# NVIDIA Showed Me Their Supercomputer Source: https://www.youtube.com/watch?v=GLGK0GKexds Recap page: https://rapidrecap.app/video/GLGK0GKexds Generated: 2025-12-27T18:03:30.492+00:00 --- ## Quick Overview NVIDIA's Nyx DGX Supercomputer, located at their Santa Clara headquarters, houses 1,192 NVIDIA Blackwell B200 GPUs totaling over 1.4 PetaWatts of power consumption, utilizing sophisticated Cold Air Containment (CAC) cooling and high-density fiber optic cabling to support AI research and internal production workloads. **Key Points:** - The NVIDIA Nyx DGX Supercomputer cluster contains 1,192 NVIDIA Blackwell B200 GPUs, with each GPU rated for up to 1,200W, resulting in a total power consumption exceeding 1,430,400W (1.43 PetaWatts). - The facility uses Cold Air Containment (CAC) units with perforated floor tiles to actively push cold air up through the racks, which is crucial given the massive heat output. - The infrastructure features extensive fiber optic cabling, strategically positioned to minimize length and maintain signal integrity, which is vital for the NVLink fabric networking. - NVIDIA's internal production and R&D teams use this supercomputer for tasks ranging from AI research to testing new chip designs, often running workloads continuously. - The cooling system is highly engineered, evidenced by the 102 dB reading outside the sealed containment room, necessitating hearing protection (PPE required beyond the point). - The presenter held an SXM interface module for the Blackwell GPU, noting its 192GB HBM3e memory and powerful 1200W-1400W thermal design power, which requires the dense liquid/air cooling solutions demonstrated. - The video concludes with a discussion of DLSS evolution, noting that newer Transformer models are replacing older CNN/RNN models for better image reconstruction fidelity compared to native rendering (TAA). ![Screenshot at 00:17: 1892 NVIDIA Blackwell B200 GPUs are shown inside the supercomputer room, highlighting the scale of the 1,192 GPU cluster that powers the system.](https://ss.rapidrecap.app/screens/GLGK0GKexds/00-00-17.jpg) **Context:** The video features a tour and technical deep dive into NVIDIA's Nyx DGX Supercomputer cluster located at their Santa Clara headquarters. The host examines the sheer scale of the system, focusing on its massive computational power, the specialized liquid/air cooling infrastructure required to manage the heat generated by the new Blackwell B200 GPUs, and the advanced networking that connects the compute nodes. The video also contrasts the performance improvements of NVIDIA's DLSS technology (moving from CNN to Transformer models) with traditional rendering techniques. ## Detailed Analysis The video provides an in-depth look at NVIDIA's Nyx DGX Supercomputer, which is situated in Santa Clara, California. The host first introduces the massive scale, revealing that the cluster contains 1,192 NVIDIA Blackwell B200 GPUs. Each GPU is capable of consuming up to 1,200W, leading to a staggering total power draw exceeding 1.43 Megawatts (1,430,400W). Managing this heat requires advanced cooling infrastructure, specifically Cold Air Containment (CAC) units, where cold air is forced up through perforated floor tiles into the hot aisles. The noise level outside the containment doors reaches 102 dB, requiring hearing protection (PPE). The system relies on extensive fiber optic cabling for high-speed NVLink fabric networking between the DGX units. The presenter shows a specific SXM interface module for the Blackwell GPU, noting its 192GB HBM3e memory and high power draw (up to 1,400W per card). The facility is used for both R&D and internal production, with nodes frequently reconfigured. Finally, the video transitions to a discussion of DLSS technology, showing comparisons where newer Transformer-based models achieve fidelity closer to native rendering than older CNN-based models, illustrating the ongoing AI advancement in graphics. ### Supercomputer Introduction & Scale - NVIDIA Nyx DGX Supercomputer located at HQ in Santa Clara - Contains 1,192 Blackwell B200 GPUs - Total power consumption exceeds 1.43MW (1,430,400W) based on 1,192 x 1,200W per GPU. ### Cooling and Infrastructure - Utilizes Cold Air Containment (CAC) with cold air supplied through perforated raised floor tiles - Noise levels outside containment reach 102 dB, requiring hearing protection - Extensive fiber optic cabling supports the NVLink fabric. ### GPU Hardware Details - Host holds an SXM interface module for the Blackwell GPU - Each GPU offers 192GB of HBM3e memory - Individual GPUs can draw up to 1,400W. ### Operational Use - The cluster is used for both R&D and NVIDIA Internal Production, with nodes frequently reconfigured for validation of new chips or server designs. ### DLSS Technology Evolution - Discusses DLSS evolution, showing that newer Transformer models (DLSS 2.0+) achieve image reconstruction fidelity closer to native rendering compared to older CNN models, especially in motion. ### Sponsor Message - Promotes Squarespace for building websites and domains, offering LTT viewers a free trial and 10% off. ![Screenshot at 00:09: 10:The host stands outside the NVIDIA Nyx DGX Supercomputer facility in Santa Clara.](https://ss.rapidrecap.app/screens/GLGK0GKexds/00-00-09.jpg) ![Screenshot at 00:17: 1192 NVIDIA Blackwell GPUs are shown inside the server room, emphasizing the massive scale of the compute cluster.](https://ss.rapidrecap.app/screens/GLGK0GKexds/00-00-17.jpg) ![Screenshot at 00:22: 1192 x 1,200 = 1,430,400W calculation overlay showing the total power draw of the GPUs.](https://ss.rapidrecap.app/screens/GLGK0GKexds/00-00-22.jpg) ![Screenshot at 01:46: 46:The host kneels next to a raised floor tile, pointing out the access hatches for high-density water-cooled deployments.](https://ss.rapidrecap.app/screens/GLGK0GKexds/00-01-46.jpg) ![Screenshot at 06:05: The host excitedly holds up a large, finned heat sink module from a DGX unit, highlighting its size and cooling capacity.](https://ss.rapidrecap.app/screens/GLGK0GKexds/00-06-05.jpg)