SiSoftware Sandra是一款可运行在32bit和64bit Windows操作系统上的分析软件，它可以对于系统进行方便、快捷的基准测试，还可以用于查看系统的软件、硬件等信息。SiSoftware Sandra所有的基准测试都针对SMP和SMT进行了优化，最高可支持32/64路平台。我们利用了其中多个性能测试模块对于被测系统的性能进行了快速的测试。
SiSoftware Sandra Pro Business 2009 | ||||
测试对象 | DELL Precision T5500工作站 | HP xw9400工作站 双路AMD Barcelona Opteron 8354 + Quadro FX4800 | HP xw9400工作站 | |
Processor Arithmetic Benchmark | ||||
Dhrystone ALU | 138679MIPS | 62091MIPS | 104576MIPS | |
Dhrystone ALU vs SPEED | 52.13MIPS/MHz | 28.22MIPS/MHz | 43.57MIPS/MHz | |
Whetstone iSSE3 | 121017MFLOPS | 49606MFLOPS | 83301MFLOPS | |
Dhrystone iSSE3 vs SPEED | 45.50MFLOPS/MHz | 22.55MFLOPS/MHz | 34.71MFLOPS/MHz | |
Processor Multi-Media Benchmark | ||||
Multi-Media Int x8 iSSE4.1 | 284.58MPixel/s | 197.95MPixel/s | 326.09MPixel/s | |
Multi-Media Int x8 iSSE4.1 vs SPEED | 106.99kPixels/s/MHz | 89.98kPixels/s/MHz | 135.87kPixels/s/MHz | |
Multi-Media Float x4 iSSE2 | 217.19MPixel/s | 120.24MPixel/s | 201.5MPixel/s | |
Multi-Media Float x4 iSSE2 vs SPEED | 81.65kPixels/s/MHz | 54.65kPixels/s/MHz | 83.96kPixels/s/MHz | |
Multi-Media Double x2 iSSE2 | 119.23MPixel/s | 66.13MPixel/s | 110.73MPixel/s | |
Multi-Media Double x2 iSSE2 vs SPEED | 44.82kPixels/s/MHz | 30.06kPixels/s/MHz | 46.14kPixels/s/MHz | |
Multi-Core Efficiency Benchmark | ||||
Inter-Core Bandwidth | 69.66GB/s | 5.21GB/s | 6.09GB/s | |
Inter-Core Bandwidth vs SPEED | 26.82MB/s/MHz | 2.42MB/s/MHz | 2.60MB/s/MHz | |
Inter-Core Latency | 17ns | 169ns | 160ns | |
Inter-Core Latency vs SPEED | 0.01ns/MHz | 0.08ns/MHz | 0.07ns/MHz | |
Memory Bandwidth Benchmark | ||||
Int Buff'd iSSE2 Memory Bandwidth | 15.08GB/s | 15.08GB/s | 17.85GB/s | |
Int Buff'd iSSE2 Memory Bandwidth vs SPEED | - | 23.18MB/s/MHz | 27.44MB/s/MHz | |
Float Buff'd iSSE2 Memory Bandwidth | 15.08GB/s | 14.97GB/s | 17.82GB/s | |
Float Buff'd iSSE2 Memory Bandwidth vs SPEED | - | 23.02MB/s/MHz | 27.40MB/s/MHz | |
Memory Latency Benchmark | ||||
Memory(Random Access) Latency | 114ns | 133ns | 124ns | |
Memory(Random Access) Latency vs SPEED | - | 0.20ns/MHz | 0.19ns/MHz | |
Speed Factor | 78.00 | 95.40 | 95.90 | |
Internal Data Cache Latency | 4clocks | 3clocks | 3clocks | |
L2 On-board Cache Latency | 10clocks | 16clocks | 16clocks | |
Cache and Memory Benchmark | ||||
Cache/Memory Bandwidth | 128.85GB/s | 69.80GB/s | 113.5GB/s | |
Cache/Memory Bandwidth vs SPEED | 49.60MB/s/MHz | 32.49MB/s/MHz | 48.43MB/s/MHz | |
Speed Factor | 25.70 | 27.10 | 43.70 | |
Internal Data Cache | 424.43GB/s | 260.25GB/s | 439.60GB/s | |
L2 On-board Cache | 395.93GB/s | 143.80GB/s | 283.78GB/s | |
Graphics Processing | ||||
Native Float Shaders | 537.10MPixel/s | 272.04MPixel/s | 274.48MPixel/s | |
Native Float Shaders vs SPEED | 542.11kPixels/s/MHz | 225.95kPixels/s/MHz | 227.97kPixels/s/MHz | |
Graphics Bandwidth | ||||
Internal Memory Bandwidth | 78.03GB/s | 59.35GB/s | 59.56GB/s | |
Data Transfer Bandwidth | 9.25GB/s | 3.08GB/s | 2.90GB/s | |
Internal Memory Bandwidth | 78.03GB/s | 59.35GB/s | 59.56GB/s | |
System to Device Bandwidth | 9.58GB/s | 3.10GB/s | 2.93GB/s | |
System to Device Bandwidth Efficiency | 61.30% | 79.36% | 75.13% | |
Device to System Bandwidth | 6.34GB/s | 2.92GB/s | 2.60GB/s | |
Device to System Bandwidth Efficiency | 40.57% | 74.82% | 66.56% | |
测试结果可以通过两部分来分析。首先，AMD自家的巴塞罗那和伊斯坦布尔进行对比，很明显使用六核心的xw9400工作站性能要好很多，在处理器测试中的各个项目都有明显的提升，内存提升则相对小了一些。而与Intel平台的比较来看，由于至强5500系列处理器规格更为先进，可以支持SSE4，而AMD的Opteron只能支持SSE3，所以两者的性能看起来差距很大。另外一方面，至强E5550在主频上也略高于皓龙Opteron 2431，因此取得优势并不奇怪。
