AMD Opteron™ 4000 Series Platform Quick Reference Guide The world’s lowest power x86 cloud processor1 just got more efficient AMD Opteron™ 4200 Series Processor END USER BENEFITS > Efficiently scales performance/watt based on the innovative Designed for enterprise workloads while still delivering a modular design and aggressive power capabilities performance punch > New power saving features, like TDP Power Cap, put the > A 33% increase in core count packs in plenty of processing customer in control of more aspects of power efficiency than performance into a smaller, more efficient, 8-core design while ever before on AMD Opteron™ processor-based servers 4,5 maintaining very aggressive power/thermal ranges > Up to 8 cores and more processing throughput7 helps your > Power per core that shatters the 5W/core mark, a new record cloud easily scale within the most fluid and spiky processing 1 for an enterprise-class x86 processor environments, easily responding to the elastic needs of clouds > New power management capabilities allow for larger parts of the processor to be almost completely powered off Bringing unparalleled efficiency to your processing, power and when not being used, dramatically reducing idle core power financial budgets consumption by up to 37% over today’s AMD Opteron™ 4100 > The lowest enterprise-class power per core with up to 8 cores KEY FEATURES Series processors and allowing active cores to run at a higher in only 35W of power, shattering the previous record1 >>New>‘Bulldozer’-core>Architecture — drives more core frequency6 density and greater throughput > 32nm design and a smaller die4 drive more efficiency at > AMD Turbo CORE technology takes advantage of additional the processor level to help ensure you are maximizing your power headroom to digitally boost all cores simultaneously by >>AMD>Turbo>CORE>Technology — allows processors performance per watt per square foot up to 300MHz and can boost up to 1GHz when only half of the to independently boost their clock speeds, scaling cores are active2, allowing applications to finish tasks quickly > Virtualized infrastructure deployments can enjoy 33% more frequency up 300MHz-1GHz automatically to respond VMs in the same power and thermal ranges5 thanks to the 2 and return to lower power states to the need for more application performance new AMD-V™ virtualization features, allowing small/medium > Straight-through computing helps ensure that there are not businesses to grow with their business needs on a single bottlenecks or compromises as up to eight threads get their >>C6>Power>State — reduces processor power virtualized platform consumption at active idle by up to 37%3 own dedicated core when workload demands increase with maximum memory channel and I/O speed (across all SKUs/ > New instructions make processing technical software commands more efficient, allowing for more computing per > ™ ™ — price points) helping to ensure that there are no bottlenecks or >AMD>Virtualization >(AMD-V )>Technology>2.0 cycle, which helps drive down processing requirements so that heightens virtualization efficiency with new feature compromises unlike Intel, who purposely throttles down web/cloud applications can more easily process their workload enhancements to the AMD-V™ suite of virtualization capabilities and return to lower power states to optimize data center rack space and help minimize > Easy on your budget through both unprecedented value management tasks Delivering new levels of enterprise scalability for demanding and low power consumption, helping hold down not only cloud applications and SMB/Infrastructure applications acquisition costs but also the long-term total cost of ownership > Scale your cloud workload with up to 8 cores in a low power processor 1 As of April 13, 2011, AMD Opteron™ processor Models 4200 EE has the lowest known power per core of any x86 server processor, at 35W TDP (35W TDP (35W/8 = 4.375/core). Intel’s lowest power per core server processor, L5630, is 40W TDP (40W/4= 10W/core). See http://www.intel. 4 Based on 316mm2 silicon die area of AMD Opteron™ 4200 Series processor versus 346mm2 die size for AMD Opteron 4100 Series processor com/’Assets/PDF/prodbrief/323501.pdf. Previous record held by AMD Opteron™ processor Models 4100 EE at 35W TDP / 6 cores = 5.83 W/core 5 Based on 8-core AMD Opteron™ 4200 Series processors at 35W, 65W and 95W TDP compared to 6-core AMD Opteron™ 4100 Series 2 Based on internal AMD engineering projections of AMD Opteron™ 4200 Series processors with up to 300MHz in P1 boost state and up to 1.4GHz processors at 35W, 65W and 95W TDP when utilizing “1VM per core” loading rule in P0 boost state over base P2 clock frequency when thermal headroom is available 6 Based on internal AMD engineering estimates 3 Based on AMD estimates as of Sept 10, 2011, AMD Opteron™ 4200 Series processors will consume 37% less power in the active idle C1E power 7 Based on internal AMD performance expectations for top bin 8-core AMD Opteron™ 4200 Series standard power processor compared to top bin state with new C6 power gating employed than do AMD Opteron™ 4100 Series processors in the active idle C1E power state. 6-core AMD Opteron™ 4100 Series standard power processors AMD Confidential — NDA Required AMD Opteron™ 4000 Series Platform Quick Reference Guide AMD Opteron™ 4200 Series Processor Product Specifications AMD Opteron™ 4200 Series Processor Product Specifications Core AMD Turbo CORE Max Memory Model Number Count Core Frequency Max Frequency Bandwidth CMOS Tech L2 Cache L3 Cache ACP Cache Sizes Total Cache: 16MB (8 core), 14MB (6 core) L1 Cache: 16KB/core + 64 KB instruction/module 4226 6 2.7 GHz 3.1 GHz 51.2 GB/s 32 nm SOI 3 x 2 MB 8 MB TBD L2 Cache: 1MB (per core) L3 Cache: 8MB (per socket) 4228 HE 6 2.8 GHz 3.6 GHz 51.2 GB/s 32 nm SOI 3 x 2 MB 8 MB TBD Process Technology 32-nanometer SOI (silicon-on-insulator) technology ™ 4234 6 3.1 GHz 3.5 GHz 51.2 GB/s 32 nm SOI 3 x 2 MB 8 MB TBD HyperTransport Two x16 links at up to 6.4GT/s per link Technology Links 4238 6 3.3 GHz 3.7 GHz 51.2 GB/s 32 nm SOI 3 x 2 MB 8 MB TBD Memory Integrated DDR3 memory controller — Up to 51.2 GB/s 4256 EE 8 1.6 GHz 2.8 GHz 51.2 GB/s 32 nm SOI 4 x 2 MB 8 MB TBD memory bandwidth per CPU for Socket C32 4274 HE 8 2.5 GHz 3.5 GHz 51.2 GB/s 32 nm SOI 4 x 2 MB 8 MB TBD Number of Channels/ Dual channel support for U/RDDR3 up to DDR3-1600 Types of Memory and ULV (1.25V) RDDR3 up to DDR3-1333 4280 8 2.8 GHz 3.5 GHz 51.2 GB/s 32 nm SOI 4 x 2 MB 8 MB TBD Die Size 316 mm2 4284 8 3.0 GHz 3.7 GHz 51.2 GB/s 32 nm SOI 4 x 2 MB 8 MB TBD Packaging Socket C32 — 1207 Organic Land Grid Array (OLGA) AMD SP5100 Southbridge Product AMD SR5650, SR5670, SR5690 I/O Hub Product Specifications Specifications Number of Model PCI ® Max TDP/ Process USB Ports 12 USB 2.0 + 2 USB 1.1 Number Processor Interface Express® PCIe Ports/ Virtualization Error Detection/Isolation Idle (w/c1e) Technology Package engines PCI Bus Support PCI rev 2.3 HyperTransport error handling, Serial ATA AHCI 1.1 SATA 3.0Gb/s with HyperTransport™ 3.0 22 lanes/ PCIe® Advanced Error 29 x 29mm SW RAID Support SR5650 v2.0 AMD-Vi (IOMMU 1.26) 12.6W/ 5.4W TSMC 65nm technology (5.2GT/s) 8 engines Reporting, PCIe® end-to-end FCBGA SATA Ports 6 (can be independently Cycle Redundancy Check disabled) HyperTransport error handling, Max TDP/Idle 4W/1W HyperTransport™ 3.0 30 lanes/ PCIe® Advanced Error 29 x 29mm SR5670 v2.0 AMD-Vi (IOMMU 1.26) 15.4W/ 5.75W TSMC 65nm technology (5.2GT/s) 9 engines Reporting, PCIe® end-to-end FCBGA Process Technology TSMC .13um Cycle Redundancy Check Package 528 ball FCBGA, 21x21mm, 0.8mm pitch HyperTransport error handling, HyperTransport™ 3.0 42 lanes/ PCIe® Advanced Error 29 x 29mm SR5690 v2.0 AMD-Vi (IOMMU 1.26) 18W/6.15W TSMC 65nm technology (5.2GT/s) 11 engines Reporting, PCIe® end-to-end FCBGA Cycle Redundancy Check ©2011 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD Opteron, AMD Virtualization, AMD-V and combinations thereof are trademarks of Advanced Micro Devices, Inc. HyperTransport is a licensed trademark of the HyperTransport Technology Consortium. Other names are for reference only and may be the trademarks of their respective owners. PID 50367A AMD Confidential — NDA Required.