What do people actually mean by "stronger cores"

[Timopomer]

People say Intel's cores are better in single core and they have "stronger cores" what does that even mean?

 

it cant be the multiplier or the frequency because they dont seem to be perfectly correlated with speed.

 

 

what is the difference?

[Scepsis]

It refers to the number of instructions per clock.

[Mo5]

IPC

[Timopomer]

ok, so are there actual numbers? statistics? some kind of comparison? 

[Scepsis]

Look at the benchmarks. That's as definitive proof as anything.

Intel is just faster than AMD on a clock per clock basis.

[minibois]

ok, so are there actual numbers? statistics? some kind of comparison? 

There are programs on which you can see how much every single core is doing.

That way people can conclude if a game uses all cores, or just 2.

Look up some single core performance reviews

[GuruMeditationError]

Stronger core = More efficient. Get more done in the given time, so more "powerful" (not related to cycle rate).

[Bogica]

Higher single core performance than AMD has. And most applications and games use 2-4 cores so higher performance per core> more cores in gaming and similar. Only professional software and a small amount of other programs use more than 4 cores.

[dandrix]

Intel has more advanced CPU architecture.

[i_build_nanosuits]

ok, so are there actual numbers? statistics? some kind of comparison? 

called: Benchmarks

[Timopomer]

called: Benchmarks

no.

 

i mean IPC compared to other things like frames per second etc. to find out if high frequency is pointless? maybe 1 core that does many instructions is better at almost everything? i know that rendering leverages more cores.

[harrynowl]

no.

 

i mean IPC compared to other things like frames per second etc. to find out if high frequency is pointless? maybe 1 core that does many instructions is better at almost everything? i know that rendering leverages more cores.

IPC and minimum FPS correlate quite well, look at some minimum FPS graphs for CPU intensive games and I expect intel i5/7s are at the top with amds towards the bottom.

 

 

The i5 and FX are both @4.7GHz but the i5 is quite a bit faster, so in workloads that don't use 8+ cores an i5 is going to be quite a bit faster

[i_build_nanosuits]

no.

 

i mean IPC compared to other things like frames per second etc. to find out if high frequency is pointless? maybe 1 core that does many instructions is better at almost everything? i know that rendering leverages more cores.

again, you have to rely on benchmarks to be able to quantify those kind of datas.

[Hieb]

It means that more instructions can be completed in a single clock cycle. Intel's IPC is around 50% faster than AMD's currently (comparing Haswell to Piledriver & Steamroller), meaning that if both an Intel CPU and AMD CPU were clocked at the same speed (measured in GHz) the Intel would perform about 50% better (theoretically; ignoring other factors like cache and design of the specific application).

 

GHz = how many cycles completed per second

IPC = how much gets done in each one of those cycles

 

Then there's also other resources. Intel's cache has lower latency than AMD's cache resulting in less delays in processing. AMD's Bulldozer/Piledriver/Steamroller CPUs have a lot of shared resources. You may have heard the term "modules". Each pair of cores in these architectures is referred to as a module, because each of these pairs share a single decoder, floating point unit and L2 cache. Result is processing gets stalled at times when one core can't access the resource it needs.

[Timopomer]

ok so now that i understand the difference, how does amd/intel control the IPC? its a piece of silcon after all.

[Scepsis]

Simple answer? Microarchitecture.

[i_build_nanosuits]

ok so now that i understand the difference, how does amd/intel control the IPC? its a piece of silcon after all.

trough architecture design, improvement and precise engeniering pioneer research and development and spending billions of dollars on it...perfecting every detail of the micro-architecture.

...and while intel is doing that, AMD is strapping more of their old slow cores on their power hungry inneficient chips.

[Timopomer]

trough architecture design, improvement and precise engeniering pioneer research and development and spending billions of dollars on it...perfecting every detail of the micro-architecture.

...and while intel is doing that, AMD is strapping more of their old slow cores on their power hungry inneficient chips.

 

ok,why arent processors compared by instructions per SECOND? wouldn't that be much more comparable? and informative?

[Scepsis]

Clock speeds are a more appealing number to market.

[i_build_nanosuits]

ok,why arent processors compared by instructions per SECOND? wouldn't that be much more comparable? and informative?

yes, but it's impossible to know exactly how many instructions a CPU can process per second, that's why...also it will depend greatly on the workloads, type of instructions, number of threads etc.

[typographie]

ok,why arent processors compared by instructions per SECOND? wouldn't that be much more comparable? and informative?

 

The non-tech-savvy mass market prefers a simple, single number to compare (like clock frequency in Hz), but you can't really boil down IPC that far. The only way to know these things is to compare processors using real-world benchmarks. Nothing else really tells you what you need to know.

[Timopomer]

The non-tech-savvy mass market prefers a simple, single number to compare (like clock frequency in Hz), but you can't really boil down IPC that far. The only way to know these things is to compare processors using real-world benchmarks. Nothing else really tells you what you need to know.

true,but i have classmates comparing phones by how many GHz the processors run at and it REALLY pisses me off because it doesnt really say anything about the speed the processor is running at.

[typographie]

true,but i have classmates comparing phones by how many GHz the processors run at and it REALLY pisses me off because it doesnt really say anything about the speed the processor is running at.

 

It can mean something, as long as it's comparing two similar architectures. It's useful to know that an i5-4440 runs at 3.3 GHz, compared to an i5-4590 at 3.7 GHz. But they're both as fast or often faster than an FX-9590 at nearly 5 GHz.

[i_build_nanosuits]

true,but i have classmates comparing phones by how many GHz the processors run at and it REALLY pisses me off because it doesnt really say anything about the speed the processor is running at.

no it does not unless they use the same micro-architecture, just like it is for PC micro-processors...you have to run a set of benchmarks in similar conditions to be able to tell how they perform compaired to another...in both single-threaded workloads AND multi-threaded workloads...also benchmarks using different sets of instructions.