Intel memory latency and bandwidth benchmark

[siDDis]

Noen som har testet dette?

https://software.intel.com/en-us/articles/intelr-memory-latency-checker

Fungerer både til Linux og Windows.

 

Veldig interessant resultatet. Her får du vite hvor raskt minnet ditt er og hvor mye overhead et NUMA oppsett vil ha.

 

Resultat server: Dual Xeon L5630 6x4GB PC3-10600R

Intel(R) Memory Latency Checker - v3.1a
Measuring idle latencies (in ns)...
                Numa node
Numa node            0       1
       0          89.4   132.4
       1         131.9    88.4

Measuring Peak Memory Bandwidths for the system
Bandwidths are in MB/sec (1 MB/sec = 1,000,000 Bytes/sec)
Using all the threads from each core if Hyper-threading is enabled
Using traffic with the following read-write ratios
ALL Reads        :      25363.8
3:1 Reads-Writes :      25131.9
2:1 Reads-Writes :      25691.8
1:1 Reads-Writes :      26879.7
Stream-triad like:      27770.1

Measuring Memory Bandwidths between nodes within system
Bandwidths are in MB/sec (1 MB/sec = 1,000,000 Bytes/sec)
Using all the threads from each core if Hyper-threading is enabled
Using Read-only traffic type
                Numa node
Numa node            0       1
       0        14952.5  9618.8
       1         9602.5 14959.1

Measuring Loaded Latencies for the system
Using all the threads from each core if Hyper-threading is enabled
Using Read-only traffic type
Inject  Latency Bandwidth
Delay   (ns)    MB/sec
==========================
 00000  108.46    24999.7
 00002  108.40    25011.0
 00008  108.55    25055.5
 00015  108.61    25074.7
 00050  108.57    25059.3
 00100  103.01    20258.9
 00200   94.35    10744.6
 00300   92.59     7317.7
 00400   91.58     5675.6
 00500   91.04     4687.5
 00700   90.42     3553.8
 01000   90.01     2703.8
 01300   89.81     2248.1
 01700   89.63     1887.0
 02500   89.46     1512.8
 03500   89.36     1285.7
 05000   89.30     1115.4
 09000   89.23      938.9
 20000   89.16      817.5

Measuring cache-to-cache transfer latency (in ns)...
Local Socket L2->L2 HIT  latency        36.1
Local Socket L2->L2 HITM latency        41.1
Remote Socket LLC->LLC HITM latency (data address homed in writer socket)
                        Reader Numa Node
Writer Numa Node     0       1
            0        -   136.5
            1    136.2       -
Remote Socket LLC->LLC HITM latency (data address homed in reader socket)
                        Reader Numa Node
Writer Numa Node     0       1
            0        -   104.8
            1    104.8       -

Resultat workstation: 4670K @ 3.8ghz 2x4GB PC3-10600

ntel(R) Memory Latency Checker - v3.1a
Measuring idle latencies (in ns)...
        Memory node
Socket       0
     0    57.5

Measuring Peak Memory Bandwidths for the system
Bandwidths are in MB/sec (1 MB/sec = 1,000,000 Bytes/sec)
Using all the threads from each core if Hyper-threading is enabled
Using traffic with the following read-write ratios
ALL Reads        :      19785.9
3:1 Reads-Writes :      19022.1
2:1 Reads-Writes :      18747.3
1:1 Reads-Writes :      18501.3
Stream-triad like:      18768.9

Measuring Memory Bandwidths between nodes within system
Bandwidths are in MB/sec (1 MB/sec = 1,000,000 Bytes/sec)
Using all the threads from each core if Hyper-threading is enabled
Using Read-only traffic type
        Memory node
 Socket      0
     0  19679.4

Measuring Loaded Latencies for the system
Using all the threads from each core if Hyper-threading is enabled
Using Read-only traffic type
Inject  Latency Bandwidth
Delay   (ns)    MB/sec
==========================
 00000  174.56    19525.6
 00002  177.05    19538.7
 00008  163.46    19516.8
 00015  158.24    19420.0
 00050  135.28    19040.6
 00100  114.88    18011.3
 00200   86.43    12093.3
 00300   80.91     8866.8
 00400   75.58     7052.5
 00500   71.22     5764.8
 00700   73.08     4488.5
 01000   67.16     3512.8
 01300   68.03     2912.0
 01700   67.68     2492.1
 02500   65.76     1996.4
 03500   71.90     1606.3
 05000   70.80     1378.7
 09000   65.97     1260.1
 20000   66.85     1088.3

Measuring cache-to-cache transfer latency (in ns)...
Local Socket L2->L2 HIT  latency        19.8
Local Socket L2->L2 HITM latency        23.1

Endret 4. august 2016 av siDDis