The ongoing drive to process workloads closer to the point of data origin creates requirements for compute at the edge. Delivering performance and security while meeting power and space constraints is key to taking full advantage of network edge usage models for benefits such as low latency and reduced costs for backhaul bandwidth. Advanced systems architectures that are optimized for edge computing and other distributed environments are enabling the push of data center resources outward, for implementations and use cases such as:
- Networking, including gateways and routers, security appliances, and storage
- 5G topologies, including C-RAN and D-RAN architectures
- Security, including Secure Access Service Edge (SASE)
- IoT, including smart operations
The age of deploying fixed-function devices to fill these roles has passed. General-purpose, open standards systems based on Intel architecture offer a flexible, cost-effective foundation for the next generation edge, including artificial intelligence (AI) in edge devices and appliances.
Introducing the Intel® Xeon® D-2700 and D-1700 Processors
Intel® Xeon® D-2700 and D-1700 processors are designed for provisioning dense compute at the edge that balances high computational throughput with low thermal design power (TDP). High per-core performance, advanced security features, and built-in hardware acceleration for crypto, AI, and compression support the requirements of demanding workloads within a density-optimized platform. The highly integrated design is packaged as a system-on-chip (SoC) based on a ball grid array (BGA) package for ease of
design-in and power efficiency.
The highly integrated design is well suited to the development of compact solutions for deployments targeting indoor, outdoor, and ruggedized environments, complemented by a newly extended range of operating temperatures. The SoC is also fully compatible with software and APIs across previous generations of Intel Xeon processors, as well as other Intel architecture and solutions. The resulting ease of design, development, and integration into existing Intel solutions enables low total cost of ownership and fast time to market for updated product offerings.
Solutions benefit from a range of hardware technologies built into the Intel® Xeon® D-2700 and D-1700 processors to accelerate workloads, including the following:
- Intel® Deep Learning Boost (Intel® DL Boost) accelerates AI workloads by eliminating unneeded precision in calculations so they can be completed more quickly.
- Intel® AES New Instructions (Intel® AES-NI) accelerate resource-intensive parts of the AES encryption algorithm in hardware.
- Intel® Advanced Vector Extensions 512 (Intel® AVX-512) boosts performance for demanding requirements such as AI and 5G workloads with ultra-wide 512-bit vector operations that work on more data per clock cycle than predecessor technologies.
- Intel® QuickAssist Technology (Intel® QAT) with inline IPSec support accelerates encryption and compression; the platform is capable of driving up to 100 Gbps crypto and 70 Gbps compression. The crypto capability also includes inline IPSec, which enables customers to free up precious compute cores for other applications.
Hardware-Based Security Innovations
In addition to crypto acceleration from Intel AES-NI and Intel QAT, the Intel® Xeon® D-2700 and D-1700 processors provide solution makers with cutting-edge security features built into the hardware that includes the following:
- Intel® Software Guard Extensions (Intel® SGX) protects data while in use by creating private, isolated areas of memory called secure execution enclaves where unencrypted data can be operated on, beyond the reach of software and users, regardless of their privilege levels.
- Memory Encryption supports existing software without modification while cryptographically protecting memory against hardware attacks using the NIST AES XTS encryption standard with hardware-generated keys from a hardened random-number generator implemented in silicon.
Advanced Integrated Ethernet Connectivity
Integrated Ethernet provides up to 100 Gbps of throughput, with connectivity options that offer links from 1GbE to 100GbE. For storage networking, the SoC platform provides Remote Direct Memory Access (RDMA) for memory transfers between systems that bypass the operating system, increasing throughput and reducing processor overhead and latency. RDMA capabilities include support for both Internet Wide Area RDMA Protocol (iWARP) and RoCEv2 (RDMA over Converged Enhanced Ethernet). This flexibility of transport protocols supports topologies of choice for storage architects.
The integrated NIC supports Dynamic Device Personalization (DDP) to provide multiple profiles that each specify optimizations and packet-handling parameters for specific traffic types, for increased throughput and traffic prioritization. Application Device Queues (ADQ) enables specific applications to reserve any number of dedicated Ethernet hardware queues, helping ensure predictable performance.
Integrated Ethernet functionality also includes an Enhanced Packet Processor component called the Network Acceleration Complex (NAC). The NAC is the next step in the evolution of packet processing and switching acceleration; it integrates the following:
- Network interface with enhanced scheduler that provides up to 100 Gbps of host throughput
- Flexible packet processor and switch accelerate inline packet processing
- Flexible connections, with up to eight ports at 25, 10, or 1 Gbps
Implementation Flexibility: One Architecture, Two Package Options
To broaden the range of usage models, the Intel Xeon SoC is available in two distinct packages: the high-core-count Intel® Xeon® D-2700 processor, optimized for performance, and the Intel® Xeon® D-1700 processor, which is optimized for cost and power consumption. The two options provide flexibility in the deployment of high-density compute and networking for various usage models.
Advanced Package (High Core Count): Intel® Xeon® D-2700 Processors
With 4–20 cores, the Advanced SoC package based on the Intel® Xeon® D-2700 processor is suited to demanding workloads, such as handling high data plane throughput. It operates at higher TDP than the Intel® Xeon® D-1700 processor and features higher memory performance and capacity, more PCI Express lanes, higher-bandwidth crypto, and compression through the Intel QAT accelerator. In addition, the Intel® Xeon® D-2700 processors have support for NAC with inline IPSec.
Standard Package (Low Core Count): Intel® Xeon® D-1700 Processors
Featuring 2–10 cores depending on the specific SKU, the Standard SoC package based on the Intel® Xeon® D-1700 processor is often deployed for control plane functions, as well as for lower-throughput usages such as customer premise equipment.
Upgrade Path from Previous Intel® Xeon® D Processors
The Intel® Xeon® D-2700 processors are the successors to Intel® Xeon® D-2100 processors, while the Intel® Xeon® D-1700 processors replace Intel® Xeon® D-1500 and D-1600 processors. In all cases, the upgrades provide significant, balanced, cost-effective improvements across compute, memory, and I/O.
Benefits Across Edge Workloads
For edge computing, the Intel® Xeon® D-2700 and D-1700 processors are more cost-effective, scalable, and secure than their predecessors.
Generation-over-Generation Performance Improvements1
Higher signaling and user-plane throughput enabled by an advanced microarchitecture
Data Plane Development Kit (DPDK) advantage from new Intel® AVX-512 instructions and built-in accelerators
Integrated High-Throughput Networking with Advanced Secure Connectivity
- Up to eight Ethernet ports with up to 100 Gbps packet processing capabilities with inline IPSec
- Assured support for line-rate requirements while adding more value via additional services and features
Lower Total Cost of Ownership
- Increased I/O bandwidth provided by PCIe 4.0 (16 GT/s) with up to 32 lanes
- Increased subscriber workload per node, for cost-effective deployment of advanced services
Integrated Cryptographic and AI Accelerators
- Improved Intel QAT offers better acceleration compared to previous generation
- New instructions for AI accelerate AI/deep learning workloads
Scalability up to 20 Cores
- Single standard architecture for NFV product portfolio, including Intel® Xeon® Scalable processors
- Reduced total platform investment with application, control, and data plane workload consolidation, with software back-compatibility
Advanced Features for Edge Deployments
The expanding range and importance of usages puts increasing demands on compute resources provisioned at the edge for performance, manageability, and data protection. Intel® Xeon® D processors provide new hardware-based technologies that accelerate workloads, streamline maintenance, and enhance security.
|Core Acceleration||Vector Byte Manipulation Instructions (VBMI)||In core compression/decompression acceleration for in-memory database workloads|
|VPMADD52 instruction||Public key crypto generation – SSL front end web server acceleration|
|New SHA extensions||Acceleration of hashing, SSL, TLS, IPsec, dedup, blockchain|
|Vector AES||Database workloads acceleration|
|Performance Tuning and Management||Intel® Resource Director Technology (Intel® RDT)||Monitoring and control of memory and Last Level Cache usage|
|Flexible All Core Turbo2 and Prioritized Base Frequency2||Higher frequencies for a subset of cores while all cores are active to manage application-level performance|
|Intel Speed Select2||Higher base frequency at lower core counts for dynamic SKUing|
|Internal block grounding||Power optimized sub-segment SKUing|
|Virtualization improvements||Enhancing performance of NFV workloads|
|Asynchronous DRAM Refresh (ADR)||ADR with enhanced battery backup significantly decreases battery size requirements|
|Security||Intel® SGX-Trusted Environment Mode (Intel® SGX-TEM)||Fine grain data protection by means of application isolation in memory|
|Intel® Total Memory Encryption – Multi-Tenant (Intel® TME-MT)||VM container isolation for multi-tenant platforms|
|Intel® Platform Firmware Resilience (Intel® PFR)||Protect, detect, and correct security threats in transit, boot, and runtime|
|New algorithms in Intel® QAT gen 3||SHA3, SM3, SM4, ChachaPoly added to accelerate IPsec, TLS, and DTLS workloads|
|Artificial Intelligence (AI)||Intel® Advanced Vector Extensions 512 (Intel® AVX-512)||Accelerates performance of computationally intensive workloads including AI/deep learning, scientific simulations, and financial analytics|
|Vector Neural Network Instructions (VNNI)||Delivers significant deep learning acceleration and power savings through the use of a single vector instruction set|
Package Specifications: Advanced versus Standard
The Intel® Xeon® D-2700 and D-1700 processors are similar in design but differentiated in physical form factor, core count, total design power, and other features. These differences in specifications allow the SoCs to be matched to usages with the corresponding performance, cost, space, and power requirements and constraints.
|Socket||SoC: Flip Chip Ball Grid Array (FCBGA) 52.5 mm x 45 mm||SoC: Flip Chip Ball Grid Array (FCBGA) 45 mm x 45 mm|
|Cores||4-20 with Intel® Hyper-Threading Technology||2-10 with Intel® Hyper-Threading Technology|
|Cache||LLC: 1.5 MB/core (max 30 MB) MLC: 1.25 MB/core||LLC: 1.5 MB/core (max 15 MB) MLC: 1.25 MB/core|
|Thermal Design Power (TDP)||64–125 watts||25–85 watts|
|Memory||4 channels DDR4 (2933 MT/s 2DIMMs per Channel, 3200 MT/s 1 DIMM per Channel) 8 Gb and 16 Gb density Up to 512 GB capacity with RDIMM3||2 or 3 channels DDR4 up to 2933 MT/s, 1 and 2 DIMMs per Channel 8 Gb and 16 Gb density Up to 384 GB capacity with RDIMM3|
|Integrated Intel® Ethernet||Up to 100 Gbps throughput options Connectivity: 1, 2.5, 10, 25, 40, 50, 100 GbE with RDMA (iWARP and RoCE v2)4||Up to 100 Gbps throughput options Connectivity: 1, 2.5, 10, 25, 40 GbE with RDMA (iWARP and RoCE v2)4|
|Integrated Intel® QAT||Intel® QAT gen 3: Up to 100 Gbps Crypto Up to 70 Gbps Compression 80kOps PKE RSA 2K||Intel® QAT gen 2: Up to 20 Gbps Crypto Up to 15 Gbps Compression 20kOps PKE RSA 2K|
|PCI Express||Total of 56 lanes thru combining 32 PCIe 4.0 + 24 HSIO PCIe 3.0 32 PCIe 4.0 full BW dedicated lanes from CPU complex (eight root ports) Bifurcation: x16, x8, x4 NTB via PCI 4.0 lanes: x16 and x8||Total of 40 lanes thru combining 16 PCIe 4.0 + 24 HSIO PCIe 3.0 16 PCIe 4.0 full BW dedicated lanes from CPU complex (four root ports) Bifurcation: x16, x8, x4 NTB via PCI 4.0 lanes: x16 and x8|
|SATA Support||Up to 24x SATA 3.0 over HSIO|
|High-Speed Flexible I/O||
24 High-Speed Flexible I/O Lanes configured as PCIe/SATA/USB
Up to 24 lanes of PCIe 3.0 (2.5, 5, 8 GT/s, Bifurcation support: x8, x4, x2; 12 root ports)
or up to 24 SATA 3.0, or up to four ports of USB 3.0
Combined HSIO bandwidth is limited to equivalent 16 PCIe 3.0 lanes of traffic
|Other Features||UART, LPC, SPI, eMMC 5.1, 2x USB 2.0, Intel® ME (Manageability Engine), SGX, TME-MT, PFR|
The Intel Processor Portfolio for Edge Computing
Open systems built on Intel® processors allow architects to tailor their solutions to the level of compute and performance that is needed, while allowing for implementation-specific space and power constraints. Intel® Xeon® D Processors participate in Intel’s broader edge processor portfolio, which also includes Intel® Xeon® Scalable processors and Intel Atom® C3000 processors. Together, these processor families meet the full spectrum of edge compute requirements, with full software compatibility across the portfolio.