Bought an 8 core got a quad

[dontsh00tmesanta]

Good Lord. AMD FX was released in 2011. How in the world are people still beating this dead horse? It isn't even a horse anymore, it's a pile of ash that used to be some bones that used to be a horse. The flesh rotted away years ago and people kept beating it, hence the pile of ash.

OP, were you living under a rock? Or were you in the cabin with the Unibomber?

I shouldn't do this, but here goes. AMD said that way back in the day the number of cores was determined by how many Integer units a cpu had. A full FX has 8 Integer units, so they call it an 8 core. By that definition they are correct. You can disagree, fine, but that's where they get it from.

Normal consumer socket Intel cpus have 4 Integer units max. Yes, an i7 has HT, but still only has the 4 Integer units, so AMD uses this Integer unit advantage in their marketing. What's the big deal? Anyone intelligent looked at the reviews back in 2011, saw which architecture was better, bought what they wanted, and moved on. How are people still fighting and/or being outraged over AMD's marketing this many years later?

I had an e8400

Lol


I don't feel the need to defend my choice to a random person.

But it was a great buy

 

[Dresdenboy]

I thought it was better from an OS thread scheduling point of view to see the FX that way?

Yes. The CPU reports itself this way due to this.

 

[Dresdenboy]

FX-8150/8350 is quad core with HT and was never different

HT? Any proof?

 

[myocardia]

HT? Any proof?

He just means AMD's version of what Intel calls HT, what you and I call CMT.

 

[Starjack]

Does anybody know if Amd Athlon II x2 270 ( my processor in the desktop system) can unlock more cores?

 

[TeknoBug]

I remember when I first bought the FX 4350 and I was confused when I saw it saying 2 cores in task manager and CPU-z. lol

 

[TheRyuu]

I thought it was better from an OS thread scheduling point of view to see the FX that way?

It is but it's just how it's being displayed anyway. It's purely cosmetic and no one should really care what it says. It blurs the line on what a core is but if core = module then it has 4 cores. But really it doesn't matter does it? All that really matters is that the OS is scheduling stuff appropriately much the same way it works with Intel CPU's and their SMT.

 

[SlowSpyder]

I'm on Win10, it shows as "AMD FX-8320E Eight-Core Processor"

 

[frozentundra123456]

It is but it's just how it's being displayed anyway. It's purely cosmetic and no one should really care what it says. It blurs the line on what a core is but if core = module then it has 4 cores. But really it doesn't matter does it? All that really matters is that the OS is scheduling stuff appropriately much the same way it works with Intel CPU's and their SMT.

This. I never understood the concern about whether FX is 8 core or 4 cores. AMD markets it a 8 cores, so that is what I would consider it. In any case, the performance "is what it is", so to me it doesnt really make much difference.

If someone buys it because it is "8 cores", but doesnt check the performance and power use in the apps they are interested in, well, too bad. Should have done their homework before the purchase. BTW, OP, this is meant as a general statement, not a personal criticism of you.

 

[SPBHM]

I'm on Win10, it shows as "AMD FX-8320E Eight-Core Processor"

that's the name of the CPU

try task manager, what does it say for
cores:
logical processors:

5960X will say 8/16, 6770K 4/8
the FX is it 8/8 or 4/8?
you would expect 8/8 from the official name.

 

[ShintaiDK]

that's the name of the CPU

try task manager, what does it say for
cores:
logical processors:

5960X will say 8/16, 6770K 4/8
the FX is it 8/8 or 4/8?
you would expect 8/8 from the official name.

I can tell you that for you. It says 4 cores, 8 logical processors. CMT isn't cores and Microsoft doesn't approve it as such. Just as SMT isnt cores either, but logical processors. On the other hand it saves AMD owners, those 3 server owners left for licensing costs with software that is still per core based.

 

[AtenRa]

CMT Module = Core 1 and Core 2

If you disable Core 1, you will get the same Single Thread performance with Core 2.

You cannot do that with SMT, only with 2x identical cores.

What you lose in CMT is Multi Thread scaling compared to two independent Cores. But that was something they already have increased substantially in SR and EXV.

 

[ShintaiDK]

How is the FP output on CMT with 1 vs 2 cores? Oh right, the same

Deal with it, it was a marketing gimmick.

 

[AtenRa]

How is the FP output on CMT with 1 vs 2 cores? Oh right, the same

Deal with it, it was a marketing gimmick.

The FP alone doesnt make a CPU Core, deal with it. The CMT was made for the Server space, where you have multiple threads and high throughput. It was not the right design choice for the Desktop back in 2010-11.

Edit: And we should not forget, you can have the same Integer + FP Throughput disabling Core 1 or Core 2.

 

[ShintaiDK]

The FP alone doesnt make a CPU Core, deal with it. The CMT was made for the Server space, where you have multiple threads and high throughput. It was not the right design choice for the Desktop back in 2010-11.

Yet Microsoft doesn't accept it as cores. Neither in the client or server space.

CMT was never the right choice for anything, there is no such thing as a successful CMT design.

 

[VirtualLarry]

CMT was never the right choice for anything, there is no such thing as a successful CMT design.

Funny how you attack CMT, when Intel's initial attempts at HT/SMT were so abysmal, that they often resulted in negative scaling. That's why MS's "best practices" recommendation for W2K Server was to disable HT.

Intel's "marketing gimmick" indeed...

 

[AtenRa]

Funny how you attack CMT, when Intel's initial attempts at HT/SMT were so abysmal, that they often resulted in negative scaling. That's why MS's "best practices" recommendation for W2K Server was to disable HT.

Intel's "marketing gimmick" indeed...

If Intel had CMT it would be the new coming of Jesus but ....... AMD bla bla bla.

 

[ShintaiDK]

If Intel had CMT it would be the new coming of Jesus but ....... AMD bla bla bla.

You show your true spirit. Not that anyone was ever in doubt.

 

[ShintaiDK]

Funny how you attack CMT, when Intel's initial attempts at HT/SMT were so abysmal, that they often resulted in negative scaling. That's why MS's "best practices" recommendation for W2K Server was to disable HT.

Intel's "marketing gimmick" indeed...

Larry, tell me what CMT designs in history have been successful. And then tell me what SMT designs that have been successful.

I can tell you in advance that CMT got 0. And SMT got above 0.

Microsoft doesn't accept CMT as "real cores" either. And this is actually a benefit for those with core based licensing.

 

[zentan]

This. I never understood the concern about whether FX is 8 core or 4 cores. AMD markets it a 8 cores, so that is what I would consider it. In any case, the performance "is what it is", so to me it doesnt really make much difference.

If someone buys it because it is "8 cores", but doesnt check the performance and power use in the apps they are interested in, well, too bad. Should have done their homework before the purchase. BTW, OP, this is meant as a general statement, not a personal criticism of you.

+1

 

[Dresdenboy]

Larry, tell me what CMT designs in history have been successful. And then tell me what SMT designs that have been successful.

I can tell you in advance that CMT got 0. And SMT got above 0.

Microsoft doesn't accept CMT as "real cores" either. And this is actually a benefit for those with core based licensing.

There's also this story about more efficient task scheduling if threads are scheduled to different modules first. BTW, did you had any thought that it might be AMD, which is reporting the cores this way? Is the core and thread number used in Windows reported by AMD's CPU driver, it's patched microcode, or something else?

And do 8 threads with CMT and turbo taken out provide less than 4x the performance of a single thread (as there is no SMT)? If you compare CMT success based on DX11 games and other average code less well threaded, then you are right, that CMT is a useless solution in a mainly low threaded world.
Some examples from the Vishera launch:

Source: http://www.golem.de/news/a8-8350-mi...-kleinem-preis-und-hohem-takt-1210-95234.html
CB-Scaling:
CMT 1T -> 8T: 6.2X (even with shared FPUs).
SMT 1T -> 8T: 3.8X or 4.5X

The real culprit is basing the CMT on weaker cores first, accompanied by a less efficient cache subsystem, which didn't do as well in average apps (CB has no avg. behaviour).

And we might do a simple test and look, if the tasks run by the OP depend better ST or MT performance and how they scale. If they don't scale better from one thread to 8 than on a SMT CPU and give significantly lower results for a similar priced same thread count Intel CPU, then the OP should have done a better research or have asked the forum first, as frozentundra already said.

 

[Zstream]

Yet Microsoft doesn't accept it as cores. Neither in the client or server space.

CMT was never the right choice for anything, there is no such thing as a successful CMT design.

Has nothing to do with Microsoft accepting it as such or not. AMD went this direction because of Oracle licensing models only.

 

[VirtualLarry]

Larry, tell me what CMT designs in history have been successful. And then tell me what SMT designs that have been successful.

I can tell you in advance that CMT got 0. And SMT got above 0.

I thought that Sun had a CMT design? Or was that the cancelled one?

Are you saying, that if Intel attempted a CMT design, that they would also be unsuccessful, that there is something inherent in the design research that is flawed?

There's something to be said for "easier to implement" and "low-hanging fruit". Intel may have chosen SMT because of that, not because it's an inherently better technology. After all, they get what, 30% additional scaling, tops, from HT?

 

[ShintaiDK]

I thought that Sun had a CMT design? Or was that the cancelled one?

Are you saying, that if Intel attempted a CMT design, that they would also be unsuccessful, that there is something inherent in the design research that is flawed?

There's something to be said for "easier to implement" and "low-hanging fruit". Intel may have chosen SMT because of that, not because it's an inherently better technology. After all, they get what, 30% additional scaling, tops, from HT?

UltraSPARC T1 CMT design flopped too and was never seen again from them. T2 and forward was SMT based. The reality is simply no CMT designs have ever been successful due to the tradeoffs. Then it doesn't matter who makes it.

 

[Essence_of_War]

UltraSPARC T1 CMT design flopped too and was never seen again from them. T2 and forward was SMT based. The reality is simply no CMT designs have ever been successful due to the tradeoffs. Then it doesn't matter who makes it. [\Quote]

SPARC is actually pretty weird. It is one type of SMT, but I think even the modern versions (since the oracle acquisition) are still more like a barrel processor than like intel or ibm's implementations of SMT.

I thought that Sun had a CMT design? Or was that the cancelled one?

Rock was the cancelled one. It was supposed to be very different.

After all, they get what, 30% additional scaling, tops, from HT?

Intel only does 2-way SMT, but if you check out anandtech's power8 reviews, you can see the returns from more-way SMT in different workloads for power8:

http://www.anandtech.com/print/9567/the-power-8-review-challenging-the-intel-xeon-

in lzma benchmarks

1 thread - 100%
2 thread - 134% (+34%)
4 thread - 188% (+54% !)
8 thread - 213% (+25%)

in c-ray there is a slight regression for 8 threads, but you get:
1 thread - 100%
2 thread - 120% (+20%)
4 thread - 151% (+31%)
8 thread - 143% (-8%)