AMD Opteron Processors In the Cloud

AMD Opteron Processors In the Cloud Pat Patla Vice President Product Marketing AMD

DID YOU KNOW? By 2020, every byte of data will pass through the cloud *Source IDC 2 AMD Opteron In The Cloud October, 2011

Cloud computing is managing the explosion of data 3 AMD Opteron In The Cloud October, 2011

Research highlights* 37% of organizations worldwide are currently operating in the cloud 63% of global cloud customers estimate they store more than $250,000 worth of data in the cloud 75% of businesses using cloud had the skills in-house to implement the technology versus only 39% of those who are currently investigating deployments 92% of cloud customers stated that the infrastructure was an important part of the decision to implement cloud computing *Based on AMD Cloud Computing research, 2011

the Cloud is here The five-year growth outlook for cloud shows an annual growth rate of 26%, which is five times faster than the rate of traditional on-premise IT offerings** Nearly a quarter of public sector workloads reside in the cloud* Federal CIO requires every Dept. to identify 3 applications that can be in the cloud and to move at least one by next year*** * Based on findings from AMD sponsored cloud computing research, 2011 ** http://www.edlconsulting.com/newsdetail.php?id=455&headline=idc:_cloud_computing_market_to_reach_$44b_in_revenue_by_2013 *** http://www.cio.gov/documents/25-point-implementation-plan-to-reform-federal%20it.pdf

U.S. Federal Government and the Cloud Federal Government s Cloud Plans: $20 Billion Shift - Cloud First policy aims to save money. -3.20-2011 The cloud computing model can significantly help agencies grappling with the need to provide highly reliable, innovative services quickly despite resource constraints. NIST releases 'Bible of cloud implementation - 11-4-2011 U.S. Federal Cloud Computing Strategy, February, 2011

end-user Needs Are driving data center changes Affordability Accessibility, Connectivity Personalization Mobility Multi-functionality Privacy & security Natural interface, Ease of use Realism User involvement in experience creation

end-user Needs Are driving data center changes Server market moving to Utility Computing Model Time-to-market expectations accelerating for innovation Visionaries predicting The Ambient Computing Systems Age Explosion of digital data changes how we process information Large-scale Mega-datacenters are highly effective Throughput Computers IN DISRUPTIVE TIMES, THE KEY TO FORECASTING FUTURE MARKET TRENDS IS UNDERSTANDING END-USER WANTS Shift from highperformance to low power, good enough computing Collision course between Social networking and Privacy/Security Shift from Wintel Clients to Cloud Clients Programming models shifting to much higher levels of abstraction

The Data Center Continues to Evolve IT has continued to change over time, as different forces impact Centralization System Cost Mainframes Minicomputers The PC Server Virtualized x86 The Cloud The Future Flexibility is needed to stay competitive Centralization is a pendulum Customers demand a better experience Flexibility User Experience Cost continues to drive down 1970 s 1980 s 1990 s 2000 s 2010 s 2020 s

You manage You manage You manage Cloud Computing Technology Stack (On-Premises) Infrastructure (as a Service) Platform (as a Service) Software (as a Service) Applications Applications Applications Applications Data Data Data Data Runtime Middleware O/S Virtualization Servers Storage Networking Runtime Middleware O/S Virtualization Servers Storage Networking Managed by service provider Runtime Middleware O/S Virtualization Servers Storage Networking Managed by service provider Runtime Middleware O/S Virtualization Servers Storage Networking Managed by service provider

How Do Cloud Workloads Scale? Cloud workloads are very different than traditional data center loads Cloud work is spiky in nature, you must be able to account for both the peaks (with more cores) and the valleys (with more power efficiency) Heavy Computation Cores Matter Power Efficiency Matters Low computation

Dense Form Factors Drive Greater Power Density Servers have continued to increase power efficiency over time But along with more efficiency has come greater density This has had an impact on total power per square foot in the data center Total Rack Power (watts) Total Server Power (watts) Source: Internal AMD estimates based on historical platform power and density within a standard 42U rack

Technology Transitions Over Time Scale Out Age Standardization Consolidation Age Virtualization Cloud Age Business Agility Accelerated Age Server fusion AMD s History of Innovation

Cloud Platform requirements Transaction Processing Cores, memory, and I/O capabilities to handle large amounts of users and data Energy Efficiency Low power consuming CPUs and platforms for scale out cloud server farms Performance and Scalability CPUs with predictable performance to handle peaks in traffic Security Hardware assisted security to keep out intruders and protect sensitive data Virtualization Hardware assisted virtualization for reduced overhead and increased isolation of VMs

AMD Bulldozer Technology the Firsts that matter First major change to AMD Opteron processor core in 8 years Innovative architecture like nothing else in the market First 16 core x86 CPU Flex FP the world s only dynamic, flexible floating point complex Revamped integrated memory controller AMD Turbo CORE technology for up to 1GHz of additional performance Driving huge benefits to customers Designed to deliver outstanding value Highly power efficient Modular architecture for future innovation and time to market advantages * = Upcoming Bulldozer architecture-based AMD Opteron processors experience all core boost of up to 500 MHz (P2 base to P1 boost state) and up to 1.3 GHz max turbo boost (half or fewer cores boost from P2 to P0 boost state)..

Introducing the new AMD Opteron 6200 and 4200 Series processors AMD Opteron 6200 Series Processor ( Interlagos ) AMD Opteron 4200 Series Processor ( Valencia ) The world s first x86 16- core processor The world s lowest x86 power-per-core 1 Scalability Up to 4 sockets with up to 16 cores Up to 2 sockets with up to 8 cores Memory 4 memory channels up to 1600 MHz memory 2 memory channels up to 1600 MHz memory Frequency Cache I/O Power Up to 3.3 GHz base frequency & up to 3.6 GHz frequency using AMD Turbo CORE technology* Cache: L1-16KB data per core + 64KB instruction per module; L2-1MB per core; L3-16MB per socket Four x16 HyperTransport technology 3.0 links @ up to 6.4GT/s per link 85W to 140 W TDP (Consistent with AMD Opteron 6100 Series) Up to 3.3 GHz base frequency & up to 3.7 GHz frequency using AMD Turbo CORE technology* Cache: L1-16KB data per core + 64KB instruction per module; L2-1MB per core; L3 8MB per socket Three x16 HyperTransport technology 3.0 links @ up to 6.4GT/s per link 35 to 95W TDP (Consistent with AMD Opteron 4100 Series) 1 As of Nov 1, 2011, AMD Opteron processor Models 4200 EE have the lowest known power per core of any x86 server processor, at 35W TDP (35W/8 = 4.375W/core). Intel 's lowest power per core server processor, L5630, is 40W TDP (40W/4 = 10W/core). See http://www.intel.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.

Designed to Drive Down Power Requirements More Low Power Memory Choices Reduces Idle CPU Power By Up to 46%** Enables More Power Control for IT Intelligent Circuit Design Low and Ultra Low Voltage Memory 1.35v DIMMs reduce voltage by 10%; 1.25v DIMMs reduce voltage by 16%* C6 power state Shuts down clocks and power to idle cores TDP Power Cap Flexibility to set power limits without capping frequency All New Design Minimizes the number of active transistors for lower power and better performance * Comparison to 1.5V standard DDR3 DIMMs ** Based on testing in AMD Performance Labs as of August, 2011, an AMD Opteron processor model 6174 (12-core 2.2GHz) consumes 11.7W in the active idle C1E power state while a stanard power AMD Opteron processor model 6276 (16-core 2.3GHz) consumes only 6.4W in the active idle C1E power state with new C6 power gating employed. System configuration: Drachma reference design kit, 32GB (8 x 4GB DDR3-1333) memory, 500GB SATA disk drive, Microsoft Windows Server 2008 x64 Enterprise Edition R2. SVR-60

TDP Power cap the ultimate power and utilization tool 6000W Rack Example Allows granular control of power envelopes Maximizes the density of servers in the data center Convert untapped space in the rack into additional density Customer-definable, as granular as 1W increments Many workloads will still operate within the same performance range despite the lower TDP 17 2U servers @ 350W max power 5,950W total power 8U of wasted space 20 2U servers @ 300W max power 6,000W total power 17% more servers

AMD and the cloud Did You Know? Nearly 2 million AMD processors are engaged in cloud computing clusters worldwide* Microsoft Windows Azure Microsoft s cloud services tuned platform, with servers using AMD Opteron processor technology Facebook Open Compute Collaborate on custom-engineered technology in Facebook s data center that is focused on delivering increased energy efficiency at a lower cost Rackspace OpenStack Project Developing open source technologies for delivering a massively scalable cloud operating environment Xen Open Cloud Project Open source enterprise-ready server virtualization and cloud computing platform, delivering the Xen Hypervisor with support for a range of guest operating systems *Source: AMD internal estimates as of Q4 2011

AMD: using the cloud to build the cloud AMD s Private Cloud 124K cores at work and counting Supports thousands of engineers worldwide 40M+ tasks executed per month 90% utilization rate for both cores and systems 100% of engineering projects in the company reside in the cloud High core count, scalability and power efficiency are driving clouds to adopt AMD There are more than 2 million AMD Opteron processors at work today in the cloud

For more information AMD and Cloud Computing On AMD s use of the Cloud AMD s Cloud Computing research

Thank You! Pat Patla Vice President Product Marketing AMD