NVIDIA GTC 2024 Jensen Accredited Samsung HMB3E 12H
This week I used to be at NVIDIA GTC in San Jose, California. It was loopy. The final massive GTC was in 2019, simply as NVIDIA was finishing its Mellanox acquisition. This 12 months, that acquisition became a key part of its new DGX techniques. The larger takeaway although, was that GTC was rocking. That’s mentioned coming from a perspective of getting been to tons of of business reveals through the years. As such, we’ve got 5 key takeaways from the occasion.
AI is Rocking 5 Key Takeaways from NVIDIA GTC 2024
Listed below are the highest 5 key takeaways from GTC 2024:
One brutally obvious remark is that AI shouldn’t be part of HPC. GTC 2024 had over 17,000 individuals. Of us have been enthusiastic about SC23 hitting 14,295 attendees in Denver, and the thought was that was swelling due to AI. After all, each reveals have off-site, non-badged attendees doing briefings, exhibiting NDA merchandise, and shutting offers. On the identical time, GTC felt full, even on the FOURTH day when normally present flooring are barren. Here’s a take a look at the hallway exterior of the present flooring after it closed, and that was regular visitors to see in the course of the 4 day occasion.
NVIDIA GTC 2024 Day 4 Occasion Ending
Develop energy techniques in the present day. The NVIDIA DGX GB200 NVL72 is a mouthful, however additionally it is a robust system that makes use of 120kW. Eight of these, plus obligatory storage, switching, and so forth is probably going over 1MW of energy in a single knowledge heart aisle. Liquid cooling goes to be necessary. Now we have checked out generations now of top-end NVIDIA liquid cooled servers. Fluid, fittings, metallic blocks, and tubes are usually not the constraints on the way forward for compute. Energy is. Most prospects in the present day can work out 30kW racks. 60kW racks are doable in lots of knowledge facilities. 120kW is a gigantic leap. On the scale of many megawatts, simply sourcing sufficient low-cost and dependable energy is a world want.
NVIDIA DGX GB200 NVL72 Entrance 1
NVIDIA shouldn’t be a gaming firm. We talk about NVIDIA GPUs as graphics processing models. The heritage was, undoubtedly, the GeForce (and Riva?) strains, skilled graphics strains, and so forth. GTC 2024 talked a bit about OVX and Omniverse. The main focus was clearly on knowledge heart AI (Blackwell), networking (NVLink/ NVSwitch and ConnectX-8 Infiniband), and Grace. Robotics (Jetson and Drive) have been definitely there. The keynote, present flooring, and so forth have been dominated by knowledge heart, robotics, and automotive. GTC has at all times had a slant on this path however after two days I used to be reflecting that if I didn’t use NVIDIA GPUs repeatedly and this was my first GTC, I might have nearly no concept that NVIDIA sells playing cards for gaming after two days within the convention.
NVIDIA GTC 2024 QCT NVIDIA Grace Blackwell 5.5kW To 6kW Tray
Robotics is the darkish horse. NVIDIA may speak rather a lot about AI in the present day, however one in all its massive bets for the longer term is that the identical LLM know-how used to flood the world with textual content, photographs, and, more and more, video goes to make its approach to robotics. Encoding motion and extra into LLMs feels like a promising approach to extra effectively program robots. If we would like Rosie the Robotic from The Jetsons, or one thing much more human-like, {hardware} that may make the most of the large knowledge heart investments in LLMs is required. Jetson Thor may be a giant one when it launches within the first half of 2025. Till then, we’ve got Jetson Nano.
NVIDIA Jetson Orin Nano Developer Equipment High
NVIDIA has a wild set of interconnects. Simply holding monitor of those is difficult. For scale out there’s Infiniband. For administration, Ethernet. GPU-to-GPU is NVLink. GPU-to- Grace CPU is Chip-2-Chip (C2C). Blackwell half to Blackwell half is through NV-HBI. NVIDIA is designing an enormous variety of interconnects simply to maneuver knowledge round. That’s one thing that loads of of us neglect when accelerator FLOPS/ TFLOPS. A startup not solely must get the on-package, and in-chassis proper, but it surely wants the scale-out materials as effectively.
NVIDIA GTC 2024 Spectrum X800 Change
Closing Phrases
Everybody I spoke to at NVIDIA GTC 2024 commented about how busy it was. NVIDIA had an worker hour on the finish which was nice. There have been many inexperienced badges strolling across the present flooring and some stopped to say hello, take selfies, after which discuss what they make, typically pointing to the product.
Here’s a higher approach to put it. I spent 4 days and a night at GTC, and I by no means received to most likely 20% of the present flooring nor a couple of reveals I needed to see. Usually, I get no less than an hour to myself after conferences and already lined a lot of the flooring. Not this journey. GTC was rocking.
Earlier article8x NVIDIA Grace Hopper Superchips in a Blade HPE Cray EX254n at GTC 2024Next articleAMD EPYC 7C13 is a Surprisingly Low-cost and Good CPU Patrick has been working STH since 2009 and covers all kinds of SME, SMB, and SOHO IT matters. Patrick is a guide within the know-how business and has labored with quite a few giant {hardware} and storage distributors within the Silicon Valley. The objective of STH is just to assist customers discover some details about server, storage and networking, constructing blocks. When you have any useful info please be at liberty to submit on the boards.