If you need to prove to your console friends that your graphics card is, beyond a shadow of a doubt, better than their new Xbox One X, then you’ll need to learn all about graphics cards and what makes them tick.
Once you know all about your TeraFLOPS you can throw down knowledge on your friends until they buckle, or until you have no friends. Either way, you’ve all learnt something.
- Transistor: The transistor is the building block upon which modern technology is built. They are the tiny switches that provide the binary on/off of electricity that is the spine of computer programming.
- Nanometer: In silicon fabrication technology the transistors are now so small they’re measured in nanometers. In general terms the production scale, measured in nm, denotes the smallest transistor used in manufacturing a given chip, where - 7, 10, 14, 16, 22nm relates to transistor size the smaller the transistor, the less power it requires.
- FLOPS: The amount of floating point operations per second - useful for measure compute performance of a device. A teraflop is one million million operations per second.
- Video memory: The memory used on a graphics card to store information the GPU requires, and is working with, such as the frame buffer and textures.
- GDDR: The graphics double data rate (GDDR) is a type of memory, and the most common form, found on current generation cards. Utilises high clock speeds rather than a wider bus. Takes up a lot of space on the graphics card board. The current generations are GDDR5 and GDDR5X, with GDDR6 following.
- HBM: High bandwidth memory (HBM) has a much larger bus than GDDR memory although lower clockspeeds. Can be stacked to save space and latency and is built directly on the graphics processor itself. We are currently on the second generation, cleverly named HBM2.
- GPU: The graphics processing unit is the central silicon chip of the graphics card that drives the computational power of the graphics card.
- Cooler: The attached cooling solution, such as fans, heatsinks and heatpipes. Can be in a blower style with one fan as intake that expels air out of the case through the ports, or a open-air coolers, these fans circulate air within the case to cool graphics card components. Liquid cooling is also possible and can allow for higher clockspeeds.
- Stream processors: Roughly analogous to the cores of a processor, the stream processors in a GPU are simpler and traditionally calculate fewer operations per second. Modern GPUs have many more of these simple ‘cores’ inside them delivering their parallel processing power. Nvidia call them CUDA cores, while AMD sometimes denote them as GCN cores.
- Compute unit: These are groups of stream processors, also known as streaming multiprocessors (SMs) by Nvidia.
- GCN: AMD's Graphics Core Next architecture.
- TDP: Thermal Design Power - the amount of heat a component will output with an extensive load, under normal circumstances. Your graphics card cooler will be rated at cooling for the TDP of your graphics card.
- PCIe slot: The peripheral component interconnect express is the current high-speed slot on a motherboard a graphics card uses. Slots are rated to the amount of lanes a device plugged into them can access, with 16x being the maximum number of lanes on a third-gen slot, although graphics cards tend to have enough usable bandwidth at 8x.
- Power connectors: Most graphics cards require 6-pin, 6+2pin connectors or even multiples of either to supply them with enough extra power to function. The PCIe slot will, however, deliver 75W via the motherboard, which is why some GPUs and PCIe SSDs don’t require extra power.
- Memory bus: A wider bus offers more bandwidth for communication between the GPU and the video memory, although a smaller bus is commonly used with high clockspeeds.
- Clockspeed: Both the memory and GPU have different clockspeeds. These represent the speed at which a processor can run as the amount of instructions they can carry out per clock cycle.. Clockspeeds cannot be directly compared between different generations or brands of graphics cards.
- DirectX: A software developed by Microsoft to provide a unified way for developers to program for different graphics cards. The graphics API, or application programming interface, is the toolset used to build applications by developers that utilise hardware components, specifically for visuals and graphics rendering.
- OpenGL: Similar to DirectX, however this is open-source and available to all.
Vulkan: an evolution of the OpenGL API, based on AMD’s discontinued Mantle API...
- GPU Encoders: Nvidia NVENC, AMD VCE, and Intel Quick Sync are all H.264 encoders, utilising graphics hardware to encode the H.264/MPEG-4 codec rather than the CPU cores.
- Overclocking: Increasing clockspeeds beyond the initial product specification. Can be applied to graphics cards, memory and CPUs. Overclocked systems perform better, but in turn use more power and generate more heat.
- VRM: Voltage regulator module. Better voltage regulation allows for greater component lifespan and better overclocks.
- PCB: Printed circuit board. The PCB contains the connections for allowing the components of a system communicate and receive power.