Why isnt a AMD 8350 better than a i76700k?

[Unhelpful]

1 minute ago, don_svetlio said:

It's not a fair comparison. If you compare the 8350 to the 2600K you will be making a more sensible argument/question.

But... But... Muh TekSyndicate and my cherry-picked benchmarks!

[RangerLunis]

40 minutes ago, Dackzy said:

The intel CPU's are better in single core performance and well newer.

Yeah that's totally true, but to make that comparison I would kinda have to know the answer to the question first, though if knew I would have totally used that as my example.

[Dackzy]

Just now, RangerLunis said:

Nothin hurt mang, sall gud. I understand those questions are pretty annoying but I really just wanted the knowledge, the 8350 is fine for lower end builds and intel for higher end. I came here understanding how each one compares to the other but not the science or information or whatever behind why.

Ohh well you see a intel cpu is just all around better specs when you look into the specs and as I said before the single core performance is much higher on a intel CPU than a AMD cpu, so unless you get a program that can use all 8 cores the 8350 has it will look like it sucks ass. We can really get into the technicalities of this, but it would take such a long time.

[Dackzy]

3 minutes ago, Dackzy said:

Ohh well you see a intel cpu is just all around better specs when you look into the specs and as I said before the single core performance is much higher on a intel CPU than a AMD cpu, so unless you get a program that can use all 8 cores the 8350 has it will look like it sucks ass. We can really get into the technicalities of this, but it would take such a long time.

and I don't really have the time and I am lazy  

[henkka_scorpio]

http://cpuboss.com/cpus/Intel-Core-i7-6700K-vs-AMD-FX-8350

 there are the specs of both and what is better in wich.

[jett473]

1 hour ago, RangerLunis said:

This is probably a dumb question and been asked quite a bit before, but why would a AMD not be better than a i76700k? I'm asking this out of complete curiosity.

 

If you look at the specs, a AMD 8350 has the same clock speed as a i76700k but a 8 whole cores compared to the i7's 4. 

 

So why does the AMD 8350 not perform as well as the I7?

 

I actually personally own a i76700k and am quite happy with it, but i'm still generally curious, could this have something to do with how much cache each processor has or just something in build quality?

clock speed means nothing when looking at CPUs that aren't in the same family (like i5 4xxx, i7 6xxx). The 8350's physical cores are rather weak, and it's a pretty old CPU (it came out in 2012, I believe). that's kind of the basic idea, really. 6700k has better single core performance, and hyper threading so it has 8 threads (in a simpler term, it's got the same performance as a 8 core.) 

[don_svetlio]

3 minutes ago, jett473 said:

clock speed means nothing when looking at CPUs that aren't in the same family (like i5 4xxx, i7 6xxx). The 8350's physical cores are rather weak, and it's a pretty old CPU (it came out in 2012, I believe). that's kind of the basic idea, really. 6700k has better single core performance, and hyper threading so it has 8 threads (in a simpler term, it's got the same performance as a 8 core.) 

I'd digress - the cores themselves appear weaker in games due to the shared FPU. In actual multi-threaded applications it's not nearly as poor as in certain games.

Also, the 8 threads =/= cores. A physical core is 40-50% faster than a HT.

[Kobathor]

1 hour ago, tbake0155 said:

Same architecture (x86), different micro-architecture

x86 is the instruction set, not the architecture. Skylake is the codename for the 14nm architecture used in the 6th gen CPU's from Intel, and Bulldozer is the codename for the 32nm architecture used in lots of AMD CPU's. Architecture is not instruction set.

[tbake0155]

9 minutes ago, don_svetlio said:

I'd digress - the cores themselves appear weaker in games due to the shared FPU. In actual multi-threaded applications it's not nearly as poor as in certain games.

Also, the 8 threads =/= cores. A physical core is 40-50% faster than a HT.

No I don't think the cores appear weaker because of the shared FPU.  Unless you have some specific instance of game developers using the AVX 256 instruction set don't just toss that out there because it sounded good to you...  As has regularly been stated on these forums by users more educated than myself, software has shamefully lagged behind hardware in it's adoption of new instruction sets, like AVX.

 

I'm open to being proved wrong but I have feeling you don't know as much about the shared FPU as you appear to to someone uneducated, although it's quite possible you understand better than I do and that you didn't explain fully.

 

The cores are weaker because they are weaker and that is because of the micro architecture and the way they were designed.  It's not because of the FPU.  The FPU sharing only causes a problem under AVX instructions that use 256 bits.  If there are AVX 256 bit instructions constantly being processed, then I would guess that the fx series would perform pretty miserably compared to intels in IPC and in multi threading.

 

But that type of scenario rarely even exists.

 

The FPU sharing and the weakness of the IPC of the cores are totally unrelated!

 

 

Also you are correct about hyper threads not being anywhere near as potent as real cores, and they are only most potent when software is actually optimized for hyper threading.  So not only are they slower than real cores, they need more optimization to be relevant.

[DXMember]

https://www.akkadia.org/drepper/cpumemory.pdf

[tbake0155]

18 minutes ago, Kobathor said:

x86 is the instruction set, not the architecture. Skylake is the codename for the 14nm architecture used in the 6th gen CPU's from Intel, and Bulldozer is the codename for the 32nm architecture used in lots of AMD CPU's. Architecture is not instruction set.

I thought I replied to this and it disappeared, sorry if it double posts.

 

I'm fairly positive you are mistaken!  The x86 instruction set defines the x86 architecture.  The micro-architecture is the actual physical implementation of transistors and components to accomplish the instruction set that defines the architecture.

 

fx and i5 chips have the same architecture (x86), although they might not have the most up-to-date instruction sets of the architecture.  Node shrinking is just a different form of micro architecture.

[Prysin]

23 minutes ago, tbake0155 said:

No I don't think the cores appear weaker because of the shared FPU.  Unless you have some specific instance of game developers using the AVX 256 instruction set don't just toss that out there because it sounded good to you...  As has regularly been stated on these forums by users more educated than myself, software has shamefully lagged behind hardware in it's adoption of new instruction sets, like AVX.

 

I'm open to being proved wrong but I have feeling you don't know as much about the shared FPU as you appear to to someone uneducated, although it's quite possible you understand better than I do and that you didn't explain fully.

 

The cores are weaker because they are weaker and that is because of the micro architecture and the way they were designed.  It's not because of the FPU.  The FPU sharing only causes a problem under AVX instructions that use 256 bits.  If there are AVX 256 bit instructions constantly being processed, then I would guess that the fx series would perform pretty miserably compared to intels in IPC and in multi threading.

 

But that type of scenario rarely even exists.

 

The FPU sharing and the weakness of the IPC of the cores are totally unrelated!

 

 

Also you are correct about hyper threads not being anywhere near as potent as real cores, and they are only most potent when software is actually optimized for hyper threading.  So not only are they slower than real cores, they need more optimization to be relevant.

the FPU is also used for Physics. Which is why in some games relying on CPU physics, FX can take a hard hit.

 

That being said, a FX 8320 at 4.6GHz will score about 9k points in Firestrike. A stock i7 4790k scores around 12k points in the physics test. So while there is a performance delta, even this isnt the only reason why FX blows.

[Kobathor]

1 minute ago, tbake0155 said:

I thought I replied to this and it disappeared, sorry if it double posts.

 

I'm fairly positive you are mistaken!  The x86 instruction set defines the x86 architecture.  The micro-architecture is the actual physical implementation of transistors and components to accomplish the instruction set that defines the architecture.

 

fx and i5 chips have the same architecture (x86), although they might not the most up-to-date instruction sets of the architecture.  Node shrinking is just a different form of micro architecture.

v.v

Quote

An instruction set, or instruction set architecture (ISA), is the part of the computer architecture related to programming, including the native data types, instructions, registers, addressing modes, memory architecture, interrupt and exception handling, and external I/O.

 

x86-64 is what AMD and Intel use now for their instruction set architecture. It's not something that is part of the CPU, like a processing core, but a logical thing that instructs the CPU on how to do things. CPU architecture (or microarchitecture, like you said) is the physical layout of the CPU. They are different things, both called "architecture." Normally, when talking about CPU's, you refer to the microarchitecture as the "architecture." 

 

So, FX and i5's do not have the same architecture, they have the same ISA.

[Coaxialgamer]

there are multiple reasons really :

 

-first , as many have mentioned, the i7 has a far better IPC , which means it can do  much more work per clock cycle ( hertz ).

 

-secondly , the fx 8350 isn't truly an "8 core" processor . What it does to achieve this core count is a bit complicated : a typical intel core ( or amd cores before bulldozer ) consist of typically 1 FPU and 1 integer unit . These are what process data . What AMD has done here is make "modules" . The fx 8350 contains 4 of these modules.  A piledriver module ( fx 8350 ) contains 2 integer units ( as you would find in 2 cores ), but they share one FPU  along with cache . This amounts to not quite two cores , without being single core either . Rather , one core and a half ( though this is oversimplified). 

This is called CMT ( clustered multi-threading ). Compared to SMT ( what intel uses for hyperthreading ), CMT penalizes performance if not taken advantage of ( where SMT only rewards you if used ).

 

 

A few things to mention : the fx 8350 is based on old SOI 32nm process tech (globalfoundries ), instead of intel's 14nm FinFet ( or tri-gate. used in the 6700k ). Now you must know that even nodes that sport the same process size ( measured in nm ) are actually very different . Now it turns out that GlobalFoundrie's 32nm process was quite a bit worse than intel's 32nm. Intel's processes are typically the best ( for example, tsmc's 16nm LP process was more closely related to intel 22nm than 14nm ). And intel is now at 14nm compared to 32nm on the fx . This means that the amd chip has larger transistors that :

-require larger chips , thus are more expensive to produce 

-produce more heat 

-require more energy to switch on and off

-leak more ( further increasing power consumption )

 

But you have to realise that the fx 8350 is based on 2011 tech (back when the 2600k was around )

 

One thing I will hand AMD however, is that it seems that bulldozer has no "cold bug" , thus usually operate at sub-zero temps , which comes in handy when doing ln2 overclocking . The current world recored for overclocking was broken using the fx 8150 ( at 8.49ghz ).

[tbake0155]

19 minutes ago, Kobathor said:

v.v

x86-64 is what AMD and Intel use now for their instruction set architecture. It's not something that is part of the CPU, like a processing core, but a logical thing that instructs the CPU on how to do things. CPU architecture (or microarchitecture, like you said) is the physical layout of the CPU. They are different things, both called "architecture." Normally, when talking about CPU's, you refer to the microarchitecture as the "architecture." 

 

So, FX and i5's do not have the same architecture, they have the same ISA.

In fact, not all FX and i5's have the same ISA... The new skylake i5 has a more up-to-date ISA than any of the FX processors, including AVX 512 bit instructions.  So all cpus with the same ISA are the same architecture, but not all cpus of the same architecture have the same ISA.

 

I'm not sure what you mean about cpu architecture and micro architecture being synonymous, because in engineering they have specific meanings as I described.  The x86 architecture goes way back, the current x86 ISA includes all of the original instructions from way back, but includes more.

 

x86-64 is x86 with 64 bits per instruction vs 32 bit, contains all of the x86 instructions, and additional 64 bit instructions.  Hopefully that helps clear things up.