FX-4100/FX-4170 vs 2500K(Intel vs AMD)

aesma

Honorable
May 26, 2012
27
0
10,530
Why 2500K performs so much better than either one of these AMD processors?

Many people say that 2500k beasts FX-4100 and FX-4170 no big deal, and I've seen the benchmarks myself too, but I wonder.. WHY is that so?
AMD has higher clock speeds and has bigger chase expect for L1 cache and of course, they are alot cheaper.
They have the same amount of cores and they both share the same 32nm technology, but they have differences in cache technology, power consumption and of course different socket and architecture. So what of these things makes 2500K and generally speaking all the Intel's processors perform so much better?

The reason why I wonder about this stuff, is that I am currently studying computer technologies independently, so that one day I can build my own computers, and because it is alot fun too.
 

Robi_g

Distinguished
Jan 28, 2012
510
0
19,010
AMD cores share some rescources, the FX4xxx has 2 bulldozer 'modules' whereas the i5 has 4 actual cores.
Also IPC matters a lot, that is performance per cycle, for example (hypothetically) the FX could do 10 things per cycle whereas the i5 (with better IPC) could do 15, those figures were made up to demonstrate this so don't quote them :p. so a higher clock speed is used to lessen this on FXs.
Difference like power consumption and cache stuff is just down to the architecture.
 

$hawn

Distinguished
Oct 28, 2009
854
1
19,060
Archtevctural differences.....a huge one at that

An intel core can do much more work per GHz than an AMD currently can. Plus, the share an FPU between 2 integer cores.
Yes they have big L3 caches, but they are quite slow, ie, quite high latencies. Plus they have poor branch predictors = more cache misses, ie more performance loss.

Also, Intels 32nm is FAR more superior and matured vs GloFo leaky 32nm :)
 

The Greater Good

Distinguished
Jan 14, 2010
342
0
18,810
" They have the same amount of cores "

Not really. AMD uses modules which is 2 integer cores and one floating point unit per. So two modules is called a quad core, but it's not in the classic definition; it's really a dual core with two extra integer cores.

The i5 is a true quad core (4 integer cores and 4 FPUs) and having two more floating point units gives the i5 a big edge over Bulldozer. Even the FX8xxx can't touch the i5 in many benchmarks.
 

aesma

Honorable
May 26, 2012
27
0
10,530
Ok that enlightens me alot, thank you guys. It seems that Intel's processors seem to work more deliberately and logically when AMD kinda just tryes to push their processors to work more effectively and make their processors LOOK more powerful?
Also did I understand right, integer core is like a virtual core, not a real, physical one?
 
Yes. Its like hyperthreading but it's garbage. CMT (Clustered Multi Threading) is just an ‘invention’ by AMD’s marketing department. They invented a term that sounds close to SMT (Simultaneous Multithreading), in an attempt to compete with Intel’s HyperThreading. HyperThreading is just a marketing-term as well, but it is Intel’s term for their implementation of SMT, which is a commonly accepted term for a multithreading approach in CPU design, and has been in use long before Intel implemented HyperThreading. People seem to think that CMT is just as valid a technology as SMT and they think that the two are closely related, or even equivalent. So here the difference between CMT and SMT; With single-threading, each thread has more ALUs (Arithmetic Logic Unit) with SMT than with CMT. With multithreading, each thread has less ALUs (effectively) than CMT. That’s why SMT works, and CMT doesn’t: AMD’s previous CPUs also had 3 ALUs per thread.
 

aesma

Honorable
May 26, 2012
27
0
10,530
Wait.. Did I get it right?
AMD has less physical cores, meaning they have less ALUs meaning that all their virtual cores share the same ALUs meaning they cannot carry us much processes simultaneously as Intel's processors? But how is that then a good thing that with multithreading each thread has less ALUs with SMT than CMT? Or did you mean by effectively that more threads can be carried thru a single ALU at the time or something like that? So more threads can be moved per ALU or wtf..
Pardon me if I didn't understood correctly, this is whole new area for me.