Nvidia's CEO, Jensen Huang, delivered an inspiring keynote at Siggraph, highlighting the company's successful bet on AI back in 2018. The gamble has paid off handsomely, with the introduction of a range of new products and partnerships. The most noteworthy among these is the "Grace Hopper" (GH) super chip, hailed by Huang as the computing world's "killer app". This chip boasts remarkable performance, being ten times faster, significantly cheaper, and having lower power requirements.

Huang stressed the democratizing impact of AI, emphasizing that this technology is accessible to everyone. The event also showcased other exciting developments. One such highlight is the Dreamscape free-roam experience, "Clockwork Forest", originating from the labs in Geneva. This innovation represents a significant shift in technology, moving away from backpack PCs. It's the result of extensive research and development by Vicon and Artanim, the Swiss research institute behind Dreamscape's VR tech platform.

While the technology is still in its prototype stage and not yet available on the market, it holds tremendous potential across various applications, including Hollywood VFX, AAA games, and life sciences and engineering. The Dreamscape experience will be open to the public at the LA location starting early August.

Siggraph also celebrated its 50th anniversary with the Time Tunnel, an immersive exhibit illustrating the milestone developments in the computer graphics industry. Created by ACM, Autodesk and a student team at the University of California, Santa Barbara, this exhibit provides an engaging look at the industry's evolution.

Another fascinating piece on display was Mort Heilig's Sensorama, a mechanical marvel. USC's Scott Fisher mentioned ongoing restoration efforts, aiming to bring the fiberglass and wood hull, which contains a movie projector and a smell-o-vision apparatus, back to life. The films created by Heilig showcased a mesmerizing handheld adventure with a beautiful girl on the California coast and a thrilling 60 mph motorcycle ride through 1956 New York. Heilig had originally priced the immersive movie theater at $6,000 (equivalent to $67,000 today), requiring 24,000 plays at $.25 per play to break even. However, the impracticality of the Sensorama's design makes it an idea ahead of its time.

The event also featured the famous Ivan Sutherland 1968 demo, which recreated his lab and initial VR experiences. This historic moment was brought back to life in VR on the Quest, providing a brief but intriguing glimpse into the early days of virtual reality.

Frontiers sessions will kick off today's schedule at 8 am with promises of astonishing insights and groundbreaking discoveries!

At the forefront of NVIDIA's priorities during the SIGGRAPH computer graphics conference on August 8th was the realm of generative AI. Key announcements in the enterprise and industrial sectors revolved around a trifecta of advancements: the Hugging Face Training Service, fueled by the might of NVIDIA DGX Cloud; the cutting-edge iteration of NVIDIA AI Enterprise (version 4.0); and the unveiling of the AI Workbench toolkit, all set to redefine the landscape of generative AI deployments.

Unveiling the 'Superchip' with HBM3e Processor for Elevated AI Development

NVIDIA seized the moment to introduce the GH200, the forthcoming addition to the Grace Hopper platform series. Anchored by the innovative Grace Hopper 'superchip', this iteration, slated as the GH200, comprises a singular server equipped with 144 Arm Neoverse cores, boasting a staggering eight petaflops of computational power and 282GB of memory, all of which operate under the HBM3e standard. This standard introduces a remarkable 10TB/sec bandwidth, an impressive threefold enhancement over its predecessor, HBM3.

Jensen Huang, the visionary founder and CEO of NVIDIA, emphasized, "To cater to the burgeoning demand for generative AI, data centers require accelerated computing platforms with specialized capabilities. The new GH200 Grace Hopper Superchip platform delivers precisely that through unparalleled memory technology, bandwidth optimization to bolster throughput, the capacity to interlink GPUs for enhanced combined performance, and a server architecture designed for effortless deployment across the entire data center."

DGX Cloud AI Supremacy, Now Accessible through Hugging Face

NVIDIA's acclaimed DGX Cloud AI supercomputing prowess takes on a new avenue by joining forces with Hugging Face, a pivotal player in the generative AI realm. This alliance facilitates individuals seeking to train and fine-tune generative AI models, available on the Hugging Face marketplace. Organizations that harbor a need for custom generative AI solutions often find themselves compelled to train models on proprietary data, a process demanding considerable bandwidth.

Manuvir Das, NVIDIA's Vice President of Enterprise Computing, articulated the harmonious collaboration, stating, "The synergy between Hugging Face and NVIDIA is innate, with Hugging Face being the ultimate repository for initial exploration, while NVIDIA DGX Cloud provides an ideal arena to engage in generative AI work utilizing these models." The DGX Cloud offering encompasses NVIDIA Networking, a high-performance, low-latency fabric, and houses eight NVIDIA H100 or A100 80GB Tensor Core GPUs on each node, collectively endowing a substantial 640GB of GPU memory.

The DGX Cloud AI training will incur supplementary charges within the Hugging Face framework, albeit precise pricing details remain undisclosed. This collaborative initiative is poised to launch in the upcoming months. Clément Delangue, the Co-founder and CEO of Hugging Face, lauded the partnership's potential, asserting, "The global community is forging novel connections and discoveries through generative AI tools, and we are merely scratching the surface of this technological revolution. Our alliance will usher in NVIDIA's forefront AI supercomputing to Hugging Face, empowering enterprises to seize control of their AI destiny via open-source resources."

AI Enterprise 4.0: A Glimpse into the Future

A forthcoming iteration of NVIDIA's AI Enterprise, the comprehensive suite encompassing AI and data analytics software tailored for crafting generative AI solutions, is slated to transition to version 4.0. The hallmark feature in this iteration is the incorporation of NeMo, an instrumental platform for bespoke tooling in generative AI curation, training customization, inference, establishing safeguards, and beyond. NeMo introduces a cloud-native framework, designed to facilitate the development and deployment of enterprise applications founded upon extensive language models.

Renowned machine learning providers, including ClearML, Domino Data Lab, Run:AI, and Weights & Biases, have collaborated with NVIDIA to seamlessly integrate their services with AI Enterprise 4.0, ensuring a synergistic experience.

AI Workbench: Elevating the Generative AI Workflow

AI Enterprise 4.0 forges an alliance with NVIDIA AI Workbench, a dedicated workspace engineered to streamline and simplify the process of launching AI applications on personal computers or home workstations. AI Workbench empowers projects to seamlessly traverse across various domains, such as PCs, data centers, public clouds, and NVIDIA's own DGX Cloud infrastructure.

As articulated by Manuvir Das, "AI Workbench offers a systematic approach to encapsulating your AI initiatives, enabling seamless migration across different environments." Developers can consolidate their models, frameworks, SDKs, and libraries from open-source repositories and the NVIDIA AI platform into a unified domain. Subsequently, they can initiate, assess, and refine generative AI products on an RTX PC or workstation, and effortlessly scale to data center and cloud computing resources when required.

Joey Zwicker, Vice President of AI Strategy at HPE, commended the potential of NVIDIA AI Workbench, emphasizing its capacity to streamline the development of generative AI projects, from inception to one-click training and deployment.

AI Workbench is slated for release in the fall of 2023 and will be offered at no additional cost as part of other product subscriptions, including AI Enterprise.

Empowering Enterprise with New RTX Workstations and GPUs for Generative AI

On the hardware front, NVIDIA introduced a novel lineup of RTX workstations, fortified with RTX 6000 GPUs and infused with enterprise-grade AI support. These workstations are meticulously engineered to meet the formidable GPU demands inherent in industrial digitization and enterprise 3D visualization.

The Ada workstation GPU family welcomes its latest members: the RTX 5000, RTX 4500, and RTX 4000. The RTX 5000 is currently available, while the RTX 4500 and RTX 4000 are slated for release in October and September of 2023, respectively.

In a parallel vein, NVIDIA unveiled an innovative architecture for OVX servers, each capable of accommodating up to eight L40S Ada GPUs. These servers seamlessly integrate with AI Enterprise software, constituting a compelling proposition for content creation, including AI-generated images for domains like graphic design, animation, and architecture.

Dan Stine, Director of Design Technology at architectural firm Lake|Flato, attested to the prowess of the RTX 5000 GPUs, noting their substantial performance boost and expansive frame buffer, culminating in superior virtual reality experiences that provide clients with a contextual and immersive encounter.

Embracing Innovation with Omniverse and OpenUSD for Digital Twinning

NVIDIA's Omniverse, a developmental platform that serves as a nexus for constructing and operating industrial digitalization applications in tandem with the 3D visualization standard, OpenUSD, is poised for substantial updates. This platform caters to a spectrum of domains, ranging from 3D animation and game development to automotive manufacturing. Several updates are directly linked to NVIDIA's recent collaboration with Universal Scene Description, an open-source format central to crafting objects and elements in the realm of 3D graphics.

Rev Lebaredian, Vice President of Omniverse and Simulation Technology at NVIDIA, encapsulated the essence of this evolution: "Industrial entities are racing to revolutionize their workflows through digitalization, ushering in a heightened demand for interconnected, interoperable 3D software ecosystems."

The latest version of Omniverse, currently in its beta phase, showcases a series of integrations harnessed through USD. Among these, a notable partnership emerges with Adobe's AI image generation application, Firefly.

Prominent companies such as Boston Dynamics have embraced Omniverse for industrial design, employing it to simulate robotics and control systems, thus underscoring its diverse applications.

In conclusion, the SIGGRAPH Conference of 2023 served as the stage for NVIDIA to unveil a constellation of transformative advancements across generative AI platforms, industrial digitization, GPUs, and beyond. With innovations spanning from the GH200 Grace Hopper Superchip to the collaborative potential between Hugging Face and DGX Cloud, and from the evolution of AI Enterprise to the introduction of AI Workbench, NVIDIA cements its role as a vanguard in propelling the realm of AI forward. This, coupled with the launch of new RTX workstations and GPUs, as well as the reinvigoration of Omniverse through OpenUSD integration, marks a pivotal juncture in AI and industrial technology. As these breakthroughs continue to reshape industries and pave the way for future advancements, the landscape of possibilities in the world of AI becomes more exciting than ever before.

Our first Keynote starts at 9 am! Hear from Darío Gil, Senior Vice President and Director of IBM Research, on what's next in quantum computing in West Hall B.

Get ready to experience the cutting-edge world of accelerated cloud GPU computing for machine learning, visual effects rendering, big data analytics, and cognitive computing as Genesis Cloud takes the stage at SIGGRAPH 2023. Celebrating the conference's 50th anniversary, Genesis Cloud is thrilled to showcase its expertise as a distinguished cloud computing provider for accelerators.

Make sure to mark your calendars and visit booth number 1129, where the Genesis Cloud team eagerly awaits to introduce you to their groundbreaking cloud computing solutions that are revolutionizing the VFX and rendering landscape with unparalleled efficiency, affordability, and sustainability.

Genesis Cloud: A European Cloud Infrastructure Pioneer

As a leading European cloud infrastructure provider, Genesis Cloud specializes in various workloads, including Machine Learning, Rendering, Blockchain, Transcoding, and IT Security. With a strong commitment to environmental responsibility, the company offers 100% green computational power, delivering a remarkable performance-to-cost ratio for any workload. Harness the power of cost-optimized GPUs, allowing you the flexibility to run your workloads on-demand or always-on, resulting in significant cost savings compared to on-premise deployments.

Unleashing the Full Potential of GPUs

Genesis Cloud leverages consumer-grade cards from Nvidia within a reliable high-performance environment supported by enterprise-grade data centers and server infrastructure. The company stays at the forefront of innovation, introducing commercial-grade cards and non-GPU accelerators tailored for specific needs, such as LLMs and Generative AI. Notably, the collaboration with Habana Labs, an Intel company, has led to the development of the Gaudi2 processor, expanding the horizons of possibilities for Genesis Cloud users.

The Testimony of Miguel Espada

With glowing testimonials from satisfied customers, it's evident that Genesis Cloud delivers on its promises. Miguel Espada, CEO and Co-founder of Espadaysantacruz Studio and co-founder of SpecialGuestX, shares his experience working with Genesis Cloud for the project "Carvana Joyrides." The project involved creating unique videos for each of Carvana's one million car sales using AI and cloud rendering. Thanks to Genesis Cloud's flexibility and scalability, they generated an astounding 1,345,948 personalized videos, totaling around 45,000 hours of content. This achievement showcases the true potential of cloud rendering with Genesis Cloud.

Empowering Creativity with Transparent Pricing

With Nvidia cards at their disposal, users of Genesis Cloud enjoy a seamless and user-friendly cloud-based environment. The company's fair and transparent pricing, starting at just $0.20 per hour for a dedicated GPU on-demand, along with zero data ingress/egress charges, makes it an attractive option for both personal and business use. The organizational feature also aids teams in collaborating on multiple projects with ease.

Connect with Genesis Cloud at SIGGRAPH

Genesis Cloud takes immense pride in being an active member of the community, with SIGGRAPH being their first-ever exhibition back in 2018, in partnership with AMD. Now, as part of SIGGRAPH's 50th-anniversary celebration, the Genesis Cloud team looks forward to engaging with fellow enthusiasts, experts, and visionaries in the West Hall of the Los Angeles Conference Center at booth 1129.

Don't miss this golden opportunity to unleash your creative vision's true potential. Visit the Genesis Cloud booth to learn how their high-performance, cost-effective solutions can take your projects to new heights. With a swift sign-up process and free credits, you can launch new instances in just two minutes and experience the future of cloud GPU computing firsthand. Join the Genesis Cloud revolution today!

Immerse yourself in mind-blowing VR narratives and breathtaking visual experiences that push the boundaries of storytelling and technology at our SIGGRAPH 2023 VR Theater Kiosks in Los Angeles. You also have the opportunity to meet the brilliant minds behind these groundbreaking VR experiences! Don't miss out on this extraordinary opportunity to witness the future of cinema and interact with the talented minds behind it all.

​https://s2023.siggraph.org/full-program/?filter1=sstype153&fbclid=IwAR1aVKz8wfz_D1dH9JIXAspi9kzmSECgo01BRoB8tUY5A0AiVnfc19YpAqM​