Specifications
Vendor
Napatech
DPU Models
F2070X (2x100GbE), F3076X
F2070X FPGA
Intel/Agilex AGFC023
F2070X SoC
Intel Xeon D-1736, 8 cores, 2.3-3.4 GHz, 15MB cache
F3076X FPGA
Intel/Agilex AGIC041
F3076X SoC
Intel Xeon 6532P-B, 32 cores, 2.20 GHz, 128MB cache
Memory
4x4GB 40-bit DDR4 with ECC — 18GB total
PCIe Interface
PCIe Gen3 x8
Max Bidirectional Bandwidth
822 Gbps (F2070X)
Crypto Offload
AES and SM4 (F2070X / F3076X)
Network Ports
2x 100GbE per DPU
SmartNIC Models
NT400D13, NT400D11, NT200A02, NT50B01, N3070X
Software Stack
Link-Capture, Link-Inline, Link-Virtualization, Link-Storage, Link-Security
Offload Domains
Networking, storage (NVMe-oF), security, virtualisation
Deployment Modes
Bare metal, cloud-native, virtualised (vSwitch / SR-IOV)
Target Segments
Telecom NFV/UPF, cyber security, network monitoring, storage acceleration
Overview
Napatech builds SmartNIC and Data Processing Unit cards that combine an FPGA with a general-purpose server CPU on a single PCIe card. That FPGA-plus-Xeon combination is the defining architectural choice: the FPGA handles deterministic, line-rate packet and data-path processing while the on-card Xeon runs the control logic and heavier software that would otherwise consume host CPU cycles.
The F2070X is the flagship DPU — a 2x100GbE PCIe card carrying an Intel/Agilex AGFC023 FPGA alongside an Intel Xeon D-1736 SoC with eight cores at 2.3-3.4GHz and 15MB of cache. It supports up to 822 Gbps of bidirectional throughput across a PCIe Gen3 x8 host interface, with AES and SM4 crypto acceleration. The F3076X scales the compute side considerably: an AGIC041 FPGA paired with a 32-core Intel Xeon 6532P-B with 128MB cache.
The SmartNIC line — NT400D13, NT400D11, NT200A02, NT50B01 and N3070X — covers the lower port-count end of the same architecture, with some models offered in both passive-cooled and link-programmable variants. Software is what makes the hardware usable: Napatech's Link software family is split by workload into Capture, Inline, Virtualization, Storage and Security packages, so the same card can be deployed as a capture appliance, an inline security accelerator, an NVMe-over-Fabric storage offload engine or a virtual switch accelerator.
Typical deployments are telecom NFV and UPF offload, cyber-security inspection, high-density network monitoring and storage acceleration in virtualised or cloud-native environments. QS Compute supplies Napatech SmartNIC and DPU hardware for B2B integration projects — contact us with your port count and offload requirement.
Key Benefits
FPGA + full Xeon SoC on one PCIe card — line-rate data-path processing plus on-card general-purpose compute. Up to 32 cores and 128MB cache (F3076X) for heavy inline workloads. AES and SM4 crypto offload built into the card. Five workload-specific software suites let one hardware platform serve capture, inline security, virtualisation, storage and security use cases.
Applications
Telecom NFV and 5G UPF offload, network monitoring and packet capture, inline cyber-security inspection, NVMe-over-Fabric storage acceleration, virtual switch acceleration, and cloud-native network function hosting.
Request a Quote — NAPATECH SMARTNIC & DPU PORTFOLIO — F2070X, F3076X, NT400D13
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →