Video Cards (GPU)

Category overview

A discrete graphics card adds dedicated video memory and processing aimed at one of two main jobs: drawing pixels (gaming, design, simulation) or running parallel compute (machine learning, scientific work). The right card depends much more on the workload than on the headline performance number.

What this category covers

Consumer GPUs (current generations from major vendors) for gaming, content creation and general workstation use; professional workstation GPUs with certified drivers for computer-aided design (CAD), 3D rendering and video work; and entry-level compute-oriented cards for development and small-scale machine learning.

How to choose a graphics card

Define the dominant workload first

For gaming, the target resolution and refresh rate sets the GPU class. For 3D rendering and video editing, GPU memory size and rendering engine support matter more than peak frame rates. For CAD, certified workstation drivers and ISV support are often required by the application.

Match GPU memory to the data set

Modern games and creative software hold larger textures and assets in GPU memory. 8 GB is the modern minimum for 1080p gaming and basic creative work. 12 to 16 GB is comfortable for 1440p, 4K and many production workflows. Machine learning and large 3D scenes can need 24 GB or more.

Confirm the system can power and fit it

Check the card's total board power and the power supply's rated wattage and connectors. Measure the card length against the case's maximum supported GPU length. Some current cards take three slots in height, which can block other PCIe slots underneath.

Check display outputs

Most cards have a mix of HDMI and DisplayPort. For multi-monitor workstations, count the displays and check the maximum supported simultaneous outputs and resolutions on the card's specifications page.

Key features to compare

Architecture and generation

A new mid-range card often outperforms an older high-end card at lower power. New architectures also add features (ray tracing acceleration, hardware encoders, better AI features) that affect specific workflows.

Cooler design

Most consumer cards use open-air axial coolers, which need a well-ventilated case. Blower-style coolers exhaust hot air directly out of the case and are common in workstation cards and dense multi-GPU servers.

Driver path

Consumer drivers are tuned for games and general use. Workstation drivers are tuned and certified for specific CAD and engineering software, which can make a real difference in stability and supported features in those applications.

What to check before buying

• Does the GPU class match the actual workload (gaming resolution, render engine, CAD software)?
• Is the GPU memory enough for current and near-future scenes or datasets?
• Does the power supply have the wattage and the correct connectors for the card?
• Will the card physically fit in the case (length, height in slots, clearance for power connectors)?
• Does the use case actually require certified workstation drivers?

Frequently asked questions

Do I need a workstation GPU or will a gaming card do?

For most creative work, gaming, video editing and 3D rendering, a consumer GPU offers better value. Workstation cards make sense when the application explicitly requires certified drivers (some CAD packages do) or when ECC (Error-Correcting Code) memory and certified support are part of the procurement requirement.

How much GPU memory is enough?

For 1080p gaming and lighter creative work, 8 GB is the modern minimum. For 1440p, 4K, larger 3D scenes and modern games with high texture quality, 12 to 16 GB is more comfortable. For heavy rendering and machine learning, 24 GB or more is often necessary.

Will my power supply handle a new GPU?

Check two figures: the PSU (Power Supply Unit)'s total wattage versus the card's recommended PSU wattage, and the PSU's available connectors versus what the card requires. A current high-end card may need a 12V-2x6 connector or several 8-pin PCIe connectors. If either falls short, replace the PSU before fitting the GPU.

Why is my new GPU louder than expected?

Common causes: the case has poor airflow and the card ramps fans to compensate; the fan curve is set aggressively in the GPU control software; or the card is mounted in a slot that restricts its bottom airflow. Improving case airflow and using a custom fan curve usually helps.

Can I use two different GPUs in one PC?

Two cards from the same vendor with the same drivers usually work, often used for compute, video editing or attaching extra displays. Mixing cards from different vendors works for displays but rarely for accelerated workloads, since each application chooses one card to render with.