4 ms·

> top speed is actually the function of the heat generated That knowledge is rather dated and probably goes back to the Pentium 4 era. Top speed is a function

> top speed is actually the function of the heat generated That knowledge is rather dated and probably goes back to the Pentium 4 era. Top speed is a function of heat, but also of the process technology and the concurrency available+used. It's best to wrap your head around it by going back to first principles: > The dynamic power consumed by a CPU is approximately proportional to the CPU frequency, and to the square of the CPU voltage: > P = C V^2 f > where C is capacitance, f is frequency, and V is voltage. [1] Capacitance goes down with smaller transistors, voltage and frequency goes up with higher sequential performance given the same architecture. So, by replacing the need for sequential performance with more concurrency (vectorization/multicore/gpu acceleration), the need for power goes down dramatically. Process technology is another big factor. On a related note: Comparing ARM with Core M just using some concurrency using benchmarks is not really comparing Apples to Apples - I expect Core M to have a higher sequential performance - so if you want to have the same performance on ARM, you'll need more concurrency in the software. [1] http://en.wikipedia.org/wiki/CPU_power_dissipation http://en.wikipedia.org/wiki/CPU_power_dissipation

seanmcdirmid 12y ago
I believe you mean parallelism, not concurrency right? Concurrency is dealing (correctly) with more things at the same time, parallelism is doing more things at the same time. The Pentium 4 had a very deep pipeline, which got very hot and wasn't fast compared to its clock. The pentium M went back to more instruction level parallelism via shallower parallel pipes (multi core came later). I don't think ARM is competitive with Intel even with parallel benchmarks; Intel does performance well and doesn't play many tricks with it, ARM's advantage has always been power efficiency and cost.
- m_mueller 12y ago
  > I believe you mean parallelism, not concurrency right? Concurrency is dealing (correctly) with more things at the same time, parallelism is doing more things at the same time. Yes, parallelism makes more sense in this context. I wanted to express it in a more general way because running multiple programs on different cores at the same time I understand as concurrency, which is also made more efficient by parallel performance rather than sequential - IMO the terms are a little bit fuzzy at times, i.e. it's not that easy to convey the ideas behind parallel programming easily such that there is no ambiguity. Edit: About ARM vs. Core M in benchmarks: After looking at it again, apparently the two are still quite far apart, you were right about their performance numbers: [1] [1] http://www.fool.com/investing/general/2014/09/22/intel-corporations-core-m-significantly-outpaces-a.aspx http://www.fool.com/investing/general/2014/09/22/intel-corpo...
Symmetry 12y ago
I don't think that anybody could reasonably read top speed being a function of heat as top speed only being a function of heat. I wouldn't really associate that idea with the Pentium IV era either since the problem then was that most people weren't thinking about power use at all. Even that equation you gave says that if you increase P you will be able to increase f, though the interplay between f and V means that your increases are sub-linear.
- m_mueller 12y ago
  Let's look at the quote again. > top speed is actually the function of the heat generated emphasis mine. > Even that equation you gave says that if you increase P you will be able to increase f, though the interplay between f and V means that your increases are sub-linear. I don't get your point. Did I write that performance and power is not related?
  - jerf 12y ago
    I feel this is one of those rare cases where some feedback about the downvote I just gave is called for. I've downvoted you because someone has asked you to apply the principle of charity to someone's post, and you've chosen to dig in instead. Even if you want to continue this otherwise-interesting discussion, please don't do it this way.
    - m_mueller 12y ago
      Thanks for the explanation. I didn't read GPs post as a call for charity on OP. Looking at this phrase.. > I don't think that anybody could reasonably read top speed being a function of heat as top speed only being a function of heat .. I still don't see it this way. My post wasn't meant to be an attack on OP, rather I wanted to expand some points about chip performance that seemed to me weren't clear. The post I'm quoting however did read uncharitable to me.
  - Symmetry 12y ago
    Again, I don't understand how you got your reading. If something is the function of something else that means that the function is singular, not that the function only takes a singular argument. Obviously there are more things that are a function of heat than just clock speed so the op was wrong in that sense. That's the sort of mistake I make in revision all the time so I overlooked it. But even taking it literally it doesn't actually hurt the point the op was trying to make.

seanmcdirmid 12y ago

I believe you mean parallelism, not concurrency right? Concurrency is dealing (correctly) with more things at the same time, parallelism is doing more things at the same time. The Pentium 4 had a very deep pipeline, which got very hot and wasn't fast compared to its clock. The pentium M went back to more instruction level parallelism via shallower parallel pipes (multi core came later). I don't think ARM is competitive with Intel even with parallel benchmarks; Intel does performance well and doesn't play many tricks with it, ARM's advantage has always been power efficiency and cost.

m_mueller 12y ago
> I believe you mean parallelism, not concurrency right? Concurrency is dealing (correctly) with more things at the same time, parallelism is doing more things at the same time. Yes, parallelism makes more sense in this context. I wanted to express it in a more general way because running multiple programs on different cores at the same time I understand as concurrency, which is also made more efficient by parallel performance rather than sequential - IMO the terms are a little bit fuzzy at times, i.e. it's not that easy to convey the ideas behind parallel programming easily such that there is no ambiguity. Edit: About ARM vs. Core M in benchmarks: After looking at it again, apparently the two are still quite far apart, you were right about their performance numbers: [1] [1] http://www.fool.com/investing/general/2014/09/22/intel-corporations-core-m-significantly-outpaces-a.aspx http://www.fool.com/investing/general/2014/09/22/intel-corpo...

Symmetry 12y ago

I don't think that anybody could reasonably read top speed being a function of heat as top speed only being a function of heat. I wouldn't really associate that idea with the Pentium IV era either since the problem then was that most people weren't thinking about power use at all. Even that equation you gave says that if you increase P you will be able to increase f, though the interplay between f and V means that your increases are sub-linear.

m_mueller 12y ago
Let's look at the quote again. > top speed is actually the function of the heat generated emphasis mine. > Even that equation you gave says that if you increase P you will be able to increase f, though the interplay between f and V means that your increases are sub-linear. I don't get your point. Did I write that performance and power is not related?
- jerf 12y ago
  I feel this is one of those rare cases where some feedback about the downvote I just gave is called for. I've downvoted you because someone has asked you to apply the principle of charity to someone's post, and you've chosen to dig in instead. Even if you want to continue this otherwise-interesting discussion, please don't do it this way.
  - m_mueller 12y ago
    Thanks for the explanation. I didn't read GPs post as a call for charity on OP. Looking at this phrase.. > I don't think that anybody could reasonably read top speed being a function of heat as top speed only being a function of heat .. I still don't see it this way. My post wasn't meant to be an attack on OP, rather I wanted to expand some points about chip performance that seemed to me weren't clear. The post I'm quoting however did read uncharitable to me.
- Symmetry 12y ago
  Again, I don't understand how you got your reading. If something is the function of something else that means that the function is singular, not that the function only takes a singular argument. Obviously there are more things that are a function of heat than just clock speed so the op was wrong in that sense. That's the sort of mistake I make in revision all the time so I overlooked it. But even taking it literally it doesn't actually hurt the point the op was trying to make.