Monday, November 25, 2024

Cerebras and G42 unveil Condor Galaxy 3, an 8 exaFLOPs AI supercomputer

Related stories

Deep Instinct Expands Zero-Day Security to Amazon S3

Deep Instinct, the zero-day data security company built on...

Foxit Unveils AI Assistant in Admin Console

Foxit, a leading provider of innovative PDF and eSignature...

Instabase Names Junie Dinda CMO

Instabase, a leading applied artificial intelligence (AI) solution for...
spot_imgspot_img

Cerebras Systems , the pioneer in accelerating generative AI, and  G42 , the leading Abu Dhabi- based technology holding group , announced the construction of Condor Galaxy 3 (CG-3), the third cluster in their AI supercomputer constellation, Condor Galaxy. Featuring 64 CS-3 systems recently announced by Cerebras – all powered by the industry’s fastest AI chip, the Wafer-Scale Engine 3 (WSE-3) – Condor Galaxy 3 will deliver 8 AI exaFLOPs with 58 million cores powered by AI.

The strategic partnership between Cerebras and G42 has already delivered 8 exaFLOPs of AI supercomputer performance through Condor Galaxy 1 and Condor Galaxy 2, each one of the largest AI supercomputers in the world. Located in Dallas , Texas , Condor Galaxy 3 brings the current Condor Galaxy network total to 16 exaFLOPs.

“With Condor Galaxy 3, we continue to realize our shared vision of transforming the world’s artificial intelligence inventory through the development of the world’s largest and fastest AI supercomputers,” said Kiril Evtimov , CTO from the G42 group. “The existing Condor Galaxy network has trained some of the best open-source models in the industry, with tens of thousands of downloads. By doubling the capacity to 16 exaFLOPs, we look forward to seeing the next wave of innovation that Condor Galaxy supercomputers can enable.

Also Read: Renesas Unveils Powerful Single-Chip RZ/V2H MPU for Next-Gen Robotics with Vision AI and Real-Time Control

At the heart of Condor Galaxy 3 are 64 Cerebras CS-3 systems. Each CS-3 is powered by the new WSE-3 of 4 trillion transistors and 900,000 AI cores. Manufactured by TSMC at the 5-nanometer node, the WSE-3 offers twice the performance of the previous generation, for the same power and price. Designed for training the industry’s largest AI models, the WSE-3 delivers a maximum AI performance of 125 petaflops per chip.

“We are proud that our recently announced CS-3 systems will play a critical role in our pioneering strategic partnership with G42,” said Andrew Feldman , CEO and co-founder of Cerebras. “Condor Galaxy 3 and subsequent supercomputers will together provide tens of exaflops of AI computing. This marks a milestone in AI computing as it delivers unparalleled processing power and efficiency.”

Condor Galaxy has trained cutting-edge generative AI models including Jais-30B, Med42, Crystal-Coder-7B, and BTLM-3B- 8K . Jais 13B and Jais30B are the world’s best bilingual Arabic models, now available on Azure Cloud. BTLM-3B- 8K is the leading 3B model on HuggingFace, delivering 7B parameter performance in a lightweight 3B parameter model for inference. Med42, developed with M42 and Core42, is a leading clinical LLM trained on Condor Galaxy 1 in one weekend and outperforms MedPaLM in performance and accuracy.

SOURCE: PRNewswire

Subscribe

- Never miss a story with notifications


    Latest stories

    spot_img