Phenom(tm) II X6 1055T with DDR2 at 800 MHz

Table of Contents Introduction Runtimes 1 core 2 cores 3 cores 4 cores 5 cores 6 cores Summary Inferences

Introduction

This page registers the performance achieved with the Phenom(tm) II X6 1055T, sporting 3 DDR2 modules of 2GB each at 800 MHz, therefore the memory communication is not fully balanced when it needs to go over the 4GB memory space.

Notes

The use of mpirun is merely as a helper application. The noavxtest* binaries are not running cooperatively.

Keep in mind that these results are not statistically balanced, since they are the result after a single run.

AVX code was commented out, because this CPU does not have the AVX feature :(

Runtimes

These were executed on Ubuntu 15.10 x86_64. Built with the native options:

    g++ -O3 -march=native noavxtest.cpp -o noavxtest
    g++ -O3 -march=native noavxtest64.cpp -o noavxtest64

1 core

32-bit:

   ./noavxtest

• x86:
  • Time taken (ms): 45832.847656

64-bit:

   ./noavxtest64

• x86_64:
  • Time taken (ms): 49884.3320000000022

2 cores

32-bit:

   mpirun -n 2 ./noavxtest

• x86:
  • Time taken (ms): 45845.625
  • Time taken (ms): 45854.257812

64-bit:

   mpirun -n 2 ./noavxtest64

• x86_64:
  • Time taken (ms): 50057.1729999999952
  • Time taken (ms): 50067.948000000004

3 cores

32-bit:

   mpirun -n 3 ./noavxtest

• x86:
  • Time taken (ms): 45918.0625
  • Time taken (ms): 45921.03125
  • Time taken (ms): 45935.003906

64-bit:

   mpirun -n 3 ./noavxtest64

• x86_64:
  • Time taken (ms): 49901.1229999999996
  • Time taken (ms): 49910.5649999999951
  • Time taken (ms): 49916.336000000003

4 cores

32-bit:

   mpirun -n 4 ./noavxtest

• x86:
  • Time taken (ms): 53917.832031
  • Time taken (ms): 53921.621094
  • Time taken (ms): 53919.816406
  • Time taken (ms): 53932.902344

64-bit:

   mpirun -n 4 ./noavxtest64

• x86_64:
  • Time taken (ms): 58558.8410000000003
  • Time taken (ms): 58564.6920000000027
  • Time taken (ms): 58570.1010000000024
  • Time taken (ms): 58585.8429999999935

5 cores

32-bit:

   mpirun -n 5 ./noavxtest

• x86:
  • Time taken (ms): 53916.703125
  • Time taken (ms): 53914.632812
  • Time taken (ms): 53919.515625
  • Time taken (ms): 53919.632812
  • Time taken (ms): 53961.320312

64-bit:

   mpirun -n 5 ./noavxtest64

• x86_64:
  • Time taken (ms): 58566.6169999999984
  • Time taken (ms): 58571.510000000002
  • Time taken (ms): 58570.7880000000005
  • Time taken (ms): 58583.7589999999982
  • Time taken (ms): 58606.153999999995

6 cores

32-bit:

   mpirun -n 6 ./noavxtest

• x86:
  • Time taken (ms): 53921.898438
  • Time taken (ms): 53920.195312
  • Time taken (ms): 53931.277344
  • Time taken (ms): 53915.027344
  • Time taken (ms): 53945.042969
  • Time taken (ms): 53911.898438

64-bit:

   mpirun -n 6 ./noavxtest64

• x86_64:
  • Time taken (ms): 58566.0869999999995
  • Time taken (ms): 58569.5670000000027
  • Time taken (ms): 58591.1010000000024
  • Time taken (ms): 58580.5239999999976
  • Time taken (ms): 58568.7870000000039
  • Time taken (ms): 58564.2750000000015

Summary

Architecture/Mode	1 core	2 cores (std-dev)	3 cores (std-dev)	4 cores (std-dev)	5 cores (std-dev)	6 cores (std-dev)
x86 (ms)	45832.85	45849.94 (6.1)	45924.70 (9.05)	53923.04 (6.75)	53926.36 (19.65)	53924.22 (12.18)
x86_64 (ms)	49884.33	49909.34 (7.68)	50062.56 (7.62)	58569.87 (11.60)	58579.77 (16.07)	58573.39 (10.36)
-	-	-	-	-	-	-
Core frequency (MHz) (cpufreq-aperf)	3276	3248	3220	2772~2800	2772~2800	2772~2800
downscale ratio (c1/cx)	1	1.0086	1.0174	1.17~1.18	1.17~1.18	1.17~1.18
x86	1	1.0004	1.0020	1.1765	1.1766	1.1765
x86_64	1	1.0005	1.0036	1.1741	1.1743	1.1742

Inferences

1. The downscale ratio on the x86/x86_64 calculations are within the expectable downscale range.
2. The speeds indicated by cpufreq-aperf, do somewhat match to the ones listed at cpu-world.com, where for 3 cores it's roughly 3300 MHz and for 4 to 6 is 2800 MHz... although it felt like when using 3 cores, it would be closer to the 3200 MHz mark.
3. When compared to the results achieved with the more recent CPUs, it feels odd that the 64-bit calculations are slower than the 32-bit ones, taking roughly 1.09 times longer to run with 64-bit.
4. For an additional reference, cpubenchmark.net gives an index of 5059 to this CPU.