NCA Data Center GPU Solutions 2 — Questions and Answers
Question 1: Which NVIDIA interconnect technology enables high-bandwidth GPU-to-GPU communication within a single server node?
- InfiniBand HDR
- NVLink (Correct answer)
- PCIe Gen 4
- RoCE v2
Correct answer: NVLink
NVLink is NVIDIA's proprietary high-speed interconnect that allows GPUs within the same node to share memory and data at much higher bandwidth than PCIe.
Question 2: What is the primary role of NVIDIA's NVSwitch in large-scale GPU deployments?
- Provides power regulation for GPU clusters
- Acts as a high-bandwidth switch enabling all-to-all NVLink connectivity among multiple GPUs (Correct answer)
- Manages GPU firmware updates across the data center
- Offloads network packet processing from GPUs
Correct answer: Acts as a high-bandwidth switch enabling all-to-all NVLink connectivity among multiple GPUs
NVSwitch is a high-speed switch chip that connects multiple GPUs via NVLink, enabling full all-to-all bandwidth across all GPUs in a node like the DGX H100.
Question 3: A data center operator needs to maximize GPU utilization for multiple tenants running different workloads simultaneously. Which NVIDIA technology is most appropriate?
- CUDA MPS
- Multi-Instance GPU (MIG) (Correct answer)
- GPU Direct Storage
- NVLink Bridge
Correct answer: Multi-Instance GPU (MIG)
MIG partitions a single GPU into multiple isolated instances, each with dedicated compute, memory, and bandwidth, enabling safe multi-tenant workloads.
Question 4: Which NVIDIA data center GPU architecture introduced the Transformer Engine for accelerating large language model training?
- Volta (V100)
- Ampere (A100)
- Hopper (H100) (Correct answer)
- Turing (T4)
Correct answer: Hopper (H100)
The Hopper architecture (H100) introduced the Transformer Engine, which dynamically applies FP8 and FP16 precision to accelerate transformer-based model training.
Question 5: In an NVIDIA DGX H100 system, how many H100 GPUs are interconnected via NVLink?
- 4
- 8 (Correct answer)
- 16
- 32
Correct answer: 8
The DGX H100 contains 8 H100 SXM5 GPUs fully interconnected via NVLink 4.0 and NVSwitch for a combined 640 GB of HBM3 memory.
Question 6: What memory technology do NVIDIA's H100 and A100 GPUs use to achieve high bandwidth in data center workloads?
- GDDR6X
- LPDDR5
- HBM2e / HBM3 (Correct answer)
- DDR5 ECC
Correct answer: HBM2e / HBM3
A100 uses HBM2e and H100 uses HBM3, both of which provide much higher memory bandwidth than GDDR solutions used in consumer GPUs.
Question 7: Which NVIDIA software platform provides a unified ecosystem of GPU-accelerated libraries, frameworks, and tools for data center AI workloads?
- CUDA-X AI (Correct answer)
- NVIDIA Omniverse
- NVIDIA Fleet Command
- NVIDIA RAPIDS only
Correct answer: CUDA-X AI
CUDA-X AI is NVIDIA's collection of GPU-accelerated libraries and SDKs (cuDNN, cuBLAS, TensorRT, etc.) that power AI and HPC workloads across data centers.
NCA Data Center GPU Solutions 2
Which NVIDIA interconnect technology enables high-bandwidth GPU-to-GPU communication within a single server node?