Hybrid Memory Cubes

HMC Production Support with Generation 10 Devices

Leveraging Intel® Stratix® 10 or Arria® 10 devices, Hybrid Memory Cube (HMC) will deliver significant advantages over solutions using conventional DRAM technology which makes it an ideal solution for next generation high-performance computing, military, and wireline communication applications.

Intel’s Generation 10 HMC solution promises to deliver significant advantages over solutions using conventional SDRAM technology.

Arria 10

  • TSMC 20 nm process
  • 15% higher performance than current high-end devices
  • 40% lower mid-range power
  • Up to 96 transceiver channels
  • 28 Gbps chip-to-chip
  • 1.9 times processor systems improvement

Stratix 10

  • Intel 14 nm Tri-Gate process
  • 2X performance increase
  • Up to 70% power savings
  • 144 transceiver channels
  • 30 Gbps chip-to-chip
  • Quad-core A53 processor from ARM
  • 3D capability for integrating SRAM, DRAM, and ASIC


Watch Video: Arria 10 Hybrid Memory Cube Demonstration

Flexible, Highest Performance, and Lowest Risk Solution

Intel extends its HMC technology leadership with the production-ready HMC Controller IP core and Arria 10 support at 10, 12.5, and 15Gbps. Talk to your Intel FPGA Sales Representative about our 15 Gbps Hybrid Memory Cube Controller IP support.

Read our white paper Addressing Next-Generation Memory Requirements Using Altera FPGAs and HMC Technology to learn about the advantages of using serial memory solutions to break the memory barrier.

Stratix V FPGA and HMC Demonstration

The successful interoperability demonstration between Intel Stratix V FPGAs and Micron’s HMC allows system designers to start exploring use models for this revolutionary technology today. HMC is the long awaited answer to limitations of conventional memory technology, providing up to 15 times the bandwidth while using 70 percent less energy and 90 percent less space than existing technologies. System Designers can start evaluating HMC benefits today with the confidence that full production support will be available with Intel’s Generation 10 portfolio.

Watch Video: Accelerating System Bandwidth with Serial Memory Solutions Using Intel FPGAs and Hybrid Memory Cube Technology

Order your Stratix V HMC demonstration boards now!

HMC Demonstration Platform


Intel: Statix V 5SGXA3 FPGAs

  • x16 full-width XCVR links between each FPGA and HMC
  • Four FPGAs on board

Micron: HMC device

  • Four separate x16 wide XCVR links
  • One x16 link to each FPGA

Interoperability Configuration Set Up

XCVR data rate = 10 Gbps
Data packets, 16B, 32B, 64B, and 128B data packets (power of two payload schemes)
Exercise four HMC links


The HMC is a revolutionary, next generation DRAM technology that promises to address some of the major challenges faced by system designers. The HMC combines the best of logic and DRAM processes into a single heterogeneous package using 3D Through Silicon Via (TSV) technology to stack multiple DRAM layers over a base logic layer. The DRAM layers handle data only, while the logic layer handles all control within the HMC. The HMC technology specification has been developed by the Hybrid Memory Cube Consortium (HMCC). Intel is one of the leading consortium members and has been collaborating with Micron Technology in defining the HMC interface specification and carrying out system-level interoperability testing. The HMC specification was ratified by the consortium in April 2013.

The HMC uses a transceiver-based interface which supports up to four serial links. Each link is built using 16 full duplex transceiver channels, resulting in up to a total of 64 channels for the interface. Each link can run at data rates ranging from 10, 12.5, 15 Gbps providing up to 1 Tbps of raw aggregated interface bandwidth in each direction.


† Tests measure performance of components on a particular test, in specific systems. Differences in hardware, software, or configuration will affect actual performance. Consult other sources of information to evaluate performance as you consider your purchase.  For more complete information about performance and benchmark results, visit www.intel.com/benchmarks.