NVIDIA-Certified Professional Agentic AI (NCP-AAI) Practice Tests

Explanation

NVIDIA GPU Operator is a Kubernetes operator that automates the deployment and management of GPU drivers, container runtime, device plugins, and monitoring components across GPU nodes. It is deployed via Helm chart and manages the entire GPU software stack. NVIDIA Container Toolkit is a component installed by the GPU Operator, not deployed separately. NVIDIA Device Plugin is also managed by the GPU Operator. NVIDIA DCGM Exporter provides metrics but does not handle driver and runtime installation.

Explanation

RAPIDS cuDF provides up to 150x speedup over pandas for GPU-accelerated DataFrame operations including groupby aggregations on large datasets. The GPU parallelism and optimized kernels enable massive acceleration for data manipulation tasks. 2-5x would be typical CPU optimization gains. 10-50x is the typical speedup for cuML machine learning algorithms. 500-1000x exceeds realistic cuDF speedups and would be more typical of cuGraph on specific graph algorithms compared to NetworkX.

Explanation

NIM_LOW_MEMORY_MODE=1 enables disk offloading to reduce GPU memory footprint when deploying large models on memory-constrained systems. This allows the model to run by swapping less-frequently-used weights to disk. NIM_KVCACHE_PERCENT controls the percentage of GPU memory allocated to KV cache but does not enable disk offloading. NIM_RELAX_MEM_CONSTRAINTS overrides memory checks on WSL2 but does not enable disk offloading. NIM_SCHEDULER_POLICY controls batching behavior, not memory management.

Explanation

The NVIDIA High-Bandwidth Interface (NV-HBI) provides 10 TB/s of bidirectional bandwidth between the two GB100 dies in the B200 package, enabling them to act as a large monolithic silicon with full cache coherency and shared L2 cache. This interconnect is based on NVLink protocol but is distinct from inter-GPU NVLink. 900 GB/s is the NVLink-C2C bandwidth in GH200. 1.8 TB/s is the 5th generation NVLink bandwidth between separate B200 GPUs. 8.0 TB/s is the HBM3e memory bandwidth in B200.

Explanation

Hopper architecture (H100) introduced the Transformer Engine which dynamically switches between FP8 and FP16 precision to optimize transformer model training and inference. This enables up to 2x speedup over FP16 while maintaining accuracy. Ampere (A100) supports TF32 and FP16 but not FP8 or Transformer Engine. Ada Lovelace supports FP8 but lacks the Transformer Engine. Blackwell extends these capabilities but Hopper was the first to introduce the Transformer Engine.