NCA-GENM Question 132
Select 3A company plans to deploy a generative AI multimodal model to support both text-to-image generation and video generation workloads. Their primary goal is to ensure optimal performance while meeting user demand for high-resolution outputs. Which combination of system components would best meet their needs?
- A
High-performance GPUs with Tensor Cores optimized for AI workloads
- B
A large, high-speed SSD storage system to manage model weights and datasets
- C
A CPU with many cores but no specific AI optimizations
- D
A high-bandwidth interconnect like NVIDIA NVLink for scaling multiple GPUs
- E
A low-power edge device with limited computational capacity
Show answer and explanation
Correct answers: A, B, D
Explanation
To meet the user needs for high-resolution outputs in text-to-image and video generation tasks, the system must include components that are optimized for AI workloads. High-performance GPUs with Tensor Cores provide the necessary compute power, high-speed SSD storage ensures data can be accessed efficiently, and a high-bandwidth interconnect like NVLink scales GPU performance across multiple devices. These components together ensure optimal performance and scalability for demanding multimodal generative AI tasks.
- A. Correct.
High-performance GPUs with Tensor Cores are essential for accelerating the heavy computation required for generative AI multimodal models, especially for tasks like text-to-image and video generation.
- B. Correct.
A large, high-speed SSD storage system ensures that the model weights, datasets, and intermediate outputs can be accessed and processed efficiently, which is critical for high-resolution output.
- C. Incorrect.
A CPU with many cores but no AI optimizations would not significantly enhance performance for AI workloads, as these tasks rely heavily on GPU acceleration.
- D. Correct.
A high-bandwidth interconnect like NVIDIA NVLink allows multiple GPUs to work together efficiently, enabling scalability and improved performance for resource-intensive workloads.
- E. Incorrect.
A low-power edge device with limited computational capacity would not be suitable for such demanding workloads, as it lacks the necessary performance and scalability for high-resolution outputs.