Efficient scale-out for the large-scale generative AI era

Scale deep learning with native integration
of 24 100 Gigabit Ethernet on every Gaudi2 accelerator

Architected to scale deep learning acceleration from one to thousands of processors, Gaudi2 offers unique advantages of substantial scaling capacity with 24 100 Gigabit Ethernet ports on every Gaudi2. Integration of Ethernet connectivity provides tangible advantages of flexibility, efficiency and cost-efficiency.

Accelerator Architecture

Gaudi2 board

Server Architecture

Gaudi2 server architecture

Advantages of Gaudi2
Integrated Connectivity

Gaudi2’s integrated networking gives the customer greater versatility and freedom to build the system that fits their needs


large scale
  • All-to-all scale-up
  • 2.4TB scale-out
  • Reduces network bottlenecks


Flexible icon
  • Non-proprietary,
    industry standard
  • Scale from one to thousands of Gaudi2s
  • Wide variety of
    Ethernet switches for build out


  • Most data centers already utilize Ethernet
  • Ease of build with industry standard
  • Cost-efficient with breadth of Ethernet choices


Control bars
  • Industry standard
  • No vendor lock in with proprietary solutions
  • Cost control through choice of system equipment

Scale-out scenarios

Required Components

Optional Components / Examples

small pod

Small Pod Architecture

16-40 Gaudi2s (2-5 servers, with 8x Gaudi2 each)

Large pod icon

Large Pod Architecture

  • Supports variable ratio of nodes-to-switches
  • 3x 400G switches + up to 128 Gaudi2s (16 servers with 8x Gaudi2 each)
Mega pod architecure

Gaudi2 MegaPod Architecture

Large clusters can be easily built using multiple MegaPods

  • 8 Gaudi2 servers & 3 switches
  • 8x Gaudi2 per node (server)
  • Leaf switches are placed close to Gaudi2 nodes to minimize communication latency
  • Copper cables can be used within pod

For access to Gaudi2 servers, see Supermicro.

For access to Gaudi2 servers in China, Contact us.

Learn more about
Networking with Gaudi

Contact us