ScienceMark Membench | ||
处理器 | 双路E5430 | 双路E5430 |
内存技术参数 | 1GB FBD ECC DDR2-667 SDRAM x4 | 2GB FBD ECC DDR2-667 SDRAM x4 |
L1带宽(MB/s) | 55965.41 | 55688.67 |
L2带宽(MB/s) | 16748.31 | 16730.90 |
内存带宽(MB/s) | 4513.57 | 4479.89 |
L1 Cache Latency(ns) | ||
32 Bytes Stride | 1.13 | 1.13 |
L1 Algorithm Bandwidth(MB/s) | ||
Compiler | 25280.32 | 25244.97 |
REP MOVSD | 25451.95 | 25451.67 |
ALU Reg Copy | 12960.86 | 12826.15 |
MMX Reg Copy | 25259.36 | 25348.59 |
SSE PAlign | 51777.17 | 48552.33 |
SSE2 PAlign | 55956.41 | 55688.67 |
L2 Cache Latency(ns) | ||
4 Bytes Stride | 1.13 | 1.13 |
16 Bytes Stride | 1.50 | 1.50 |
64 Bytes Stride | 4.51 | 4.51 |
256 Bytes Stride | 4.51 | 4.51 |
512 Bytes Stride | 4.89 | 4.89 |
L2 Algorithm Bandwidth(MB/s) | ||
Compiler | 11852.65 | 11882.42 |
REP MOVSD | 12551.47 | 12530.11 |
ALU Reg Copy | 8624.66 | 8590.76 |
MMX Reg Copy | 13400.21 | 13403.55 |
SSE PAlign | 16733.54 | 16715.36 |
SSE2 PAlign | 16748.31 | 16730.90 |
Memory Latency(ns) | ||
4 Bytes Stride | 1.13 | 1.13 |
16 Bytes Stride | 4.51 | 4.51 |
64 Bytes Stride | 19.17 | 19.17 |
256 Bytes Stride | 59.40 | 60.90 |
512 Bytes Stride | 68.04 | 68.04 |
Memory Algorithm Bandwidth(MB/s) | ||
Compiler | 3199.52 | 3184.41 |
REP MOVSD | 3234.82 | 3220.63 |
ALU Reg Copy | 2811.05 | 2780.84 |
MMX Reg Copy | 2942.42 | 2966.63 |
MMX Reg 3dNow | --- | --- |
MMX Reg SSE | 4030.85 | 4022.10 |
SSE PAlign | 4345.19 | 4320.64 |
SSE PAlign SSE | 4463.26 | 4443.62 |
SSE2 PAlign | 4346.90 | 4321.07 |
SSE2 PAlign SSE | 4462.34 | 4442.27 |
MMX Block 4kb | 4076.96 | 4062.40 |
MMX Block 16kb | 4502.76 | 4476.12 |
SSE Block 4kb | 4093.63 | 4083.46 |
SSE Block 16kb | 4513.57 | 4476.69 |
在随后的测试中,我们对比了4GB内存和8GB内存两种容量对于系统性能的影响。与上面测试不同的是,我们本次采用的并不是逐步增加插接的内存数量,而是选择了小容量的内存,通过1GB*4和2GB*4两种方式进行测试。而从得到的结果来看,测试数据之间并没有明显的变化,无论是L1、L2、内存带宽,还是内存本身的延迟差别都不大,可以说4GB和8GB所测得的成绩差距微乎其微。这跟我们之前的测试结论截然不同,那么这又是为什么呢?