Tenstorrent Blackhole P150b AI Accelerator – 32GB GDDR6 | 664 TFLOPS | 120 Tensix Cores | 512GB/s | Passive Cooling | PCIe 5.0

Enterprise-Grade AI Acceleration for High-Performance Servers & Scalable AI Infrastructure

Experience next-generation artificial intelligence computing with the Tenstorrent Blackhole P150b, a high-performance, passively cooled PCIe AI accelerator designed for rack-mounted servers, AI research laboratories, enterprise computing infrastructure, and multi-accelerator deployments.

Powered by Tenstorrent’s innovative Blackhole Tensix architecture, the P150b combines 120 Tensix AI processing cores, 16 SiFive X280 RISC-V processor cores, 32GB GDDR6 memory, 512GB/s memory bandwidth, and up to 664 TFLOPS of BLOCKFP8 compute performance.

Unlike conventional desktop AI accelerator cards, the Blackhole P150b uses a passive heatsink engineered for server environments with high-static-pressure forced-air cooling.

Its four dedicated 800Gbps QSFP-DD interconnect ports enable direct communication between compatible Blackhole accelerators, making it a compelling platform for scalable AI research, distributed machine learning, and high-density server installations.

Key Features & Highlights

• 664 TFLOPS Peak BLOCKFP8 Compute – High-throughput AI acceleration for supported machine learning applications.
• 120 Tensix AI Cores – Dedicated processing resources for tensor-intensive computing.
• 16 SiFive X280 RISC-V Cores – Programmable processor architecture integrated into the Blackhole chip.
• 32GB GDDR6 Memory – High-capacity onboard memory for AI models and data processing.
• 512GB/s Memory Bandwidth – High-speed memory access for demanding computational workloads.
• 180MB On-Chip SRAM – Fast local memory for efficient AI processing.
• PCI Express 5.0 x16 – High-bandwidth connection to compatible server motherboards.
• 4 × QSFP-DD 800Gbps Interconnects – Direct accelerator-to-accelerator connectivity.
• Passive Server Cooling – Designed for rack-mounted systems with sufficient forced airflow.
• 300W Total Board Power – High-performance accelerator operating envelope.
• Up to 1.35GHz AI Clock – Advanced Blackhole processing architecture.
• Open-Source Software Ecosystem – Access to Tenstorrent development tools and programming resources.

Powerful Blackhole Tensix AI Architecture

The Tenstorrent Blackhole P150b incorporates an advanced AI processing architecture designed for modern machine learning and neural network applications.

At its core are 120 specialized Tensix processing cores engineered to perform tensor operations efficiently.

The processor also integrates 16 SiFive X280 RISC-V cores, providing additional programmable computing capabilities.

664 TFLOPS AI Processing Performance

The Blackhole P150b delivers up to 664 TFLOPS of theoretical BLOCKFP8 computing throughput.

This performance makes the accelerator suitable for exploring:

• Artificial intelligence inference
• Neural network development
• Transformer-based AI architectures
• Generative AI applications
• Computer vision processing
• Custom AI kernel optimization
• Distributed machine learning research

Performance depends on supported numerical formats, model architecture, software implementation, and system configuration.

32GB GDDR6 High-Bandwidth Memory

Modern artificial intelligence workloads require fast and efficient memory access.

The Blackhole P150b features an advanced memory subsystem with:

32GB GDDR6 Memory

Provides onboard storage for model parameters, intermediate tensors, and supported computational workloads.

512GB/s Memory Bandwidth

Enables high-speed transfer of data between onboard memory and processing resources.

256-bit Memory Interface

Supports efficient data movement across the accelerator’s memory subsystem.

180MB On-Chip SRAM

Provides fast local memory resources for supported processing operations.

This combination makes the P150b particularly attractive for advanced AI development and multi-accelerator computing environments.

Four 800Gbps QSFP-DD Ports for Multi-Accelerator Scalability

One of the defining capabilities of the Tenstorrent Blackhole P150b is its high-speed accelerator interconnect architecture.

The card includes four passive QSFP-DD ports, each providing 800Gbps connectivity.

These ports support direct communication between compatible Tenstorrent Blackhole P150a and P150b accelerator cards.

Benefits of Multi-Card Connectivity

Scalable AI Computing

Build supported multi-card configurations to increase aggregate computational resources.

Expanded Distributed Memory

Combine the memory resources of multiple accelerator cards for supported distributed workloads.

High-Bandwidth Communication

Dedicated interconnects support fast communication between compatible Blackhole processors.

Enterprise AI Infrastructure

Deploy multiple accelerators in compatible rack-mounted servers for research and development.

Flexible AI Research

Explore distributed model execution, multi-device computing, and advanced AI software optimization.

QSFP-DD ports are intended exclusively for supported Tenstorrent Blackhole accelerator connections, not general-purpose Ethernet networking. Compatible cables and supported software are required.

Passive Cooling Designed for Rack-Mounted Servers

The Blackhole P150b is specifically engineered for high-density server environments.

Unlike the Blackhole P150a, which includes an active blower-style cooling solution, the P150b uses a passive heatsink.

Advantages of Passive Server Cooling

• No dedicated onboard cooling fan
• Designed for centrally managed server airflow
• Suitable for compatible high-density rack-mounted systems
• Enables server integrators to manage cooling through chassis fans
• Supports multi-accelerator installations with appropriate thermal design

Important: Passive cooling does not mean the accelerator can operate without airflow. The P150b requires sufficient high-static-pressure forced-air cooling from the host server.

It should not be installed in an ordinary desktop computer without a validated cooling solution.

Applications & Use Cases

1. Enterprise AI Servers

Deploy compatible Blackhole accelerators in rack-mounted infrastructure for specialized artificial intelligence workloads.

2. Large Language Model Research

Explore supported transformer architectures, language model inference, and distributed AI processing.

3. Deep Learning & Neural Networks

Develop, test, and optimize supported neural network workloads.

4. Generative Artificial Intelligence

Experiment with compatible generative AI models and inference applications.

5. AI Research Laboratories

Support universities, research institutions, and advanced engineering teams working on AI accelerator technology.

6. Multi-Accelerator Computing

Interconnect multiple P150a and P150b cards using supported QSFP-DD cables and configurations.

7. Custom AI Software Development

Explore hardware-level programming and custom AI kernel development using Tenstorrent’s open-source software stack.

8. High-Performance Computing Research

Investigate specialized parallel computing workloads that can be implemented using the Tensix architecture.

Complete Technical Specifications

System Requirements & Installation

The Tenstorrent Blackhole P150b is intended for compatible server installations.

Recommended Hardware Environment

• CPU: Compatible x86-64 host processor
• Motherboard: PCIe 5.0 x16 slot without bifurcation recommended
• Power Supply: ATX 3.1 certified or newer
• External Power: One compatible 12V-2x6 connector per card
• Cooling: Server chassis with sufficient high-static-pressure forced airflow
• Expansion Clearance: Dual-slot space per accelerator
• Operating System: Supported Linux distribution and Tenstorrent software stack
• Interconnect: Compatible QSFP-DD cables for multi-card deployments

For installations involving multiple accelerator cards, verify total server power capacity, chassis airflow, PCIe slot availability, interconnect topology, and software compatibility.

Why Choose the Tenstorrent Blackhole P150b?

1. Enterprise-Oriented Design

Engineered for rack-mounted servers and high-density AI computing environments.

2. Powerful AI Acceleration

Delivers up to 664 TFLOPS of theoretical BLOCKFP8 compute throughput.

3. High-Capacity Memory

32GB GDDR6 memory with 512GB/s bandwidth supports demanding AI development workloads.

4. Multi-Card Scalability

Four dedicated 800Gbps QSFP-DD ports support compatible accelerator interconnections.

5. Passive Server Cooling

Integrates with properly designed server cooling infrastructure without requiring an onboard fan.

6. Open-Source Software

Tenstorrent’s development ecosystem enables experimentation and custom AI software development.

7. Advanced RISC-V Architecture

Combines dedicated Tensix AI cores with programmable RISC-V processor cores.

Frequently Asked Questions

Q1. What is the Tenstorrent Blackhole P150b?

The P150b is a passively cooled PCIe AI accelerator designed for rack-mounted servers, machine learning research, and scalable multi-card computing.

Q2. How much memory does the Blackhole P150b have?

It features 32GB GDDR6 memory with 512GB/s memory bandwidth.

Q3. What is the difference between Blackhole P150a and P150b?

Both offer the same core specifications, including 32GB GDDR6, 120 Tensix cores, 664 TFLOPS BLOCKFP8 performance, and four 800Gbps QSFP-DD ports.

The primary difference is cooling: P150a uses active cooling for desktop workstations, while P150b uses passive cooling for rack-mounted servers.

Q4. Does passive cooling mean the P150b operates without fans?

No. It requires forced airflow supplied by the server chassis.

Q5. Can the P150b be used in a desktop PC?

It is not recommended for conventional desktops unless the system has a validated high-pressure forced-air cooling solution.

Q6. Can multiple P150b cards be connected together?

Yes. Compatible P150a and P150b cards can be interconnected using supported QSFP-DD cables.

Q7. Does the P150b support NVIDIA CUDA?

No. It uses Tenstorrent’s software ecosystem rather than NVIDIA CUDA.

Q8. Is the P150b suitable for large language models?

It can be used for supported language model workloads. Actual compatibility depends on model architecture, memory requirements, software support, and configuration.

Q9. What is the maximum power consumption?

The specified total board power is 300W.

Q10. Is the P150b suitable for AI data centres?

Yes, it is designed for compatible server environments and can be deployed in properly engineered AI computing infrastructure.

Buy Tenstorrent Blackhole P150b AI Accelerator in India

Build scalable AI computing infrastructure with the Tenstorrent Blackhole P150b – a high-performance, server-oriented AI accelerator designed for advanced machine learning and artificial intelligence research.

Featuring 32GB GDDR6 memory, 664 TFLOPS peak BLOCKFP8 compute performance, PCIe 5.0 connectivity, four 800Gbps QSFP-DD ports, and passive server cooling, the P150b is designed for organizations developing specialized AI computing environments.

Order Now | Request Best Price | Enquire for Availability | Contact for Enterprise & Bulk Orders

Final pricing, availability, warranty, shipping, and compatibility are subject to confirmation.

Primary keyword: Tenstorrent Blackhole P150b India

Secondary keywords: Buy Tenstorrent P150b India, Blackhole P150b Price India, Tenstorrent 32GB AI Accelerator, Passive Cooled AI Card, Server AI Accelerator India, 664 TFLOPS AI Processor, PCIe 5.0 AI Accelerator, RISC-V AI Hardware, Tenstorrent Data Centre Accelerator, QSFP-DD AI Card.