EEMBC

EEMBC, the Embedded Microprocessor Benchmark Consortium, was a non-profit, member-funded organization established in 1997 to create objective and application-oriented performance benchmarks for embedded computing. Its work addressed a longstanding difficulty in comparing embedded processors and platforms: performance in this market depends not only on raw CPU capability, but also on compiler behavior, memory systems, operating environments, power consumption, peripheral activity, and the characteristics of the application being executed. The consortium’s central purpose was to develop benchmark suites that hardware and software companies could use to evaluate embedded processors, compilers, and system implementations under clearly specified conditions. Rather than relying solely on synthetic measures, EEMBC emphasized workloads intended to represent practical embedded applications. Members could participate in benchmark development, vote during stages of the technical approval process, and obtain early access to benchmark code and specifications before public release. This membership model was designed to combine industry expertise with a degree of procedural consistency and comparability. EEMBC’s scope expanded as embedded computing moved from relatively simple microcontrollers and single-core processors toward multicore, connected, energy-constrained, mobile, automotive, and heterogeneous systems. Its benchmark families included AutoBench for automotive, industrial, and general-purpose workloads; Networking for packet-processing applications; MultiBench for multicore scalability; CoreMark for embedded CPU performance; and FPMark for floating-point workloads. Other initiatives addressed web browsing, Android devices, low-power microcontrollers, connected Internet of Things devices, automotive advanced driver-assistance workloads, and embedded security. Among its best-known projects was CoreMark, a benchmark intended to provide a compact and standardized way to measure central-processing-unit performance in embedded systems. EEMBC also developed ULPMark to measure the energy behavior of ultra-low-power microcontrollers, including core activity, sleep modes, and selected peripheral operations. IoTMark extended the organization’s work to system-level performance and energy efficiency in connected devices, with early work focused on Bluetooth and later work on Wi-Fi. ADASMark addressed compute-intensive flows associated with automotive surround-view and other heterogeneous embedded workloads, while SecureMark considered the performance, energy, and memory effects of security implementations. The organization’s name originated as the EDN Embedded Microprocessor Benchmark Consortium. EEMBC separated from EDN Magazine in 2012 while retaining the EEMBC name. In 2023, EEMBC merged with the Standard Performance Evaluation Corporation, commonly known as SPEC. Its activities subsequently became part of the SPEC Embedded Group, ending EEMBC as an independent consortium. The EEMBC name therefore represents a former standards-oriented benchmarking organization whose technical work and benchmark lineage continued within SPEC’s embedded-computing structure.

EEMBC was formed in 1997 as the EDN Embedded Microprocessor Benchmark Consortium. Its creation reflected the need for more useful and consistent ways to compare embedded processors and related software. Embedded systems were deployed across automotive, industrial, networking, consumer, and other specialized applications, but their real-world performance could vary significantly according to compiler settings, memory architecture, operating-system support, and workload design. EEMBC sought to address this problem through benchmark suites based on application-oriented code and documented evaluation methods. The consortium operated as a non-profit organization funded by its members. Participation gave companies an opportunity to contribute to benchmark specifications and implementations, vote during development and approval stages, and receive early access to benchmark materials. This structure made EEMBC both a technical standards forum and a service for companies evaluating embedded platforms. Its intended audience included processor designers, semiconductor companies, compiler and tool vendors, system developers, and other organizations involved in embedded computing. Its early and continuing benchmark work covered several major classes of embedded workloads. AutoBench addressed automotive, industrial, and general-purpose single-threaded code. Networking focused on packet-processing activity, while MultiBench examined the scalability of multicore processors through multithreaded workloads. CoreMark became a prominent compact benchmark for embedded CPU performance. Later suites broadened the scope beyond conventional processor throughput. BXBench measured the user experience of web browsing from an interaction with a URL through page rendering, rather than focusing only on JavaScript execution. AndEBench-Pro provided a standardized method for assessing Android-platform performance. Energy efficiency became increasingly important as embedded products incorporated battery-powered and ultra-low-power devices. EEMBC created FPMark for single- and double-precision floating-point workloads across different data-set sizes, and ULPMark for measuring the energy behavior of microcontrollers. ULPMark included separate emphasis on processor-core activity and sleep modes, as well as peripheral operations such as analog-to-digital conversion, serial peripheral interfaces, real-time clocks, and pulse-width modulation. As embedded systems became connected, EEMBC expanded into system-level benchmarking. IoTConnect was designed to measure performance and energy associated with Internet of Things connectivity. Its first phase, IoTMark-BLE, focused on Bluetooth Low Energy, while IoTMark-Wi-Fi addressed energy efficiency in 802.11 connectivity. The organization also developed ADASMark for compute-intensive application flows used by heterogeneous automotive systems, initially including surround-view workloads. SecureMark examined the performance, energy, and memory consequences of security implementations in IoT and other embedded devices. The organization’s relationship with EDN changed in 2012, when EEMBC separated from EDN Magazine but kept the established acronym and brand identity. This preserved continuity for its benchmark families while making the consortium independent of the publication with which it had originally been associated. In 2023, EEMBC merged with the Standard Performance Evaluation Corporation. The resulting SPEC Embedded Group assumed responsibility for embedded benchmarking activities, bringing EEMBC’s work into the broader SPEC organization and ending EEMBC’s existence as a standalone consortium.

1. 2023Merger with SPEC

   EEMBC merged with the Standard Performance Evaluation Corporation and became part of the SPEC Embedded Group.

   Source
2. 2012EEMBC becomes independent of EDN Magazine

   EEMBC separated from EDN Magazine while retaining the consortium’s established name and acronym.

   Source
3. 2008MultiBench introduced

   EEMBC introduced MultiBench, a multithreaded benchmark intended to evaluate the scalability of multicore embedded processors.

   Source
4. 1997Consortium established

   The EDN Embedded Microprocessor Benchmark Consortium, later known as EEMBC, was formed as a non-profit, member-funded organization devoted to embedded-system benchmarking.

   Source

An industry consortium and benchmark developer focused on repeatable, application-based evaluation of embedded hardware and software, with particular attention to processor performance, energy efficiency, connectivity, automotive workloads, and system-level behavior.

• CoreMarkEmbedded CPU benchmark

  CoreMark is a compact benchmark for measuring the performance of central processing units used in embedded systems. It was designed to provide a standardized workload and a more relevant comparison point for embedded processors than broad, non-embedded performance measures alone. CoreMark became one of EEMBC’s most recognizable benchmark projects.

• AutoBenchAutomotive and industrial benchmark

  AutoBench is a single-threaded benchmark suite covering code associated with automotive, industrial, and general-purpose embedded applications. Its purpose is to evaluate processor and platform behavior against workloads representative of practical embedded deployments.

• MultiBenchMulticore benchmark2008

  MultiBench uses multithreaded code to examine the scalability of multicore processors. It addresses the difference between nominal core counts and the performance that a system can achieve when embedded workloads are divided across multiple processing threads.

• ULPMarkUltra-low-power microcontroller benchmark

  ULPMark measures energy behavior in ultra-low-power microcontrollers. Its benchmark areas include core activity and sleep modes, as well as selected peripheral operations such as analog-to-digital conversion, serial peripheral communication, real-time-clock activity, and pulse-width modulation.

• IoTMarkIoT system benchmark

  IoTMark evaluates performance and energy use associated with connecting Internet of Things devices. The family began with IoTMark-BLE, focused on Bluetooth Low Energy, and later included IoTMark-Wi-Fi for 802.11 energy-efficiency testing.

• ADASMarkAutomotive heterogeneous-computing benchmark

  ADASMark focuses on computationally intensive application flows used by embedded heterogeneous-computing architectures. Its initial workload area included automotive surround-view processing, making it relevant to advanced driver-assistance and vehicle-vision platforms.

• SecureMarkEmbedded security benchmark

  SecureMark measures the performance, energy, and memory impact of security implementations. It is intended to help developers assess the system costs associated with securing embedded and Internet of Things devices.

• BXBenchWeb browsing system benchmark

  BXBench measures web-browsing user experience across the interaction with a URL and the eventual rendering of the page. The benchmark is designed to evaluate the complete experience rather than limiting measurement to JavaScript execution.

• AndEBench-ProAndroid system benchmark

  AndEBench-Pro provides a standardized method for evaluating Android-platform performance. It was distributed as a free Android application and was intended to support comparable assessment of Android-based embedded and mobile platforms.

• 2023Merge with the Standard Performance Evaluation CorporationM&A

  EEMBC and SPEC both maintained industry benchmarking programs, including work relevant to embedded computing.

  What changed. EEMBC merged with SPEC, with its activities moving into the SPEC Embedded Group.

  Aftermath. EEMBC ceased operating as an independent consortium, while its embedded benchmark work continued under the SPEC structure.

  Source
• 2012Separate from EDN MagazineStrategy

  EEMBC had originally been associated with EDN Magazine through the name EDN Embedded Microprocessor Benchmark Consortium.

  What changed. The consortium separated from EDN Magazine but retained the EEMBC name and acronym.

  Aftermath. The change established EEMBC as an independent organization while preserving continuity for its benchmark programs.

  Source

• 2023EEMBC merges with SPEC to form the SPEC Embedded Group

  EEMBC merged with the Standard Performance Evaluation Corporation, with its embedded benchmarking activities continuing under the SPEC Embedded Group.

  M&AOther
  Source
• 2012EEMBC separates from EDN Magazine

  The consortium became separate from EDN Magazine while retaining the EEMBC name, which had originated from the EDN Embedded Microprocessor Benchmark Consortium.

  Other
  Source

1. EEMBC