Competitive Comparison: Energy Efficient Performance
Dual-Core Intel® Xeon® Processor-based Platforms vs. AMD Opteron

Get higher performance and capability along with lower energy consumption, reducing costs and improving density from every server in your infrastructure based on the new Dual-Core Xeon® processor.

The new Dual-Core Xeon processor-based systems deliver leading performance and better performance per watt. Key platform feature improvements include:

• Lower power 64-bit dual-core processors based on Intel® CoreTM microarchitechture to improve performance per watt
• Flexibility to do more with every system by leveraging server visualization now hardware assisted with Intel® Virtualization Technology.
• New dual independent point-to-point bus and faster front side bus speed driving up to 3x bandwidth improvements versus prior Intel® Xeon® processor-based platforms.
• Fully buffered DDR2 DIMM memory (FB-DIMM) for more throughput higher capacity and improved reliability.
• Platform supports future qual-core processors for continued performance gains and investment protection.

Leading Energy Efficiency
Why does system power matter? IT managers are faced with the challenge of a growing need for higher compute density and constant pressure to lower overall spending. The cost of electricity is the second highest data center costs, only second to labor. Lowering the power (watts) of each system helps increase compute density within a rack or a fixed footprint and also helps reduce operational cost associated with powering both the servers directly and the associated cooling.

How is Intel driving new levels of power efficiency? Intel has a comprehensive approach to improving power efficiency resulting in better optimized data centers and lower overall cost. First, Intel has already incorporated strained silicon into its 65 nm processor technology to improve transistor performance and decrease power. Intel is shifting to the Intel® CoreTM microarchitecture, further reducing the power consumed by the processor. Three power levels of Woodcrest are being offered: 80, 65, and 40 watts. Additionally, the Intel® power toolkit helps IT managers fit more servers within each rack. Finally, demand-based switching with Enhanced Intel SpeedStep® technology helps lower utility costs.

Breakthrough Performance
Intel® XeonTM processor-based platforms vs. AMD Opteron. For more information on published performance results, please visit: www.intel.com/performance/server/xeon.
(On the graphs below, the higher numbers equate better performance.)

Comparing the Architectures
Key architecture advantages:

• Dual Independent Point-to-Point Bus: Up to 3x system bandwidth increase vs. previous generation, resulting in higher performance.
• Intel® Smart Cache: 4M shared cache fully accessible by each core lowering memory latency and increasing performance.
• Fully Buffered DIMMs: Increases memory capacity and throughput, simultaneously read and write helps reduce latency at peak load, and allows for maximum capacity at full bandwidth.

Advantages:

• 2X cache and fully accessible by both cores (fewer cache misses)
• Wider bus solely for data (higher data bandwidth)
• Full-speed memory at full capacity

Dual-Core Intel® XeonTM processor (5100 Series)

Limitations:

• Smaller cache and only one-half accessible to each core
• Sacrifice capability for bandwidth
• Requires both processors for max memory BW and capacity
• Shared link for Data and Address (lower data bandwidth)

Opteron (Socket 940)

Uni-direction, encoded packet (shared address and data bus)

16 bits (each direction)
8 GB/s
2x1M Discrete/Only	1M Accessible per Core
DDR-400	DDR-266
12.8 GB/s	8.5 GB/s
32 GB	64 GB

Value of Fully Buffered DDR2
Why FB-DIMM? Fully Buffered DIMM (FB-DIMM) is the next evolution in memory technology for server and workstations that value both capacity and bandwidth. With older memory technologies, as the signaling rates are increased, the number of DIMMs supported per channel decreases. Therefore, there was a trade-off between overall memory capacity and bandwidth. FB-DIMM offers the no-compromise solution by providing maximum capacity at the highest bandwidth.

What is FB-DIMM? Previous memory technologies used a shard parallel interface with all DIMMs on the same bus. FB-DIMM is based on a high-speed point-to-point interface that uses a buffer, referred to as the AMB (Advanced Memory Buffer), to decouple the memory interface from the DRAM components. Decoupling allows the memory bus to run at full-speed no matter how many DIMMs are installed, and allows simultaneous reads and writes while still using industry standard commodity DDR2 DRAMs to keep down costs.

Reliability Built In: FB-DIMM technology offers better RAS (reliability, availability, serviceability) by complementing Intel’s existing enhanced data ECC protection (error-correcting code) with enhanced CRC (Cyclic Redundancy Checking). Enhanced CRC checks the transfer of all address, commands, and data and automatically retries when an error is detected, allowing for uninterrupted operation in case of transient errors. Finally, there is an AMB pass-through mode whereby if part of the AMB falls, it switches into repeater mode. That DIMM is effectively turned off, but the memory requests are passed in both directions.

For more information on the new Dual-Core Intel Xeon processor, visit www.intel.com/xeon.