Memory Performance:
We've had to opt to use the latest version of Sisoft Sandra XII SP1 for testing, because the current version of Everest doesn't support Phenom properly. Instead of loading all the cores, Everest 4.2 only loads a single core (instead of all cores) when using Phenom and thus reports lower-than-expected performance.
-
Phenom 9900 (4x2.6GHz, 2.0GHz HTT)
-
Phenom 9700 (4x2.4GHz, 2.0GHz HTT)
-
Phenom 9600 ES (4x2.3GHz, 2.0GHz HTT)
-
Phenom 9500 ES (4x2.2GHz, 2.0GHz HTT)
-
Core 2 Extreme QX9650 (4x3.00GHz, 1333MHz FSB)
-
Core 2 Extreme QX6850 (4x3.00GHz, 1333MHz FSB)
-
Athlon 64 X2 6000+ (2x3.00GHz, 1.0GHz HTT)
-
Core 2 Duo E6850 (2x3.00GHz, 1333MHz FSB)
-
Core 2 Duo E6750 (2x2.67GHz, 1333MHz FSB)
-
Core 2 Quad Q6700 (4x2.67GHz, 1066MHz FSB)
-
Core 2 Quad Q6600 (4x2.40GHz, 1066MHz FSB)
-
11211.0
-
11200.0
-
11142.0
-
11062.0
-
7589.0
-
7003.5
-
6986.0
-
6975.5
-
6853.5
-
5643.0
-
5626.0
0
2500
5000
7500
10000
12500
MB/s
As expected, the Phenoms obliterate everything else in dual channel memory mode, however the usual performance increase we see from the older Athlon 64 X2 isn't there according to Sandra. One thing that's worth mentioning here is that there is a difference between ganged and unganged memory modes -- we have investigated this later on in the review.
-
Athlon 64 X2 6000+ (2x3.00GHz, 1.0GHz HTT)
-
Core 2 Extreme QX6850 (4x3.00GHz, 1333MHz FSB)
-
Core 2 Extreme QX9650 (4x3.00GHz, 1333MHz FSB)
-
Core 2 Duo E6750 (2x2.67GHz, 1333MHz FSB)
-
Core 2 Duo E6850 (2x3.00GHz, 1333MHz FSB)
-
Core 2 Quad Q6700 (4x2.67GHz, 1066MHz FSB)
-
Core 2 Quad Q6600 (4x2.40GHz, 1066MHz FSB)
-
Phenom 9900 (4x2.6GHz, 2.0GHz HTT)
-
Phenom 9700 (4x2.4GHz, 2.0GHz HTT)
-
Phenom 9600 ES (4x2.3GHz, 2.0GHz HTT)
-
Phenom 9500 ES (4x2.2GHz, 2.0GHz HTT)
-
67.0
-
68.0
-
70.0
-
71.0
-
71.0
-
81.0
-
82.0
-
87.0
-
87.0
-
87.0
-
88.0
0
10
20
30
40
50
60
70
80
90
nanoseconds (lower is better)
Interestingly, Phenom's memory latency results were rather unexpected, as we're normally used to the integrated memory controller reducing memory latency. This is probably down to the L3 cache being an extra high-speed data storage pool, meaning that it takes longer for data to actually reach the memory.
Want to comment? Please log in.