Specifications
Product
Tenstorrent Wormhole n150s AI accelerator PCIe card
Part Number
TC-02001
Architecture
Wormhole, RISC-V based Tensix cores
Tensix Cores
72
On-chip SRAM
108 MB distributed SRAM
Memory
12 GB GDDR6, 288 GB/s bandwidth
TDP
160 W
Host Interface
PCIe Gen4 x16
Software Stack
TT-Metalium, TT-NN, TT-Forge (open source)
Model Support
LLMs, CNNs and vision transformers via TT-NN / TT-Forge
Configuration
Single Wormhole ASIC (n150) — the n300 card carries two ASICs
Card Family
Wormhole PCIe cards: n150d / n150s / n300d / n300s
Target Workloads
AI inference and model development at the edge/desktop
Availability
In stock — quote for volume
Overview
The Tenstorrent Wormhole n150s (order code TC-02001) is a single-ASIC PCIe accelerator card built around 72 Tensix cores — RISC-V based compute tiles with their own vector and matrix engines. Tenstorrent publishes the whole Wormhole PCIe card family openly, and the n150s is the single-chip, actively-cooled member.
The design trades on memory density close to compute: 108 MB of distributed on-chip SRAM sits next to the Tensix cores, backed by 12 GB of GDDR6 at 288 GB/s and a 160 W envelope. For inference workloads the open-source TT-Metalium / TT-NN / TT-Forge stack lets teams write and tune their own kernels rather than depend on a closed compiler.
QS Compute supplies Wormhole cards alongside the Blackwell-generation Tenstorrent parts and the mainstream NVIDIA and AMD accelerators, so customers can build heterogeneous inference nodes.
Key Benefits
72 Tensix RISC-V cores with 108 MB SRAM keep activations on-chip for LLM and vision inference. Open-source TT-Metal stack avoids vendor lock-in and lets teams tune kernels. 160 W / single slot-class power enables dense edge inference nodes. 12 GB GDDR6 at 288 GB/s holds larger models and batched weights.
Applications
Edge and on-premises LLM inference; vision and transformer inference nodes; research and university AI clusters; heterogeneous inference servers needing low power per token; and model-development workstations using the open TT-Metal stack.
Request a Quote — TENSTORRENT WORMHOLE N150S — 72-CORE RISC-V AI ACCELERATOR PCIE CARD
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →