South Korea plans on taking on global leader Nvidia in the AI chips market. Here’s how

South Korea plans on taking on global leader Nvidia in the AI chips market. Here’s how. Source: Nvidia

World largest GPU to power tomorrow’s AI tech today

GPUs are used to source out processing power for graphic based workloads, putting significantly less pressure on the Central Processing Unit (CPU). This speeds up image rendering, making output on display devices smoother.

With artificial intelligence (AI), machine learning, and other emerging technologies on the rise – and the need for more powerful GPUs growing, hardware companies are racing to meet demand (and raising customer expectations).

Recently, Nvidia announced the “world’s largest GPU” at its annual GPU Technology Conference.

Their latest offering, dubbed the DGX-2, is a GPU supercomputer, fitted with 16 Tesla V100 32GB GPUs, which gives you a total of 81.920 CUDA Cores and 2000 Teraflops of Tensor Cores. Each of the 16 GPUs comes with 32GB of system memory, totaling a cool 512GB of HBM2 memory clocking in at 14.4TB/sec.

To put that in perspective, the “machine” enables you to transfer 1440 HD movies per second.

The whole system is powered by two Intel Xeon platinum CPUs, with 1.5TB system memory and 30TB NVMe SSDs internal storage.

In short, that’s a lot of power and it’s what will help make industrial grade AI, deep learning algorithms, and real-time big data analytics viable in the future.

This “monstrous” GPU (which, in fact, is a whole system on its own) runs on 10,000 watts of energy, which powers a 2 Petaflop brain. Around 10 times faster than its previous generation DGX-1, the whole setup weighs almost 160 kilos and costs about US$400,000.

In essence, the DGX-2 is the only commercially available machine that claims to power any AI innovation in the market (or in the lab).

The recent boom in cryptocurrencies has caused the prices of GPU to skyrocket as these processors could also power cryto-mining workstations.

However, Nvidia is keen on working with partners to create AI-based applications.

Recently, for example, Nvidia demoed a cloud supercomputing technology it calls CLARA. Targeted at doctors and medical professionals, the system is capable of transforming imaging capabilities. It can simulate a full color, 3D image with blood pressure and other diagnostics using two-dimensional ultrasound image from 15 years ago.