Don't let costs define the limits of your research.
Achieve 2.4x cost efficiency with Wormhole™ to accelerate your experiments and innovations.
2.4x higher cost efficiency than the H200. Train and experiment with large-scale AI models effortlessly, even on limited budgets. The Tenstorrent Wormhole™ AI processor, integrated into the liquid-cooled dg5W, delivers outstanding energy efficiency and thermal control.
※ Equipped with AI processors supplied through an official Tenstorrent partnership.High-performance, low-power architecture
Wormhole™ offers exceptional performance per dollar with outstanding efficiency and cost-effectiveness.
Specialized AI performance
Optimized for LLM and machine learning workloads with matrix operations and high-bandwidth memory.
Flexibility
Maximize development flexibility with the TT-Forge™🔗 and TT-Metalium™🔗 open-source SDKs.
Performance per cost (based on Llama 3.1 70B)
※ Higher values indicate better performance
Tenstorrent n300 (1.0x)
Tenstorrent n300 (1.0x)
NVIDIA H200 (0.4x)
NVIDIA H200 (0.4x)
NVIDIA RTX 4090 (0.4x)
NVIDIA RTX 4090 (0.4x)
RTX A6000 (0.3x)
RTX A6000 (0.3x)
Key features of the dg5W x Tenstorrent product.
SPEC
20 ~ 53dB (Full load)
Quietness
660mm x 220mm x 445mm
Size
W x D x H
(660mm x 220mm x 551mm including handle and wheels)
AMD EPYC 9124
CPU
4 x Tenstorrent Wormhole n300
AI Processor
4 x 64GB DDR5-5600 ECC/REG
Memory
1TB M.2 NVMe SSD
HDD
7 PCIe 5.0 x16
PCIe
Windows, RHEL, Ubuntu, Rocky Linux
OS
2 x 10GbE Base-T
Network
IPMI
For 3 years
Warranty
4 x 1,300W(5.2kW) 80 Plus Platinum
Power
Manual download
Manual & Warranty