If Moore's law implies that the amount of transistors we can fit into a certain amount of space keeps getting bigger and bigger, is it really necessary to decrease the sophistication of processor logic? Why can't we have the best of both worlds - several processors that are all just as sophisticated as the processors that came before the multi-core era?
This comment was marked helpful 0 times.
placebo
Given my knowledge of what's out there, I'd venture to say that even a singleton core on a modern processor is far more complex+powerful than those that came before the multicore era (e.g. one core of an i7 Ivy Bridge chip is faster and more sophisticated than say any Pentium 4), simply because of the many advancements that have come since then (SSE4 & AVX for example).
The impression I got from class is that you can always build a more complex single core CPU with the same number of transistors as today's multicore CPUs. But while those fancy add-ons would blow the multicore chip away in single-threaded applications, they'd only get you so far in multi-threaded apps since the power wall would limit the chip's highest clock frequency (arguably the single most important factor of single core chips). It wouldn't be able to compete with true multicore processors (which are still pretty complex in themselves).
This comment was marked helpful 1 times.
kayvonf
@tborenest: your thinking is correct. As @placebo says, it's really a matter of how sophisticated a single core is today, vs. how sophisticated a single core at today's transistor counts could be. You could implement very sophisticated branch predictors, prefetchers, huge out-of-order windows, and add in a huge data cache, or you could choose more modest implementations of all these features and place a few cores on the chip. Studies show the former would (for most single-threaded workloads) increase your performance a few percentage points, or maybe at best a factor of two. However, in the the latter case, you now have four or six cores to give a multi-threaded workload a real boost. I should change this slide to say: Use increasing transistor count for more cores, rather than for increasingly complex logic to accelerate performance of a single thread.
As keep in mind that by multi-threaded can mean multiple single-threaded processes active in the system, running on different cores. This is helpful on your personal machine when multiple programs are running, and it is very helpful on servers, where it's common to run several virtual machines on the same physical box.
If Moore's law implies that the amount of transistors we can fit into a certain amount of space keeps getting bigger and bigger, is it really necessary to decrease the sophistication of processor logic? Why can't we have the best of both worlds - several processors that are all just as sophisticated as the processors that came before the multi-core era?
This comment was marked helpful 0 times.
Given my knowledge of what's out there, I'd venture to say that even a singleton core on a modern processor is far more complex+powerful than those that came before the multicore era (e.g. one core of an i7 Ivy Bridge chip is faster and more sophisticated than say any Pentium 4), simply because of the many advancements that have come since then (SSE4 & AVX for example).
The impression I got from class is that you can always build a more complex single core CPU with the same number of transistors as today's multicore CPUs. But while those fancy add-ons would blow the multicore chip away in single-threaded applications, they'd only get you so far in multi-threaded apps since the power wall would limit the chip's highest clock frequency (arguably the single most important factor of single core chips). It wouldn't be able to compete with true multicore processors (which are still pretty complex in themselves).
This comment was marked helpful 1 times.
@tborenest: your thinking is correct. As @placebo says, it's really a matter of how sophisticated a single core is today, vs. how sophisticated a single core at today's transistor counts could be. You could implement very sophisticated branch predictors, prefetchers, huge out-of-order windows, and add in a huge data cache, or you could choose more modest implementations of all these features and place a few cores on the chip. Studies show the former would (for most single-threaded workloads) increase your performance a few percentage points, or maybe at best a factor of two. However, in the the latter case, you now have four or six cores to give a multi-threaded workload a real boost. I should change this slide to say: Use increasing transistor count for more cores, rather than for increasingly complex logic to accelerate performance of a single thread.
As keep in mind that by multi-threaded can mean multiple single-threaded processes active in the system, running on different cores. This is helpful on your personal machine when multiple programs are running, and it is very helpful on servers, where it's common to run several virtual machines on the same physical box.
This comment was marked helpful 1 times.