[Event] Nvidia GTC 2022 keynote | Hopper | di. 22 mrt 16:00

[Balance]

Nvidia GTC 2022 keynote Dinsdag 22 maart 2022 vanaf 16:00 Inleiding Nvidia's opening keynote van de GTC 2022 conferentie vindt vandaag plaats om 16:00 Nederlandse tijd. In dit topic can discussie plaatsvinden omtrent het event, slow-chat is toegestaan! NVIDIA CEO Jensen Huang kicks off #GTC22 with a keynote that presents the latest breakthroughs in #AI, data science, high performance computing, graphics, edge computing, networking, and autonomous machines. Deep dive into announcements and discover more content at https://www.nvidia.com/gtc Nvidia GH100-gpu Een nieuwe GPU-architectuur die Ampere op moet volgen wordt verwacht, alsmede de eerste GPU met die architectuur, de GH100. Die lanceert waarschijnlijk in de Nvidia H100-kaart voor servers en HPC. Such as the fact that GH100 is a large monolithic GPU with six Graphics Processing Clusters. In fact, we can also count a total of 144 Streaming Multiprocessors. As reported earlier today, one of the GPCs is different, because it also carries the render engine and supports graphics. Furthermore, the chip supports up to 6 stacks of High-Bandwidth-Memory, unfortunately since the full specs are not yet available, we are not sure if all 6 stacks are used. Nvidia DGX H100 Station & SuperPod Een nieuw workstation desktop wordt ook verwacht, de DGX H100 Station. Het huidige DGX Station A100 bevat vier A100-gpus, de opvolger zou dus vier H100 GPUs kunnen krijgen. Daarnaast lijkt een nieuwe DGX SuperPod server met 20+ H100 GPUs ook in de pijpleiding te zitten. NVIDIA will be rolling out a number of products based on GH100 GPU, such an SXM based H100 card for DGX mainboard, a DGX H100 station and even a DGX H100 SuperPod. More importantly, NVIDIA is also announcing PCIe-based H100 model at the same time.

[Balance]

Nvidia H100 specs

• TSMC N4 process
• 80 miljard transistoren
• HBM3 geheugen met 3 TB/s bandbreedte
• PCIe 5.0
• 700 watt TDP

[Balance]

Nvidia blogs:

• H100 Transformer Engine Supercharges AI Training, Delivering Up to 6x Higher Performance Without Losing Accuracy
• NVIDIA Hopper GPU Architecture Accelerates Dynamic Programming Up to 40x Using New DPX Instructions