bit-tech.net

Intel Core i7 Memory Performance

Memory Timings

Lavalys Everest 4.60.1540 Beta Memory Performance

Website: Lavalys

Timings: Everest Memory Read Test

  • 1600MHz 9-9-9-24-1T
  • 1600MHz 8-8-8-21-1T
  • 1600MHz 8-7-8-18-1T
  • 1708MHz OC 9-9-9-24-2T
  • 1600MHz 8-8-8-21-2T
  • 1333MHz 6-6-6-15-1T
  • 1333MHz 7-7-7-18-1T
  • 1066MHz 5-5-5-15-1T
  • 1333MHz 8-8-8-21-1T
  • 1333MHz 9-9-9-24-1T
  • 1066MHz 6-6-6-15-1T
  • 1066MHz 7-7-7-18-1T
  • 1066MHz 9-9-9-24-1T
  • 1066MHz 8-8-8-21-1T
  • 16933
  • 16762
  • 16708
  • 16307
  • 16211
  • 14876
  • 14234
  • 14167
  • 14119
  • 13805
  • 13722
  • 13337
  • 13113
  • 12981
0
2500
5000
7500
10000
12500
15000
17500
MB/s (higher is better)

Timings: Everest Memory Write Test

  • 1708MHz OC 9-9-9-24-2T
  • 1600MHz 8-8-8-21-1T
  • 1600MHz 9-9-9-24-1T
  • 1600MHz 8-7-8-18-1T
  • 1600MHz 8-8-8-21-2T
  • 1333MHz 7-7-7-18-1T
  • 1333MHz 9-9-9-24-1T
  • 1066MHz 9-9-9-24-1T
  • 1066MHz 5-5-5-15-1T
  • 1066MHz 8-8-8-21-1T
  • 1333MHz 6-6-6-15-1T
  • 1066MHz 7-7-7-18-1T
  • 1066MHz 6-6-6-15-1T
  • 1333MHz 8-8-8-21-1T
  • 14663
  • 13842
  • 13790
  • 13754
  • 13728
  • 11969
  • 11968
  • 11967
  • 11965
  • 11950
  • 11912
  • 11909
  • 11902
  • 11888
0
2500
5000
7500
10000
12500
15000
MB/s (higher is better)

Timings: Everest Memory Copy Test

  • 1708MHz OC 9-9-9-24-2T
  • 1600MHz 8-8-8-21-1T
  • 1600MHz 8-7-8-18-1T
  • 1600MHz 8-8-8-21-2T
  • 1600MHz 9-9-9-24-1T
  • 1333MHz 7-7-7-18-1T
  • 1333MHz 6-6-6-15-1T
  • 1333MHz 8-8-8-21-1T
  • 1066MHz 7-7-7-18-1T
  • 1333MHz 9-9-9-24-1T
  • 1066MHz 5-5-5-15-1T
  • 1066MHz 6-6-6-15-1T
  • 1066MHz 8-8-8-21-1T
  • 1066MHz 9-9-9-24-1T
  • 20103
  • 19945
  • 19607
  • 19238
  • 19216
  • 17392
  • 17287
  • 16787
  • 16575
  • 16460
  • 16419
  • 16214
  • 15936
  • 15670
0
5000
10000
15000
20000
MB/s (higher is better)

The results for 1,600MHz memory are a particularly high jump because the CPU memory controller frequency has jumped from 2.66GHz to 3.2GHz - the lowest available at that memory clock. Strangely we found the reduction in memory timings didn't do much for read performance as Everest consistently spat out a result for C9 that was faster than C8 - Corsair's rated timings. Likewise, Everest also didn't seem to like the overclock all that much either as the rated raw read performance is lower than most of the standard 1,600MHz scores - in copy performance, it's up the top though.

The 1T performance setting does seem to make a little bit of difference in read and copy, but write performance is not affected by memory timings - just like it is with Core 2 CPUs which were affected by front side bus clock, the Core i7 memory controller clock directly affects write performance. In the graph above there are three clear steps, from 2.67GHz, to 3.2GHz to finally 3.4GHz with the 1,708MHz memory overclock.

For the most part, despite the few anomalous results, it does correlate to the gentle progression we expect when we either lower the timings or increase the frequency - there is some overlapping but it seems that, for the most part, increasing the frequency does more than decreasing the memory timings, but both benefit.

Timings: Everest Memory Latency Test

  • 1708MHz OC 9-9-9-24-2T
  • 1600MHz 8-7-8-18-1T
  • 1600MHz 8-8-8-21-1T
  • 1600MHz 8-8-8-21-2T
  • 1333MHz 6-6-6-15-1T
  • 1066MHz 5-5-5-15-1T
  • 1600MHz 9-9-9-24-1T
  • 1333MHz 7-7-7-18-1T
  • 1333MHz 8-8-8-21-1T
  • 1066MHz 6-6-6-15-1T
  • 1333MHz 9-9-9-24-1T
  • 1066MHz 7-7-7-18-1T
  • 1066MHz 8-8-8-21-1T
  • 1066MHz 9-9-9-24-1T
  • 27.2
  • 30.7
  • 31.7
  • 32.6
  • 32.8
  • 33.3
  • 33.9
  • 35.5
  • 36.0
  • 36.4
  • 39.4
  • 39.4
  • 41.5
  • 44.3
0
10
20
30
40
Nanoseconds (lower is better)

There's a nanosecond difference between 1T and 2T command rates, and again the gentle progression we'd expect towards lower latencies as clock speeds are increased and timings are dropped, but the frequency of the memory controller does not have as much an effect as optimising the timings, as 1,066MHz at C5 outperforms 1,600MHz at C9.