Observations from GTC’19

by

March 27, 2019

OmniScale Media was at GTC’19 last week to see all of the interesting current and future technology and applications being showcased at the event. As a technology marketing company with a growing roster of clients relevant to artificial intelligence (and other forms of advanced computing), an event like GTC is great for us to reconnect with old friends in the industry, make new friends and business partners, and learn about how the technologies that we are involved with marketing are shaping the way forward. The event was successful on all fronts (we wish we could have seen even more of our old friends – we missed a few of you!), and there was plenty to take away from the event.

Below are some of my takeaways from the confab:

NVIDIA has a great read on the future – What a fresh insight! NVIDIA knows what they’re doing. Okay, that’s not a ground-breaking insight, but it must be said because very interesting things happened in March around this event. It started when NVIDIA announced that it had bought Mellanox for $6.9 billion. Our friend, Tiffany Trader at HPCwire reported on it here, and as you can see from the quotes by NVIDIA’s CEO Jensen Huang, a big reason they did it is because they understand that data centers are in a period of metamorphosis as the need for compute gets pushed to the edges of the network, where the data is being generated. More on that in the next point, but when you consider that autonomous vehicles generate 4,000 GB (4TB) of data each day that needs processing out in the wild, it’s easy to see how data centers will be affected.

Just as a point of interest, in Tiffany’s article, you’ll see a quote from NVIDIA’s Huang about their relationship with Intel: “people think we’re antagonistic, but its just not true.” We found that quote to be interesting, particularly when hours before Huang took the stage for his always anticipated keynote, Intel released big news about the 2021 delivery of Aurora, the first exascale supercomputer in the United States, stealing headlines as NVIDIA kicked off its big event. Some could reasonably see this as “antagonistic.” We were amused. (By the way, be sure to check out Nicole Hemsoth’s article about the Aurora, its evolution and significance here.)

If we heard criticism on the show floor, it was that many people thought Huang spent too little time in his keynote on the Mellanox acquisition and its significance, and that there was “nothing really all that new” announced in the presentation – not that we necessarily agree, but this is the buzz we caught repeatedly as we traveled the conference.

If we heard criticism on the show floor, it was that many people thought Huang spent too little time in his keynote on the Mellanox acquisition and its significance, and that there was “nothing really all that new” announced in the presentation – not that we necessarily agree, but this is the buzz we caught repeatedly as we traveled the conference.

Edge Computing is the Next Wave – We touched on it in the previous observation, but this development deserves attention because it’s only going to get more significant as time marches forward. There was a lot of talk about the need for crunching data closer to where it is being generated. We saw wireless sensors with GPU processors, robots collecting video data, autonomous vehicles (including a self-driving semi-truck with its own datacenter in the cabin). Pushing the network to the edges was a clear theme at the show.

One of our own clients, ScaleMatrix, participated in news “at the edge” by announcing the acquisition of Instant Data Centers. Rich Miller at Data Center Frontier did a great write-up on this news and why it’s so significant, but to summarize it in our own words, we are told that when AI is being trained, PCI cards are running at max capacity. The more cards there are, the faster the AI can train, so organizations that are looking for first mover competitive advantages want to load up their racks, which creates more heat – and the need for better cooling solutions. ScaleMatrix, through their DDC Cabinet Technology, has figured out modular scalable cooling, with the ability to cram 52 kW of power density into a single cabinet. It’s no wonder that they are getting so much attention from some of the biggest leaders in AI, like Nvidia (DGX-Ready Partnership) and HPE (Center of Excellence Partnership). Michael Feldman at The Next Platform did another great write-up that gives the full picture. Anyone involved with AI training should consider giving ScaleMatrix a serious evaluation.

 

Autonomous Semi-Truck – A self-driving hauler with a data center in the cabin. 

Autonomous Semi-Truck – A self-driving hauler with a data center in the cabin.

Autonomous X is much closer than I even realized – You name it, if it can be automated, it’s coming! Having worked around advanced computing for the last 20 years, I’m not a stranger to the idea of automation. But I can admit that I have been skeptical to a large degree as to how close this all really is. My tune is changing. It’s not that automation is coming – it’s here. We saw demonstrations of completely automated factory floors, the AGROBOT Robotic Strawberry Harvester, robotic Arms that put items on robotic serving bots that can navigate busy areas, robotic inspection droids that can climb metal walls and inspect hard to reach areas. It’s one thing to read about this stuff – its another thing to see the working demonstrations at an event like this and realize just how close the automated industrial revolution (Industrial Revolution 4.0, anyone?) actually is. (By the way, if you want to see some of the coolest stuff going on in AI and automation, check out Venture Beat’s Kyle Wiggers twitter feed at @Kyle_L_Wiggers. He’s got his finger on the pulse and is a great read).

 

 

Blockchain? Never heard of it… For all the hype about blockchain in 2018, it was virtually nowhere to be seen on the GTC show floor – not even around cryptocurrency mining, where GPUs have become a runaway cottage industry. If blockchain was there, we missed it, and we would have loved to have seen it because we believe that for an AI future, trustless, open-source ledgers will be necessary. We know that data governance is a big issue where “Big Data” is concerned, and that while blockchain and similar distributed ledger technologies won’t necessarily be required for every automated function, there will be plenty where having a trustless, immutable ledger will add value (or even be necessary) for accuracy and security. I’m not exactly sure what this absence says about either the maturity or feasibility of blockchain as a solution, but anecdotally, it has been our experience that people working on high end computing applications aren’t really taking blockchain seriously at this point – (a friend of OmniScale’s working on HPC in genomics once commented that blockchain is “magical math tokens,” which we found hilarious, though we are still believers.)

We personally believe that blockchain-related technologies will have a big role to play in the future in everything from economics, to healthcare, computing and beyond, but we also understand that the technology has its kinks and has a lot to prove still.

Virtual Reality will have surprising applications – There was a lot of virtual reality on display on the conference floor, which is not surprising at all. I went to my first Supercomputing show in Dallas in November 2000, and I can remember seeing a virtual reality display way back then. What’s changed over the years is that there are increasingly more applications for virtual reality that add value. Whether it was VR for automotive demos to help train automated vehicles, or VR for training humans (and even accessing data), there is clearly more and more application for this technology being employed.

 

One of the neat things we saw was was a worm-like robot from Sarcos Robotics that had magnetic tracks to climb oil derricks, or through small accesses and even pipelines. Right now the robots are being remotely controlled, but we were told that they will soon be automated. The VR application is that as the robot collects video data from all the locations it surveys, a human can put on a VR helmet and travel down these corridors to examine obstructions, evaluate structure, and audit safety. While this application may be obvious to some out there, its one of those applications that I hadn’t given much thought to, and it makes me consider what other industrial uses of VR are out there that will surprise us.

Isaac Lopez is a Co-Founder of OmniScale Media and an enthusiast for everything high-tech having been around marketing for advanced computing and applications for 20 years. High tech companies looking for marketing support can contact us here.

Pin It on Pinterest