We provide high-performance modular memory subsystems to customers in diverse
industries that require enterprise and storage class memory solutions to empower
critical business decisions. We have a history of introducing disruptive new
products, such as one of the first load reduced dual in-line memory modules
("LRDIMM") based on our distributed buffer architecture, which has
been adopted by the industry for DDR4 LRDIMM. We were also one of the first
to bring NAND flash memory ("NAND flash") to the memory channel with
our NVvault non-volatile dual in-line memory modules ("NVDIMM") using
software-intensive controllers and merging dynamic random access memory integrated
circuits (“DRAM ICs” or "DRAM") and NAND flash to solve
data bottleneck and data retention challenges encountered in high-performance
computing environments. We recently introduced a new generation of storage class
memory products called HybriDIMM to address the growing need for real-time analytics
in Big Data applications, in-memory databases, high performance computing and
advanced data storage solutions. We also resell NAND flash, DRAM products and
other component products to end-customers that are not reached in the distribution
models of the component manufacturers, including storage customers, appliance
customers, system builders and cloud and datacenter customers.
Due to the ground-breaking product development of our engineering teams, we
have built a robust portfolio of over 100 issued and pending U.S. and foreign
patents, many seminal, in the areas of hybrid memory, storage class memory,
rank multiplication and load reduction. Since our inception, we have dedicated
substantial resources to the development, protection and enforcement of technology
innovations we believe are essential to our business. Our early pioneering work
in these areas has been broadly adopted in industry-standard registered dual
in-line memory modules (“RDIMM”), LRDIMM and NVDIMM. Our objective
is to continue to innovate in our field and invest further in our intellectual
property portfolio, with the goal of monetizing our intellectual property through
a combination of product sales and licensing, royalty or other revenue-producing
arrangements, which may result from joint development or similar partnerships
or defense of our patents through enforcement actions against parties we believe
are infringing them.
The global high-performance memory module market is driven by increasing demand
from data center and enterprise storage applications for improved input/output
performance, lower latency and data retention capabilities in the event of unexpected
system failure. Global data center IP traffic is expected to reach 163 zettabytes
of data in 2025, up from 16.1 zettabytes in 2016, according to International
Data Corporation. The proliferation of mobile devices, social media platforms,
cognitive/artificial intelligence systems and cloud-based software applications
is resulting in the creation of unprecedented amounts of unstructured data.
In order to manage and analyze this data, we believe new computing and memory
architectures need to be developed to satisfy the needs in the industry.
In high-performance computing environments, such as cloud-based computing and
Big Data applications, a system's overall processing speed is limited to the
ability of the central processing unit ("CPU") to access data cached
in memory. Memory speeds have failed to keep pace with improvements in CPU processing
speeds, resulting in buffering delays encountered in highly intensive computing
environments. To mitigate challenges arising from differences in CPU and memory
clock speeds, data center operators have increased the number of servers in
their facilities as well as the memory content in each server. Memory capacity
is expanded through the use of DIMMs, generally incorporating up to 16 GB of
DRAM per module with today’s technology and moving up to 64/128 GB of
DRAM per module and beyond. Our technology enables an intelligent controller
to be integrated onto the DIMM, in order to manage the rapid flow of data between
the CPU and memory. The number of DIMMs incorporated into a server increases
in correlation with the number of processing cores in the CPU. DDR4 DIMMs incorporate
our load-reduction technology to mitigate the trade-off between operating speed
and memory capacity inherent in prior generations of server DIMM. These load-reduced
DIMMs, or LRDIMMs, are now the pre-dominant memory technology used in high-capacity
servers and high-performance computing clusters.
Our commercially available memory subsystem products and other products that
we sell include:
Component and Other Product Resales
In addition to our own products, we resell certain component products that
we purchase for the purpose of resale. We have purchased certain of these products,
including NAND flash and DRAM products, from Samsung under the terms of the
JDLA. We have also sourced these products from other suppliers to the extent
sufficient product is not available from Samsung to meet customer demand or
in the event of other Samsung supply issues. In 2017 and 2016, resales of these
products represented approximately 85% and 66% of our net product revenues,
respectively. Additionally, we sell excess component inventory to distributors
and other users of memory integrated circuits.
Storage Class Memory
Using an industry standard DDR4 LRDIMM interface, we believe HybriDIMM is the
industry’s first storage class memory product capable of operating in
existing Intel/AMD x86 servers without BIOS and hardware changes. HybriDIMM
unifies DRAM and NAND flash in a plug-and-play module, delivering terabyte storage
capacities operating at DRAM-like nanosecond memory speeds. HybriDIMM’s
architecture combines an on-DIMM co-processor with our software-defined data
management algorithm. HybriDIMM’s feature-set encompasses the NVDIMM functionalities
adopted by the industry. HybriDIMM dramatically improves application performance
by reducing data access latency by up to 1,000 times versus the fastest existing
storage solution known to us, and provides up to an 80% cost reduction compared
to the highest existing memory density for in-memory applications known to us.
We publicly demonstrated a HybriDIMM prototype in August 2016 and we sampled
HybriDIMM to select customers in the second half of 2017, and we are now working
with certain customers to transition to volume production.
Nonvolatile Memory
Our Vault product line enables customers to accelerate data running through
their servers and storage and reliably protect enterprise-level cache, metadata
and log data by providing near instantaneous recovery in the event of a system
failure or power outage. Our nonvolatile memory offering includes:
EXPRESSvault PCIe (EV3). EV3 is a plug-and-play memory card for server appliances
which provides data acceleration and data protection. It is compatible with
industry standard PCIe 3.0 x8 lane slots, eliminating the need for proprietary
hardware and extra motherboard space. EXPRESSvault ships with Linux and Windows
drivers compatible with standard PCIe peripheral command sets.
NVvault DDR3 NVDIMM (NV3). NV3 provides data acceleration and data protection
in a Joint Electron Device Engineering Council (“JEDEC”) standard
DD3 interface for integration into industry-standard server and storage solutions.
NV3 incorporates our patented Vault cache-to-flash controller in conjunction
with general purpose CPUs, redundant array of independent disks (“RAID”)
adapters and other devices.
NVvault DDR4 NVDIMM (NV4). NV4 relies on the NVDIMM-N interface specification
to provide data acceleration and data protection in a JEDEC standard DDR4 interface.
We have demonstrated NV4 to deliver a 200% performance improvement in transaction-based
workloads compared to PCIe NAND devices.
Specialty DIMMs and Embedded Flash
A small portion of our net product revenues is from OEM sales of specialty
memory modules and flash-based products, the majority of which are utilized
in data center and industrial applications. When developing custom modules for
an OEM system launch, we engage with our OEM customers from the earliest stages
of new product development definition, providing us valuable insight into their
full range of system architecture and performance requirements. This close collaboration
has also allowed us to develop a significant level of systems expertise. We
leverage a portfolio of proprietary technologies and design techniques, including
efficient planar design, alternative packaging techniques and custom semiconductor
logic, to deliver memory subsystems with persistence, high density, small form
factor, high signal integrity, attractive thermal characteristics, reduced power
consumption and low cost per bit.