Tenstorrent Previews Large Compute Cluster, Generates Video Faster Than Real Time

2026-04-21 · Source: Big Data & AI News - EE Times · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

Tenstorrent demonstrated its new BlackHole-generation Galaxy servers, showcasing a video generation demo that produced a five-second, 720p video from a text prompt in 2.4 to 3 seconds. This performance, achieved using an optimized Wan2.2-14B model developed with partner Prodia, is approximately 10 times faster than the same model on other leading hardware, including Nvidia and xAI’s grok-imagine-video. The demo utilized four Galaxy servers, housing 128 Tenstorrent BlackHole chips. Tenstorrent emphasizes its unified hardware architecture, which integrates compute, memory, and networking to run large software programs without proprietary interconnects, aiming for general-purpose AI computing rather than specialized hardware.

Key takeaway

For Machine Learning Engineers developing video generation applications, Tenstorrent's new Galaxy servers offer a significant performance advantage, generating five-second videos in under three seconds. You should consider evaluating this platform for projects requiring high-speed video inference, especially given its general-purpose architecture and open-source software stack, which promises adaptability to future model advancements and straightforward porting of existing diffusion models.

Key insights

Tenstorrent's new hardware accelerates video generation by 10x, breaking the real-time barrier for five-second videos.

Principles

• Unified hardware architecture enhances scalability.
• General-purpose AI computing resists model evolution obsolescence.

Method

An optimized Wan2.2-14B model, developed by Prodia, runs on Tenstorrent's BlackHole-generation Galaxy servers, leveraging 128 BlackHole chips for parallel processing of iterative denoising.

In practice

• Generate 5-second, 720p videos in under 3 seconds.
• Port existing diffusion models to Tenstorrent's open-source stack.

Topics

• Tenstorrent
• Video Generation
• BlackHole Chips
• Galaxy Servers
• Wan2.2-14B Model

Best for: Machine Learning Engineer, Computer Vision Engineer, CTO, AI Hardware Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• Tenstorrent’s Galaxy Blackhole AI servers escape the event horizon — Tenstorrent's Galaxy Blackhole offers a scalable, cost-effective AI compute platform with competitive performance for large models.
• Vibe coding is about to get so fast — GPT 5.3 Spark leverages Cerebras wafer-scale chips for unprecedented coding inference speed.
• Collaborative Few-Step Distillation and Low-Bit Quantization for Wan2.2 Dual-Expert Video Diffusion Models — Co-designing distillation and low-bit quantization compresses video diffusion models, surpassing full-precision at fewer steps.
• NVIDIA RTX Accelerates 4K AI Video Generation on PC With LTX-2 and ComfyUI Upgrades — NVIDIA RTX GPUs enhance 4K AI video generation on PCs via LTX-2 and ComfyUI optimizations.
• Here’s how our TPUs power increasingly demanding AI workloads. — TPUs are Google's custom chips optimized for high-speed AI model computation.
• Qualcomm in Talks to Buy Tenstorrent to Expand AI Chip Capabilities — Qualcomm's potential acquisition of Tenstorrent signals strategic expansion in the competitive AI chip market.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Big Data & AI News - EE Times.