为了体现出超线程对系统性能的影响，我们特地关闭了超线程又作了一遍测试。
SiSoftware Sandra Pro Business 2009 | |||
测试对象 | Dawning I620r-G Nehalem-EP 双路Intel Gainestown Xeon E5540 2.53GHz | Dawning I620r-G | |
Processor Arithmetic Benchmark 处理器架构测试 | |||
| Dhrystone ALU | 129014MIPS | 130767MIPS | |
| Dhrystone ALU vs SPEED | 51.05MIPS/MHz | 51.75MIPS/MHz | |
Whetstone iSSE3 | 111000MFLOPS | 68158MFLOPS | |
| Dhrystone iSSE3 vs SPEED | 43.93MFLOPS/MHz | 26.97MFLOPS/MHz | |
Processor Multi-Media Benchmark 处理器多媒体测试 | |||
Multi-Media Int x16 iSSE4.1 | 269.08MPixel/s | 228.02MPixel/s | |
Multi-Media Int x16 iSSE4.1 vs SPEED | 106.48kPixels/s/MHz | 90.23kPixels/s/MHz | |
Multi-Media Float x8 iSSE2 | 206.19MPixel/s | 172.03MPixel/s | |
Multi-Media Float x8 iSSE2 vs SPEED | 81.60kPixels/s/MHz | 68.08kPixels/s/MHz | |
Multi-Media Double x4 iSSE2 | 113.93MPixel/s | 89.72MPixel/s | |
Multi-Media Double x4 iSSE2 vs SPEED | 45.09kPixels/s/MHz | 35.50kPixels/s/MHz | |
Multi-Core Efficiency Benchmark | |||
Inter-Core Bandwidth | 63.30GB/s | 25.88GB/s | |
Inter-Core Bandwidth vs SPEED | 25.65MB/s/MHz | 10.49MB/s/MHz | |
Inter-Core Latency (越小越好) | 22ns | 58ns | |
Inter-Core Latency? vs SPEED (越小越好) | 0.01ns/MHz | 0.02ns/MHz | |
Memory Bandwidth Benchmark 内存带宽测试 | |||
Int Buff'd iSSE2 Memory Bandwidth | 9.02GB/s | 32.59GB/s | |
Float Buff'd iSSE2 Memory Bandwidth | 8.90GB/s | 32.56GB/s | |
Memory Latency Benchmark 内存延迟测试 | |||
Memory(Random Access) Latency (越小越好) | 96ns | 92ns | |
Speed Factor (越小越好) | 59.40 | 57.90 | |
Internal Data Cache | 4clocks | 4clocks | |
L2 On-board Cache | 10clocks | 10clocks | |
L3 On-board Cache | 52clocks | 51clocks | |
Cache and Memory Benchmark 缓存及内存测试 | |||
Cache/Memory Bandwidth | 122.06GB/s | 120.64GB/s | |
Cache/Memory Bandwidth vs SPEED | 49.46MB/s/MHz | 48.89MB/s/MHz | |
Speed Factor (越小越好) | 22.80 | 23.10 | |
| Internal Data Cache | 398.74GB/s | 401.21GB/s | |
| L2 On-board Cache | 368.03GB/s | 362.61GB/s | |
.NET Arithmetic Benchmark .NET架构测试 | |||
Dhrystone .NET | 29299MIPS | 28774MIPS | |
Dhrystone .NET vs SPEED | 11.59MIPS/MHz | 11.39MIPS/MHz | |
Whetstone .NET | 69736MFLOPS | 44516MFLOPS | |
Whetstone .NET vs SPEED | 27.60MFLOPS/MHz | 17.62MFLOPS/MHz | |
.NET Multi-Media Benchmark .NET多媒体测试 | |||
Multi-Media Int x1 .NET | 53.25MPixel/s | 46.38MPixel/s | |
Multi-Media Int x1 .NET vs SPEED | 21.07kPixels/s/MHz | 18.35kPixels/s/MHz | |
Multi-Media Float x1 .NET | 23.09MPixel/s | 13.30MPixel/s | |
Multi-Media Float x1 .NET vs SPEED | 9.14kPixels/s/MHz | 5.26kPixels/s/MHz | |
Multi-Media Double x1 .NET | 45.02MPixel/s | 24.73MPixel/s | |
Multi-Media Double x1 .NET vs SPEED | 17.81kPixels/s/MHz | 9.79kPixels/s/MHz | |
SiSoftware Sandra对比，用蓝色标出了性能特出的项目
只有极少数的项目中，关闭超线程获得了更好的测试成绩。Nehalem-EP的超线程比起Pentium 4时代有了不少的改进，你不应该将其关闭。
