problem is you just don't want to accept their definition, and you insist on "explaining" it to us, when in fact it doesn't matter and the issue is just you. Accept the definition and move on.

AMD duplicated enough resources such that it is far more than Intel's hyperthreading. A claim of 80% is far better than Intel's 5-30%. Do they market it as a real core, or as some super turbo hyperthreading? They decided on the real core approach. They made the design and the execution, that's their call. Get over it. There is no need for a crusade against this non-issue.

Otherwise, it's hard to continue a healthy debate when you insist on your own terms and wonder why people don't see your point.