Mango DPU
Powered by AMD™Programmable DPUBoostX™ Platform
One Programmable DPU Platform for Every Data Path.
Built into every BoostX™ DPU.
Whatever the workload, every BoostX™ DPU shares the same strengths.
Programmable foundation.
An FPGA based fabric adapts to evolving protocols and new workloads.
Line rate offload.
Engines offload the data path at line rate with ultra low latency, freeing host CPU cores.
Standards compatible.
Standard PCIe and Ethernet, so every card drops into existing infrastructure.
Three cards. One platform.
Each BoostX™ card specializes the shared foundation for its workload.
BoostX™ RoCE AI
The intelligent, UEC ready RNIC built for massive AI clusters.
AI Cluster Networking400GUEC Ready
BoostX™ NRT
NVMe / RoCEv2 target. The full duplex DPU for AI scale storage.
AI Storage200GHeadless JBOF
BoostX™ Base
The programmable foundation for next generation DPUs. Build your own dataplane.
Custom Dataplane400GProgrammable IP
* Every BoostX™ card shares one FPGA based programmable DPU foundation. Specific hardware acceleration features can be enabled and customized to meet unique application workloads and customer requirements.
Same foundation. Different purpose.
The same FPGA based DPU underneath. Only the acceleration engines change.
Acceleration engines, different per card
AI Networking
BoostX™ RoCE AI
RDMA, UEC ready transport engine
Storage
BoostX™ NRT
NVMe and RoCEv2 target engine
Programmable Base
BoostX™ Base
Your own custom IP blocks
Everything below is identical on every card
Shared foundation
Programmable FPGA fabric
Reprogrammable data path, line rate offload, minimal host CPU
Host interface
PCIe Gen5
Network interface
Standard Ethernet, up to 400GbE
Need a Boost?
Our team is at the ready to create a customized plan for you to optimize and scale your business.