i7 950 with DDR3 1333 MHz

Table of Contents Introduction Runtimes 1 core 2 cores 4 cores 8 cores Summary Inferences

Introduction

This page registers the performance achieved with the Intel i7-950, sporting 6 DDR3 modules of 4 GB each at 1333 MHz. The machine used is part of blueCAPE's IT pool.

Notes

HyperThreading was turned on in the BIOS settings.

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.

Runtimes

These were executed on CentOS 6.6 x86_64, using a custom build of GCC 4.8.4. 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.sh

• x86:
  • Time taken (ms): 29130.0

64-bit:

   ./noavxtest64.sh

• x86_64:
  • Time taken (ms): 28270.0

2 cores

32-bit:

   mpirun -n 2 ./noavxtest.sh

• x86:
  • Time taken (ms): 30140.0
  • Time taken (ms): 30320.0

64-bit:

   mpirun -n 2 ./noavxtest64.sh

• x86_64:
  • Time taken (ms): 29250.0
  • Time taken (ms): 29480.0

4 cores

32-bit:

   mpirun -n 4 ./noavxtest.sh

• x86:
  • Time taken (ms): 30240.0
  • Time taken (ms): 30310.0
  • Time taken (ms): 30360.0
  • Time taken (ms): 30360.0

64-bit:

   mpirun -n 4 ./noavxtest64.sh

• x86_64:
  • Time taken (ms): 29250.0
  • Time taken (ms): 29330.0
  • Time taken (ms): 29370.0
  • Time taken (ms): 29570.0

8 cores

32-bit:

   mpirun -n 8 ./noavxtest.sh

• x86:
  • Time taken (ms): 56810.0
  • Time taken (ms): 56590.0
  • Time taken (ms): 56650.0
  • Time taken (ms): 56810.0
  • Time taken (ms): 56600.0
  • Time taken (ms): 56720.0
  • Time taken (ms): 56660.0
  • Time taken (ms): 56720.0

64-bit:

   mpirun -n 8 ./noavxtest64.sh

• x86_64:
  • Time taken (ms): 56560.0
  • Time taken (ms): 56730.0
  • Time taken (ms): 56720.0
  • Time taken (ms): 56610.0
  • Time taken (ms): 56560.0
  • Time taken (ms): 56480.0
  • Time taken (ms): 56530.0
  • Time taken (ms): 56480.0

Summary

Architecture/Mode	1 core	2 cores (std-dev)	4 cores (std-dev)	8 cores (std-dev)
x86 (ms)	29130.0	30230.0 (90.0)	30317.5 (49.1807889323)	56695.0 (79.8435971134)
x86_64 (ms)	28270.0	29365.0 (115.0)	29380.0 (117.898261226)	56583.75 (90.9584383111)
-	-	-	-	-
Core frequency (MHz) (cpufreq-aperf)	3305	3182	3182	3182
downscale ratio (c1/cx)	1	1.0386	1.0386	1.0386
x86	1	1.03776175764	1.04076553381	1.94627531754
x86_64	1	1.0387336399	1.03926423771	2.00154757694

Inferences

1. The downscale ratio on the x86/x86_64 calculations are all exactly on the mark for the downscale range, with the exception when relying on the HyperThreading feature.
   • Note: the frequencies were revised with cpufreq-aperf, to compare with the ones at cpu-world.com.
2. On this machine that has HyperThreading turned on, it was possible to assess that this feature has very little impact on improved performance.
   • On 32-bit, it meant roughly 7% more processing power: 2*30317.5/56695.0 - 1.0 = 0.0695
   • On 64-bit, it meant roughly 4% more processing power: 2*29380.0/56583.75 - 1.0 = 0.0385
3. For an additional reference, cpubenchmark.net gives an index of 5646 to this CPU, but keep in mind that this index accounts for HyperThreading. The estimated index without HyperThreading is at a 1.3 ratio, namely 4343.
4. An interesting result is that the 64-bit timings are only slightly faster than the 32-bit ones.