Furiosa RNGD - Gen 2 data center accelerator
<p><strong><span class="legendSpanClass"><span class="xn-location">SANTA CLARA</span></span></strong> &#8212; <a href="https://c212.net/c/link/?t=0&;l=en&;o=4239861-1&;h=1999214758&;u=https%3A%2F%2Ffuriosa.ai%2F&;a=FuriosaAI" target="_blank" rel="nofollow noopener">FuriosaAI</a>, an emerging leader in the AI semiconductor space, has unveiled RNGD (pronounced &#8220;Renegade&#8221;), a leading AI accelerator, at Hot Chips 2024. RNGD is positioned to be the most efficient data center accelerator for high-performance large language model (LLM) and multimodal model inference, disrupting an AI hardware landscape long defined by legacy chipmakers and high-profile startups. Founded in 2017 by three engineers with backgrounds at AMD, Qualcomm, and Samsung, the company has pursued a strategy focused on rapid innovation and product delivery which has resulted in the unveiling and fast development of RNGD.</p>
<div class="pull-right inline-gallery-container col-md-8 col-sm-7 col-xs-12">
<div class="gallery inline-gallery">
<div class="row">
<div class="col-sm-12"></div>
<div class="col-sm-12 clearfix">
<figure>
<div class="image lightbox-item " data-src="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?p=publish" data-asset-type="photo" data-asset-id="Life_After_Debt_Aug_17_Event.jpg" data-asset-label="General" data-sub-html="June Paik, Co-Founder and CEO of FuriosaAI." data-tweet-text="June Paik, Co-Founder and CEO of FuriosaAI." data-facebook-share-text="June Paik, Co-Founder and CEO of FuriosaAI." data-linkedin-text="June Paik, Co-Founder and CEO of FuriosaAI." data-download-url="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?p=publish" data-pinterest-text="June Paik, Co-Founder and CEO of FuriosaAI." data-twitter-share-url="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?p=twitter" data-linkedin-share-url="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?p=linkedin" data-facebook-share-url="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?p=facebook" data-pinterest-share-url="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?p=facebook"><a class="tabfocus" role="button"><img id="imageid_2" class="gallery-thumb img-responsive" title="June Paik, Co-Founder and CEO of FuriosaAI." src="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?w=400" data-getimg="https://mma.prnewswire.com/media/2489185/FuriosaAI_june_ceo.jpg?w=400" /></a></div><figcaption>June Paik, Co-Founder and CEO of FuriosaAI.</figcaption></figure>
</div>
</div>
</div>
</div>
<p>Furiosa successfully completed the full bring-up of RNGD after receiving the first silicon samples from their partner, TSMC. This achievement reinforces the company&#8217;s track record of fast and seamless technology development. With their first-generation chip, introduced in 2021, Furiosa submitted their first MLPerf benchmark results within 3 weeks of receiving silicon and achieved a 113% performance increase in the next submission through compiler enhancements.</p>
<p>Early testing of RNGD has revealed promising results with large language models such as GPT-J and Llama 3.1. A single RNGD PCIe card delivers 2,000 to 3,000 tokens per second throughput performance (depending on context length) for models with around 10 billion parameters.</p>
<p>&#8220;The launch of RNGD is the result of years of innovation, leading to a one-shot silicon success and exceptionally rapid bring-up process. RNGD is a sustainable and accessible AI computing solution that meets the industry&#8217;s real-world needs for inference,&#8221; said <span class="xn-person">June Paik</span>, Co-Founder and CEO of FuriosaAI. &#8220;With our hardware now starting to run LLMs at high performance, we&#8217;re entering an exciting phase of continuous advancement. I am incredibly proud and grateful to the team for their hard work and continuous dedication.&#8221;</p>
<p><b>RNGD&#8217;s key innovations include:</b></p>
<ul >
<li>A non-matmul, Tensor Contraction Processor (TCP) based architecture that enables a perfect balance of efficiency, programmability and performance.</li>
<li>Programmability through a robust compiler co-designed to be optimized for TCP that treats entire models as single-fused operations.</li>
<li>Efficiency, with a TDP of 150W compared to 1000W+ for leading GPUs</li>
<li>High-performance, with 48GB of HBM3 memory delivering the ability to run models like Llama 3.1 <span class="xn-money">8B</span> efficiently on a single card.</li>
</ul>
<p>The chip is currently sampling to early access customers, with broader availability expected in early 2025.</p>

SAN FRANCISCO -- Wispr, the voice-to-text AI that turns speech into clear, polished writing in every…
SAN FRANCISCO -- Numeric, an AI accounting automation platform, has raised a $51 million Series…
Apple has announced 45 finalists for this year’s App Store Awards, recognizing the best apps…
The University of California (UC) and the California Nurses Association (CNA) have reached a tentative…
SAN FRANCISCO -- House Rx, a health tech company focused on making specialty medications more accessible and…
Britain's King has given an award to the King of NVIDIA! NVIDIA founder and CEO…