英特尔至强Sandy Bridge首测(图)(1)(17)
本次测试的至强E3-1275是E3系列的高端产品，具备了4核心8线程，性能非常强悍。下面，我们将关闭E3-1275的超线程功能，观察在这个状态下它的性能表现，以便总结超线程技术对于Sandy Bridge处理器的影响。
| SiSoftware Sandra Lite 2011 | ||
| 产品名称 | 至强E3-1275开启） | 至强E3-1275关闭） |
| 平台类型 | 单路Sandy Bridge | 单路Sandy Bridge |
| Processor Arithmetic Benchmark 处理器算术运算测试 |
||
|---|---|---|
| Dhrystone ALU | 136.74GIPS | 106.77GIPS |
| Dhrystone ALU vs SPEED | 35.98MIPS/MHz | 28.86MIPS/MHz |
| Whetstone iSSE3 | 83.43 GFLOPS | 49.57 GFLOPS |
| Dhrystone iSSE3 vs SPEED | 21.96 MFLOPS/MHz | 13.40MFLOPS/MHz |
| Processor Multi-Media Benchmark 处理器多媒体测试 |
||
| Multi-Media Int x16 iSSE4.1 | 201.26MPixel/s | 163.79MPixel/s |
| Multi-Media Int x16 iSSE4.1 vs SPEED | 78.26 kPixels/s/MHz | 63.69kPixels/s/MHz |
| Multi-Media Float x8 iSSE2 | 153MPixel/s | 111.37 MPixel/s |
| Multi-Media Float x8 iSSE2 vs SPEED | 59.48 kPixels/s/MHz | 43.3kPixels/s/MHz |
| Multi-Media Double x4 iSSE2 | 83.54MPixel/s | 60.16MPixel/s |
| Multi-Media Double x4 iSSE2 vs SPEED | 32.49 kPixels/s/MHz | 23.4kPixels/s/MHz |
| Multi-Core Efficiency Benchmark 处理器效能测试 |
||
| Inter-Core Bandwidth | 16 GB/s | 8.84GB/s |
| Inter-Core Bandwidth vs SPEED | 4.29 MB/s/MHz | 2.45MB/s/MHz |
| Inter-Core Latency(越小越好) | 42.4ns | 40.2ns |
| Inter-Core Latency vs SPEED(越小越好) | 0.01ns/MHz | 0.11ns/MHz |
| .NET Arithmetic Benchmark .NET算术运算测试 |
||
| Dhrystone .NET | 18GIPS | 15.81GIPS |
| Dhrystone .NET vs SPEED | 5MIPS/MHz | 4.16MIPS/MHz |
| Whetstone .NET | 50.1 GFLOPS | 29.34GFLOPS |
| Whetstone .NET vs SPEED | 13.92MFLOPS/MHz | 7.72MFLOPS/MHz |
| .NET Multi-Media Benchmark .NET多媒体测试 |
||
| Multi-Media Int x1 .NET | 37.72 MPixel/s | 29.17MPixel/s |
| Multi-Media Int x1 .NET vs SPEED | 9.93MPixel/s | 7.68kPixels/s/MHz |
| Multi-Media Float x1 .NET | 13.7MPixel/s | 8.16MPixel/s |
| Multi-Media Float x1 .NET vs SPEED | 3.61kPixels/s/MHz | 2.15kPixels/s/MHz |
| Multi-Media Double x1 .NET | 27.22 MPixel/s | 16.24MPixel/s |
| Multi-Media Double x1 .NET vs SPEED | 7.16 kPixels/s/MHz | 4.27kPixels/s/MHz |
对比开启与关闭超线程的测试数据我们发现，处理器计算性能方面，开启超线程之后会有30%-50%左右的性能提升，多媒体方面的性能提升为30%左右。而在.NET测试中，这个数值被缩小到了15%-30%，效能测试的时候两者的差距大约为40%。总体而言，在开启超线程之后，Sandy Bridge至强处理器在运算性能上会有30%左右的提升，这个数值与Nehalem与Westmere的成绩是差不多的，事实上我们也没发现Sandy Bridge在超线程方面有什么特别大的改动。
评论暂时关闭