question about CPU in general

[Jacob Morris]

This is just a educational question, but why is it that an Intel i5-4570 (3.2 Ghz - 3.6 Ghz quad core) gets a higher CPU benchmark score than an AMD FX-6300 (3.5 Ghz - 4.1 Ghz hex core)?. Does the Intel just do more work in each clock cycle or what? I have just been wondering about it,

 

Thanks

[helping]

Yep, it's what you think

 

tl;dr different architectures 300 Intel mhz =/= 300 AMD mhz =/= 300 ARM mhz

[Izaya Orihara]

Core per clock Intel has about double the single core performance. Intel Cores are stronger in general. 

[technovore]

Think of cores as wheels. there might be 4 wheels in both CPU's and they might both spin at the same speeds, but one set of wheels could be larger than another set of wheels..

I did not come up with this comparison, I read it somewhere..

[TheRetiredSlave]

This is just a educational question, but why is it that an Intel i5-4570 (3.2 Ghz - 3.6 Ghz quad core) gets a higher CPU benchmark score than an AMD FX-6300 (3.5 Ghz - 4.1 Ghz hex core)?. Does the Intel just do more work in each clock cycle or what? I have just been wondering about it,

 

Thanks

 

Yep,, Intel's architecture is stronger, so that quadcore i5 is more powerful even though it has less cores than the FX hexcore..

[Jacob Morris]

Thanks everyone for the explanations

[Nineshadow]

Even though they both are x86,they are quite different.Intel has way better single core performance.

 

Also,the AMD isn't really 6 cores.It has 3 modules,which I would call the actual cores,which have 2 cores each.That means those cores have to share resources,memory,etc.It's just a physical method of hyperthreading,really.

[Faa]

Also,the AMD isn't really 6 cores.It has 3 modules,which I would call the actual cores,which have 2 cores each.

But if you say that each module has 2 cores and you have 3 modules how can it be a fake 6core? The only fake 8core is the i7, not the 8350's.

[DildorTheDecent]

But if you say that each module has 2 cores and you have 3 modules how can it be a fake 6core? The only fake 8core is the i7, not the 8350's.

None of them are really true 8 cores. the 8350 (i think) has 4 modules with 2 cores each while the i7 has 4 cores and 8 threads.

[deathjester]

But if you say that each module has 2 cores and you have 3 modules how can it be a fake 6core? The only fake 8core is the i7, not the 8350's.

 

Cus they share cache on the chip section. It is just physical hyperthreading. If they didn't share resources then the 8350 would outperform the I7 at rendering. They don't. Slightly better than HT, but per core performance is much slower, so HT > integer cores. The less threads/cores used the more Intel is ahead, which is why I5 >8350 in games.

 

With perfect optimization they are close to being the same due to single core performance being so much faster on Intel. In a few years? This won't even be a discussion. We will be discussing which chip renders better and what has the best price point. That is in a couple of years though. ATM I5 is a better chip for gaming. 8320/50 is a better chip for rendering.  This is the whole reason AMD made Mantle. Optimization also helps Intel's but it helps the AMD's a lot more, except for maybe the I3. The I3 with 2C 4 threads skyrockets under Mantle. The I7 is overkill. 

 

[Faa]

Cus they share cache on the chip section. It is just physical hyperthreading. If they didn't share resources then the 8350 would outperform the I7 at rendering. They don't. Slightly better than HT, but per core performance is much slower, so HT > integer cores. The less threads/cores used the more Intel is ahead, which is why I5 >8350 in games.

 

With perfect optimization they are close to being the same due to single core performance being so much faster on Intel. In a few years? This won't even be a discussion. We will be discussing which chip renders better and what has the best price point. That is in a couple of years though. ATM I5 is a better chip for gaming. 8320/50 is a better chip for rendering.  This is the whole reason AMD made Mantle. Optimization also helps Intel's but it helps the AMD's a lot more, except for maybe the I3. The I3 with 2C 4 threads skyrockets under Mantle. The I7 is overkill. 

Cache being shared doesnt make it a false 8 core and it's not necessarly the main problem, see below. The L3 cache is always shared between all cores if we have to believe this diagram (http://www.realworldtech.com/includes/images/articles/sandy-bridge-1.png?71da3d) but BD is sharing L2 even between 2 cores where as Intel doesnt.

Just to start off, having 8 integer cores makes your OS recognize 8 cores. We're seeing here 2x 128bit FP's being available to 2 cores, they can run each independetely or combine together like for Superpi but the main problem is the FPU's are shared between the 2 cores that's complety out of the tradition. Being somewhat the same as hyperthreading? Not even close. If you have an app with 2 threads and an instruction like 3+2 + 5/7, 3+2 would push into the ALU and 5/7 in the FPU - AMD would do this in 2 cycles where as Intel would do this with Hyperthreading within a single cycle that would make a 100% performance difference.

 

 

None of them are really true 8 cores. the 8350 (i think) has 4 modules with 2 cores each while the i7 has 4 cores and 8 threads.

Why not, because it uses some garbage FPU's? Because they use 2x 128bit FP's for 2 cores where as Intel is more mature and uses a single 256bit FP per core? The Integer cores aren't fake at all, if they only had 4 integer cores you would only see 4cores or threads in Windows aka the fx4300. AMD didnt really care about FP performance aslong as they put enough integer cores for some gimmick corecount sales.

[deathjester]

 

Cache being shared doesnt make it a false 8 core and it's not necessarly the main problem, see below. The L3 cache is always shared between all cores if we have to believe this diagram (http://www.realworldtech.com/includes/images/articles/sandy-bridge-1.png?71da3d) but BD is sharing L2 even between 2 cores where as Intel doesnt.

Just to start off, having 8 integer cores makes your OS recognize 8 cores. We're seeing here 2x 128bit FP's being available to 2 cores, they can run each independetely or combine together like for Superpi but the main problem is the FPU's are shared between the 2 cores that's complety out of the tradition. Being somewhat the same as hyperthreading? Not even close. If you have an app with 2 threads and an instruction like 3+2 + 5/7, 3+2 would push into the ALU and 5/7 in the FPU - AMD would do this in 2 cycles where as Intel would do this with Hyperthreading within a single cycle that would make a 100% performance difference.

 

 

Why not, because it uses some garbage FPU's? Because they use 2x 128bit FP's for 2 cores where as Intel is more mature and uses a single 256bit FP per core? The Integer cores aren't fake at all, if they only had 4 integer cores you would only see 4cores or threads in Windows aka the fx4300. AMD didnt really care about FP performance aslong as they put enough integer cores for some gimmick corecount sales.

 

 

It shares resources. Performance suffers compared to cores that don't share resources between 2 cores. In the end performance is all that matters. It is physically a core, but also physically not a core. You can argue this til you are blue in the face and both sides of the argument are correct. It is on 4 chip sections not 8, and it has to share cache. One core on a chip section can be put in a wait state for cache while the other core is busy. 

 

So it is really what do you consider a core?  If going by performance and how they both have to share resources on logical cores and physical cores, you might as well call Intel's I7 a 8 core, because it does the same thing, but not physically and the I3/HT a 4 core, but that wouldn't make sense because the I5 is a true 4 core. 

 

In the end core count between the two means nothing. We should be looking at benchmarks. Buy the best chip on the benchmarks (make sure they are from multiple sources) for the stuff you will be doing and games you play and call it a day. Core count between the two means nothing because the 8350 can be faster in rendering then an I5 and get half the fps in a mmo on double the "cores".   

 

If you want to test how much better one is then the other as far as almost physical 8 core and logical core (and this wouldn't make it a better chip because the I7 has higher total in multi core) I guess you could take single core performance, multicore performance in cinebench and divide it by 8 on the I7 and 8350.

 

You could do same on a hexa and say just how much better a solid core is then a HT core. That is just rendering though. That is what makes this "blah blah is better because of cores" so stupid. What makes something better is benchmarks.