Announcement

Collapse
No announcement yet.

Performance-Per-Watt & How The Raspberry Pi 2 + Pi Zero Compare To Old NetBurst CPUs

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Performance-Per-Watt & How The Raspberry Pi 2 + Pi Zero Compare To Old NetBurst CPUs

    Phoronix: Performance-Per-Watt & How The Raspberry Pi 2 + Pi Zero Compare To Old NetBurst CPUs

    After starting to run some Raspberry Pi Zero benchmarks this weekend, I'm back today with more benchmarks. In this article is also an interesting comparison showing the performance of the Raspberry Pi Zero and Raspberry Pi 2 against old "Northwood" Pentium 4 and Celeron processors from the Socket 478 NetBurst days. The many results in this article also include power consumption and performance-per-Watt metrics for this $5 ARM single board computer.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    The manufacturer of the Watts Up Pro warns: "Below 60 watts, current and power factor decrease in accuracy". I'd take the measurements for the RPi's with a grain of salt.

    Comment


    • #3
      Thanks for the write up, I find this stuff very interesting (have a Pi 2 colocated in Amsterdam ).

      As for performance per watt, the numbers are actually a bit surprising. I would have expected Intel to have a much better architecture. The more efficient Intel chips are all built on a 22nm or better process whereas the BCM2835 and BCM2836 are based on TSMC's low power 40nm process. It seems like most improvements are simply through economies of scale (bigger chips being more efficient) and smaller gate sizes, hence the Pi 2 comparing rather well in that regard.

      As for the comparison with the pentium 4, the numbers are inline with what I expect. The Pi 1 with 700 Mhz was roughly equivalent to a Pentium 2 with 300 Mhz. The Pi 2 is 6 times faster so that would be a 1.8 GHz Pentium 2. Netburst suffered from inefficient Pipelines so clock for clock would be somewhat lower than a P2.
      Last edited by bstrobl; 08 December 2015, 11:33 AM.

      Comment


      • #4
        Michael, maybe you could make the benchmark charts slightly more interactive with :hover on the bars to tint it slightly on hover. I don't know.

        Comment


        • #5
          Originally posted by uid313 View Post
          Michael, maybe you could make the benchmark charts slightly more interactive with :hover on the bars to tint it slightly on hover. I don't know.
          Patches are welcome... As I've said before, far from being a graphics artist / UI/UX expert.
          Michael Larabel
          https://www.michaellarabel.com/

          Comment


          • #6
            Cool stuff, thanks Michael, I love seeing geeky benchmarks like this. Really shows how IPC has improved so much over the past decade. I read somewhere that a DEC Alpha EV56 at 500 Mhz - a very high end piece of equipment in 1998, is about on-par with a modern iPad.

            Comment


            • #7
              I find the FFTW results for the zero surprising. 333Mflops/s for a 1GHz ARM v6 core? That's amazing. The Pi2 must have been memory bandwidth limited as it has way better floating point hardware (and four times as many).

              Comment


              • #8
                Originally posted by willmore View Post
                I find the FFTW results for the zero surprising. 333Mflops/s for a 1GHz ARM v6 core? That's amazing. The Pi2 must have been memory bandwidth limited as it has way better floating point hardware (and four times as many).
                I'm guessing it's using VFP rather than NEON for the Pi 2. The Pi 2 should turn much much higher numbers.

                Comment


                • #9

                  Comment


                  • #10
                    This is an amazing look back at IPC and power/performance! thanks Michael!

                    Comment

                    Working...
                    X