Original 45nm Intel® Core™ Microarchitecture

Original 45nm Intel® Core™2 Processor Performance

Intel Xeon® processors

New servers, workstations, and high performance computing (HPC) systems are built with new quad-core Intel Xeon processors 5400 series that are based on the 45nm PenrynΔ core technology. Intel's 45nm technology packs 820 million transistors into the Intel Xeon processor 5400 series. The chip is smaller than the previous-generation Intel Xeon processor 5300 series (214mm² vs. 286mm²), which had 582 million transistors. More transistors on new 45nm technology means more capability, performance, and energy efficiency.

Increased performance

The Intel Xeon 5400 series, based on technology from the Penryn family of processors, featuring a larger 12-MB L2 cache, delivered a strong performance gain to the already stable and shipping server platform based on the Intel 5000 series chipset. A drop-in into the existing platform, the 5400 series, added up to a 21 percent performance increase over the previous-generation, quad-core Intel Xeon processor 5300 series for mainstream server benchmarks at the highest frequency level (comparing Xeon X5460 at 3.16GHz to Xeon X5365 at 3GHz). Figure 13 shows the comparison on a range of server benchmarks. Figure 14 shows performance comparisons to the previous generation at the same clock frequency (3GHz) achieving up to a 19 percent performance increase and highlighting the benefits of the improvements in the Penryn family of processors.

Figure 13: Comparison of server benchmarks with those of the previous generation.

Figure 14: Comparison of server benchmarks to those of the previous generation at the same frequency on the same platform

The Penryn family of processors’ architecture presents several compiler optimization opportunities. These opportunities include tuning for the new ISA, a larger cache, and hardware pre-fetching. The overall gain for the SPEC CPU2006 benchmark suite is shown in Figure 14 as 15–16 percent on the ‘rate’ benchmark. But gains across the individual components could be as high as 57 percent. Figure 15 shows some of the highlights across the Integer and Floating-point component workloads. All the results shown are on the peak result metric (SPECint_rate2006 and SPECfp_rate2006).

Figure 15: Gains on specific SPEC CPU2006 ‘rate’ components at same frequency

To leverage the full potential of the 45nm microarchitecture, a new platform targeting the HPC market segment was launched with the new Intel 5400 chipset that could run at the faster FSB speed of 1600MHz. The additional bandwidth delivered by the platform is critical for the HPC segment. Figures 16a and b show results on key HPC workloads on segments such as manufacturing, financial services, energy, weather and climate modeling, electronic design automation (EDA), and life sciences. It is important to note that the gains achieved by these applications were due to a combination of microarchitectural enhancements and the increased platform bandwidth.

Figure 16: (a) HPC benchmarks .(b) HPC benchmarks

Improved energy efficiency

Servers based on the quad-core Intel Xeon processor 5400 series also maximize data centers performance and density through improved energy efficiency. As shown in Figure 17 , these processors can deliver up to 38 percent more performance per watt in the same platforms and at the same system power level. The platform power in this chart is based on measured average power value at the steady-state window of the benchmark run. For this comparison we used the most energy-efficient mainstream processor SKU from each of the processor families. In this case, that would mean the Intel Xeon E5450 running at 3GHz compared to the Intel Xeon E5345 running at 2.33GHz, both at an 80W TDP rating. The 45nm, Hi-k-based processor also lowers the idle power significantly. Results on the new industry standard SPECpower*s_ssj2008 benchmark, which is the first comprehensive benchmark to measure energy-efficient performance across a load-line including idle power, highlights the energy-efficient performance of the Penryn family of processors. Table 2 shows the top-10 list for this benchmark, all occupied by platforms based on the Penryn family of processors’ architecture.

Figure 17: Energy efficiency-SPECjbb2005 Benchmark

click image for larger view

Enhanced virtualization

Virtualization partitions or compartmentalizes a single computer so that it can run separate operating systems and software. These partitions can better leverage multi-core processing power, increase efficiency, and cut costs, by letting a single machine act as many virtual ‘mini’ computers. Consolidating applications onto fewer systems not only results in better multi-core utilization but improves performance density. In the Penryn family of processors, virtual machine transition (entry exit) times show an improvement of between 25 percent and 75 percent. Based on virtualization benchmark results on different VMMs, the Penryn family of processors provide up to a 20 percent performance gain when compared to previous-generation platforms. Figure 18 shows comparisons of different benchmarks such as VMmark and vConsolidate running various VMMs such as VMware ESX Server, Parallels Virtuozzo, and Virtual Iron 4.0.

Figure 18: Virtualization performance

Back to Top