NVIDIA (NVDA 2.81%) held its annual GPU Technology Conference (GTC 2018) between March 25 and 29, where it unveiled a lot of new products, chips, and technologies. Let's take a closer look at the chipmaker's six biggest revelations.
1. A medical supercomputer
NVIDIA unveiled Project Clara, a "medical imaging supercomputer" powered by the chipmaker's vGPUs (virtual GPUs) in the cloud. Clara's use of virtual horsepower makes the platform -- which connects to a wide variety of medical instruments -- easily scalable.
NVIDIA founder and CEO Jensen Huang. Image source: NVIDIA.
NVIDIA believes Project Clara will modernize older imaging scanners, which can't be updated with the latest technologies. Those legacy devices, which have limited AI and image processing capabilities, can remotely connect to Clara to gain significant performance boosts.
2. The Quadro GV100
For the professional visualization market, NVIDIA unveiled the Quadro GV100, a Volta-based card with 32GB of HBM2 memory, 5,120 CUDA cores, and 640 Tensor cores for deep learning purposes. That's a big jump from last year's Quadro GP100, which had 16GB of HBM2 memory, 3,584 CUDA cores, and no Tensor cores.
A single card will cost $9,000. Two cards can be joined with NVLink 2, which is similar to its SLI feature for mainstream cards. NVIDIA plans to use the Quadro GV100 to showcase its RTX technology, which introduces real-time ray-tracing for media applications. NVIDIA claims that the Quadro V100 can "do complex renders up to 10x faster than with a CPU alone," and that workstation users "can create interactive, photoreal visualizations of massive 3D models -- all in real time."
Image source NVIDIA.
3. The DGX-2
Last year, NVIDIA introduced the DGX-1 supercomputer, which put the power of "400 servers" in a single box. That $149,000 box is powered by eight Volta-based Tesla GPUs and two 20-core Intel (INTC 1.04%) Xeon processors.
NVIDIA unveiled its big brother, the DGX-2, at GTC. The $399,000 rack is powered by 16 Volta-based Tesla GPUs, which gives it twice the number of CUDA cores at the DGX-1, and two Xeon CPUs. It's also equipped with NVIDIA's suite of deep learning software for data scientists.
4. The new Tesla V100
NVIDIA also upgraded its Tesla V100 GPUs from 16GB of HBM2 memory to 32GB. That upgrade could strengthen NVIDIA's data center business, which more than doubled its revenues annually last quarter. It could also exacerbate the current shortage of memory chips worldwide.
Image source: NVIDIA.
Several major customers -- including IBM, Hewlett-Packard Enterprise, and Dell EMC -- plan to install the new Tesla V100 in their data centers during the second quarter. Tom's Hardware expects the units to "fly off shelves as the data center continues its transition to AI-centric architectures."
5. New Drive computers
On the automotive front, NVIDIA showcased Drive Pegasus, its newest and most powerful version of the Drive PX supercomputer for autonomous cars. The Pegasus is powered by two Tesla V100s and two Xavier Tegra systems on a chip (or SoCs), which crunch 320 trillion operations per second to analyze the constant flow of data from 360-degree cameras and LIDAR systems. The Pegasus, which remains in development, is aimed at level 5 (fully autonomous) vehicles.
NVIDIA also showcased Pegasus' little brother, Drive Xavier, which handles 30 trillion operations per second with a custom eight-core ARM64 CPU and a Volta GPU. NVIDIA claims that the Xavier is the "world's most powerful SoC," and will also power level 5 driverless vehicles.
6. A VR-driven vehicle
NVIDIA is commonly associated with driverless cars, but the chipmaker introduced a brand new approach to vehicles with a VR-driven car. During the conference, NVIDIA synchronized its Holodeck software, which places users in VR environments, with an HTC Vive VR headset, driving simulation controls, and a Ford Fusion parked behind the convention center.
NVIDIA's Holodeck. Image source: NVIDIA.
During the demo, a driver in a VR headset entered the "virtual car" on the Holodeck and drove the empty Fusion around the parking lot. NVIDIA CEO Jensen Huang admitted that the demo was merely a proof of concept, but the technology has fascinating applications for any industry that could benefit from VR-controlled machines.
The bottom line
NVIDIA's GTC announcements reinforce its reputation as the chipmaker of the future. These new products -- which focus on cloud-based computing, professional visualization, AI, cars, and automation -- could propel the stock to even higher levels over the next few years.
