Frequently Asked Questions
The most-asked questions about nvidia and amd graphics processors.
What is a graphics processor (GPU) and what does it actually do?
A GPU is a specialized chip built to crunch enormous numbers of parallel math operations, originally for rendering images and video. Today it also drives AI inference, scientific simulation, and video encoding. NVIDIA and AMD each design these chips from the ground up, competing on architecture, driver quality, and feature sets.
Who are the two main companies in the discrete GPU market?
NVIDIA, founded in 1993 by Jensen Huang and two MIT classmates, and AMD (originally Advanced Micro Devices, founded in 1969), which entered the GPU space by acquiring ATI Technologies in 2006. They are the only two vendors shipping discrete graphics cards at scale for both consumers and data centers.
What are the current consumer product lines from each company?
NVIDIA sells its GeForce RTX series (40-series and 50-series), while AMD offers the Radeon RX line (7000-series and 9000-series). Both target the same gaming and creative-workload segments at overlapping price tiers.
What is CUDA and why does it matter to the rivalry?
CUDA is NVIDIA's proprietary parallel-computing platform that lets developers write kernels running directly on its GPU silicon. AMD's counterpart, ROCm, has been closing the gap but still trails in library breadth and community adoption, giving NVIDIA a persistent edge in AI and HPC workloads.
What are DLSS and FSR, and how do they compare?
DLSS (Deep Learning Super Sampling) is NVIDIA's AI-upscaling tech that uses dedicated tensor cores to render at a lower resolution and reconstruct a sharper frame. FSR (FidelityFX Super Resolution) is AMD's open-source alternative that runs on any GPU; both aim to boost frame rates while preserving visual quality, though DLSS generally edges ahead on fidelity.
What is ray tracing in the context of GPUs?
Ray tracing simulates how light bounces and reflects through a scene, producing more realistic shadows, reflections, and global illumination than traditional rasterization. NVIDIA popularized real-time hardware ray tracing with its RTX 20-series in 2018, and AMD followed with its RDNA 2 architecture in 2020.
Where should a newcomer start learning about GPU architecture?
A good entry point is understanding the basic building blocks: streaming multiprocessors on NVIDIA or compute units on AMD, which contain the ALUs, texture units, and cache that execute your workloads. From there, exploring how the memory hierarchy (L1/L2 cache, frame buffer) and the driver layer interact will give a solid mental model.
What are the most notable generational leaps in GPU history?
The shift from fixed-function pipelines to fully programmable shaders in the mid-2000s (GeForce 6800, Radeon X800) fundamentally changed how games were built. More recently, the introduction of hardware ray-tracing cores and dedicated AI acceleration units in 2018–2020 marked another paradigm shift in what a GPU can accomplish in real time.
Do NVIDIA and AMD GPUs work in the same systems, or are there compatibility issues?
Both use the standard PCIe interface, so a GeForce card will physically fit in the same slot as a Radeon card in any modern motherboard. The differences are software-level: driver ecosystems, API support (Vulkan, DirectX, CUDA/ROCm), and vendor-specific features like NVENC versus AMF video encoding.
What role do data-center and AI GPUs play in the rivalry?
NVIDIA dominates the AI training and inference market with its Hopper- and Blackwell-generation chips, which generate the bulk of its revenue. AMD counters with its Instinct MI300X and MI325X accelerators, competing on price and open-ecosystem appeal, though it holds a much smaller share of the AI hardware market.