HARDWARE & DEVICES
Access the power of the next generation of AI chips through one of our strategic partnerships and achieve the computational speed needed for your AI applications.
Cerebras
THE WAFER-SCALE ENGINE
Marvel at the size of Cerebras’ WSE-3 – a wafer-scale chip with 4 trillion transistors providing the speed and inference needed for future AI applications.
Fast
State-of-the-art output speed (566 tokens/s) and latency (time to first token: 0.23 s) on Llama 3.1 70B
Scalable
Seamlessly move from training 1B to 1T parameter models. Achieve linear performance improvements when scaling your cluster from one to thousands of Cerebras systems.
SambaNova
THE SN40L RDU
Accelerate your AI journey with the SN40L – a Reconfigurable Dataflow Unit (RDU) based on “Cerulean” architecture. SambaNova’s SN40L boasts with unbelievably fast inference and incredible efficiency.
Fast
Over 10 x faster tokens/second/user*:
-
Llama 3.1 8B 16-bit 1066 T/S
-
Llama 3.1 70B 16 bit 460 T/S
-
Llama 3.1 405B 16 bit 132 T/S
*Compared to standard GPUs.
Energy-efficient
19 kW per rack, air cooled – max. 11 kW for inference.
10 x more efficient compared to current GPU solutions.
Scalable
The only enterprise platform that scales in the cloud or on-premises.
SpiNNcloud Systems
HYBRID AI: SPINNAKER 2
The third wave of AI introduces you to the possibilities of neuromorphic computing: a brain-like, massively parallel computing technology with asynchronous processing. Combined with statistical AI, the hybrid SpiNNaker 2 system delivers supercomputing at large scale for real-time AI applications.
Fast
Ultra-low latency capabilities (< 1 ms) thanks to modular and light-weight network-on-chip communication.
Energy-efficient
Combining the strengths of deep neural networks, symbolic AI, and neuromorphic models to enhance the robustness and energy efficiency.
Scalable
Available as server, rack, or modular container solution — ready to go with WAIYS liquid cooling systems.