GTC 2024 preview
Next week, NVIDIA CEO Jensen Huang will take the stage at the SAP Center in San Jose to showcase the company's vision for the next year, and likely (see below) a few announcements too.
Expectations are sky-high, like NVIDIA's stock price, which is up ~278% since GTC 2023.
@March 13, 2024
WHAT TO EXPECT
New Chips
NVIDIA is highly anticipated to unveil its next-generation GPU architecture, codenamed "Blackwell". These GPUs are expected to power data centers, supercomputers, and high-end workstations, promising a significant jump in AI processing power, energy efficiency, and performance. Early leaks suggest the chip will be called B100 (the H100 was announced at GTC 2022, and the A100 at GTC 2020). Expected to be built on TSMC's N3 (3nm) process, the chip is rumored to be a 1000W part!
NVIDIA's consumer graphics line will also get an architectural update to a new RTX50-series. Again, on a TSMC N3 node, with rumors it will have bleeding edge GDDR7 memory, and potentially up to 512-bit memory bus. I would expect updates to their DLSS (deep learning super sampling) technology as well.
We may in fact see some other products too;
next-gen SuperChip (GB200)
next-gen NVLink (that would be NVLink 5)
next-gen AI+graphics chip (B40)
next-gen networking @ 800G and beyond (Quantum3, Bluefield4 and Spectrum5
Leak from late 2023
In addition to the cool hardware, a lot of the conference will be discussions around CUDA, developer tooling updates, partner and customer use cases, and hopefully visual computing (Omniverse).
Advancements in AI
Generative AI Innovations: Expect a strong focus on generative AI tools and frameworks for tasks like image and text creation, code generation, and more. NVIDIA may introduce new software that makes it easier for developers to work with this disruptive technology.
Large Language Models (LLMs): NVIDIA is deeply involved in large language model development. Announcements, partnerships, or new tools to optimize and deploy LLMs are probable.
Industry-Specific AI Tools NVIDIA will showcase AI applications across various industries like healthcare, robotics, autonomous vehicles, and more.
Software Innovations
CUDA Updates: CUDA is NVIDIA's core programming platform for parallel computing. We should expect updates and new features to make coding for NVIDIA GPUs even more powerful and accessible. With AMD threatening with very competent silicon (MI300X), NVIDIA's moat is increasingly important,
Omniverse Expansion: Omniverse, NVIDIA's platform for virtual collaboration and real-time simulation, will likely see significant announcements. New features, integrations, and industry use-cases are possibilities.
Cloud Computing & Developer Tools: NVIDIA is heavily invested in cloud computing. Expect advancements in cloud-based AI development platforms and enhanced tools to simplify the developer experience.
This will be my first time at GTC, if you are headed and want to meet up, reach out!