Video Cards: Functions, Specifications, and Connections

1. The Video Card and the GPU

First, let's clarify two terms that are often used interchangeably but have distinct meanings: video card (or graphics card) and GPU.

As the article explains:

• The Graphics Processing Unit (GPU) is the specialized processor—the "brain"—designed for graphics-related calculations. It's a single chip.
• The video card (or graphics card) is the entire add-in board that you install into a PCIe slot. It houses the GPU, its own dedicated memory (VRAM), display output ports (like HDMI), and a cooling solution (heatsink and fans).

You will also encounter Integrated GPUs (iGPUs), which are not on a separate card. Instead, they are built directly into the same package as the CPU. These are common in laptops and budget desktops where high-end graphics performance is not a priority. Our focus today will be on discrete video cards, which offer significantly more power.

2. The Function of a GPU: The Power of Parallelism

Why do we need a separate, powerful processor just for graphics? It comes down to a fundamental difference in architecture compared to a CPU. A CPU is designed for serial processing—handling a wide variety of complex tasks one after another very quickly. A GPU, on the other hand, is a master of parallel processing.

The following video uses a great analogy to illustrate this core difference.

As the video highlights, a GPU contains thousands of smaller, simpler cores. This design is perfect for tasks that can be broken down into many identical, simultaneous calculations—a concept known as "embarrassingly parallel." Rendering a 3D scene, which involves calculating the color and position of millions of pixels, is a perfect example.

This massive parallel processing capability is also why GPUs have become crucial for scientific computing and data analysis, accelerating tasks in machine learning and complex simulations far beyond what a CPU alone could achieve.

3. Key GPU and Video Card Specifications

When you look at a video card's product page, you'll see a list of technical specifications. Let's break down the most important ones.

a) Processing Cores and Architecture

The GPU itself is the most important component. The two main manufacturers are NVIDIA (with its GeForce line) and AMD (with its Radeon line). Each generation of GPUs has a specific architecture (e.g., NVIDIA's "Ada Lovelace" or AMD's "RDNA 3").

Within the GPU are the cores that do the work. The Branch Education video gives a detailed look at this internal structure.

• CUDA Cores (NVIDIA) / Stream Processors (AMD): These are the fundamental, general-purpose cores of the GPU. A higher number generally indicates more processing power.
• Ray Tracing (RT) Cores: Specialized cores designed to accelerate the complex calculations needed for realistic lighting and reflections (ray tracing).
• Tensor Cores (NVIDIA) / AI Accelerators (AMD): Specialized cores that accelerate matrix operations, which are fundamental to AI and machine learning workloads.

b) Clock Speed

Measured in Megahertz (MHz) or Gigahertz (GHz), clock speed indicates how many cycles the GPU cores can execute per second. You'll often see two numbers:

• Base Clock: The guaranteed minimum speed.
• Boost Clock: The maximum speed the GPU will automatically reach under load, as long as it stays within its power and temperature limits. This dynamic adjustment is a feature often called GPU Boost (NVIDIA) or similar names by AMD.

c) Video RAM (VRAM)

Just like a CPU needs system RAM, a GPU needs its own high-speed memory, called VRAM, to store textures, models, and other data it needs for rendering.

Key VRAM specifications include:

• Amount: Measured in gigabytes (GB), e.g., 8 GB, 16 GB. More VRAM is needed for higher resolutions (like 4K) and high-quality textures in games.
• Type: This indicates the generation of the memory technology, such as GDDR6 or the faster GDDR6X.
• Memory Bus Width: Measured in bits (e.g., 128-bit, 256-bit), this is the "width" of the road between the GPU and the VRAM. A wider bus allows more data to be transferred simultaneously.
• Bandwidth: The total data throughput of the memory, measured in gigabytes per second (GB/s). It's a product of the memory speed and bus width.

d) Power and Performance Limits

Modern video cards are powerful, but that performance is managed within strict thermal and power envelopes. Understanding this is key to interpreting performance.

The key takeaway here is that a video card's performance is a dynamic balance of clock speed, power consumption, and temperature. The card will automatically push itself to the highest possible clock speed until it hits one of its predefined limits.

4. Connection Types

A video card interacts with the rest of the system through three types of connections.

a) Motherboard Interface

As we've discussed, a discrete video card plugs into a PCIe x16 slot on the motherboard. This provides both a physical connection and a high-bandwidth data path to the CPU and system memory. While there are different PCIe generations (Gen 3, 4, 5), they are backward and forward compatible. A new Gen 5 card will work in a Gen 4 slot, just at Gen 4 speeds. For most current applications, the performance difference between generations is minimal.

b) Power Connectors

The PCIe slot can provide up to 75 watts of power. However, high-performance video cards need much more than that. This additional power is supplied directly from the PC's power supply unit (PSU) via dedicated PCIe power connectors.

Common connectors include:

• 6-pin: Can supply up to 75 watts.
• 8-pin (or 6+2 pin): Can supply up to 150 watts.
• 12VHPWR (12+4 pin): A newer standard found on very high-end NVIDIA cards, capable of delivering up to 600 watts over a single cable.

A card may have one or a combination of these connectors.

c) Display Connectors

These are the ports on the back of the card where you plug in your monitor(s). The two modern standards are:

• HDMI (High-Definition Multimedia Interface): Very common on TVs and monitors.
• DisplayPort: The most capable standard for PC monitors, typically supporting higher resolutions and refresh rates than HDMI.

Conclusion

You now have a solid understanding of the role and characteristics of a modern video card. This component is a highly specialized parallel-processing powerhouse that is critical for graphics-intensive applications and increasingly important for data science.

Here are the key takeaways from today's lesson:

• A video card is an expansion board containing a GPU, VRAM, and a cooling system.
• GPUs use a massively parallel architecture with thousands of cores, making them ideal for tasks like rendering 3D graphics and AI calculations.
• Key specifications include the GPU model, the number and type of processing cores, clock speed, and the amount and type of VRAM.
• Video cards connect to the motherboard via a PCIe x16 slot, draw auxiliary power through 6-pin, 8-pin, or 12VHPWR connectors, and output to monitors via HDMI or DisplayPort.

In our next lesson, we will focus on the component responsible for powering the video card and everything else in the system. We will identify power supply specifications including wattage, efficiency ratings, and modular designs.