Compare GPUs

📖 Technical TermsCompare CPUs Instead →

No GPUs selected

Select GPUs from the dropdown above to start comparing

📖 Technical Terms Explained

Performance & Specifications

Performance Index
A normalized benchmark score combining gaming performance across multiple resolutions (1080p, 1440p, 4K), ray tracing capabilities, and power efficiency. Calculated from real-world gaming and rendering tests. Higher numbers indicate better overall performance. Used to compare GPUs across different generations and architectures, providing a single metric to understand relative gaming and compute capabilities.
TFLOPS (Teraflops)
Measures a GPU's theoretical floating-point operations per second - essentially raw mathematical compute power. While higher TFLOPS suggests more processing capability, it doesn't directly translate to gaming performance due to differences in architecture efficiency, memory bandwidth, and driver optimization. More useful for comparing GPUs within the same architecture family. For example, a 20 TFLOPS RDNA 3 GPU may outperform a 30 TFLOPS older architecture in actual games.
TGP/TDP (Total Graphics Power)
The total power consumption of the GPU under typical gaming load, measured in watts. Higher TGP generally indicates higher performance but requires better power supplies and cooling. Desktop GPUs range from 75W (low-end) to 450W+ (RTX 4090). Mobile GPUs use configurable TGP (80-175W) allowing laptop manufacturers to balance performance and battery life.
Boost Clock
The maximum frequency the GPU cores can reach under load when thermal and power conditions allow. Modern GPUs dynamically adjust clock speeds based on temperature and power limits. Listed boost clocks are minimum guaranteed, with actual speeds often exceeding specifications. For example, an RTX 4080 rated at 2.5 GHz may boost to 2.7+ GHz with good cooling.

Memory

VRAM (Video Memory)
Dedicated memory on the graphics card storing textures, 3D models, frame buffers, and rendering data. More VRAM is required for higher resolutions and quality settings. Typical requirements: 1080p = 6-8GB, 1440p = 8-12GB, 4K = 12-16GB+. Insufficient VRAM causes stuttering and texture pop-in. Content creation and AI workloads often require 16-24GB for large projects.
Memory Type (GDDR6 / GDDR6X / GDDR7)
The technology and generation of video memory. Newer generations offer higher bandwidth and efficiency. GDDR7 (latest, RTX 50 series) provides highest speeds up to 28 Gbps. GDDR6X (RTX 30/40 series) offers 19-24 Gbps. GDDR6 (mainstream) provides 14-18 Gbps. Higher bandwidth improves performance at higher resolutions and with demanding textures.
Memory Bus Width
The data highway width between GPU and memory, measured in bits. Wider buses transfer more data simultaneously. Common widths: 384-bit (high-end), 256-bit (mid-range), 192-bit (mainstream), 128-bit (budget). Combined with memory speed, determines total memory bandwidth. For example, 256-bit bus at 18 Gbps provides 576 GB/s bandwidth, while 128-bit at same speed only achieves 288 GB/s.
Memory Bandwidth
Total data transfer rate between GPU and VRAM, measured in GB/s (gigabytes per second). Calculated from memory bus width × memory speed. Higher bandwidth improves performance at higher resolutions and with complex scenes. High-end GPUs achieve 800+ GB/s, while budget cards may have 200-300 GB/s. Insufficient bandwidth creates bottlenecks limiting GPU performance.

Technologies

Ray Tracing / RT Cores
Hardware-accelerated realistic lighting simulation creating accurate reflections, shadows, and global illumination in games. Dedicated RT cores handle ray tracing calculations separately from standard rendering. NVIDIA RTX GPUs use RT cores, AMD uses Ray Accelerators, Intel uses Ray Tracing Units. Performance impact varies: 20-50% FPS reduction in ray-traced games, mitigated by DLSS/FSR upscaling.
DLSS / FSR / XeSS (AI Upscaling)
AI-powered technologies that render games at lower resolution then intelligently upscale to native resolution, dramatically improving performance while maintaining visual quality. DLSS (NVIDIA) uses dedicated Tensor cores. FSR (AMD) works on any GPU through software. XeSS (Intel) uses XMX engines. Typically provides 40-100% FPS boost with minimal quality loss, enabling higher settings or resolutions.
Frame Generation
Advanced AI technology creating entirely new frames between real rendered frames, potentially doubling frame rates in supported games. DLSS 3 Frame Generation (NVIDIA RTX 40+), FSR 3 Frame Generation (AMD), and similar technologies predict motion and generate intermediate frames. Works best at already high frame rates (60+ FPS base). May introduce minor latency but makes 4K high-refresh gaming more accessible.
CUDA Cores / Stream Processors / Compute Units
The processing units performing graphics calculations. NVIDIA uses CUDA cores, AMD uses Stream Processors organized into Compute Units, Intel uses Xe-cores. More cores generally improve performance within the same architecture, but direct comparisons across architectures are meaningless. For example, 5,888 CUDA cores (RTX 4070) may outperform 7,680 cores (RTX 3070 Ti) due to architectural improvements.
Tensor Cores / AI Accelerators
Specialized cores dedicated to AI and machine learning workloads. NVIDIA Tensor cores power DLSS, enable AI denoising, and accelerate content creation. Found in RTX series GPUs. Each generation improves efficiency - 4th gen Tensor cores (RTX 40 series) are 2-3x faster than previous generation. Essential for AI applications, 3D rendering with AI denoising, and maximum gaming performance with DLSS.

GPU Variants

Ti / SUPER / XT (Enhanced Variants)
Enhanced versions with improved specifications over standard models. NVIDIA Ti (Titanium) and SUPER variants offer more cores, higher clocks, or more VRAM. AMD XT models provide refined specifications and better binning. Performance improvements typically range 10-20% over base models. For example, RTX 4070 Ti SUPER includes 16GB VRAM versus 12GB standard 4070 Ti, plus additional cores.
Laptop GPU Suffixes
Mobile GPU classifications indicating power profiles and target laptop types. HX = maximum performance (80-175W), desktop replacement gaming laptops. H = standard gaming (45-115W), mainstream gaming laptops. HS = thin-and-light balance (35-65W), premium ultraportables. S = ultraportable optimized (35-50W), slim gaming laptops. Higher power = better performance but worse battery life and thermals.
Mobile vs Desktop GPUs
Laptop GPUs share names with desktop counterparts but have significantly different performance due to power constraints. For example, RTX 4070 Laptop GPU (140W max) performs closer to desktop RTX 4060 Ti than desktop RTX 4070 (200W). Always check laptop-specific benchmarks. Desktop GPUs offer 30-60% more performance than mobile equivalents due to higher power budgets and better cooling.

Architecture

Ada Lovelace (NVIDIA RTX 40 Series)
NVIDIA's current generation architecture featuring 3rd gen RT cores, 4th gen Tensor cores, DLSS 3 Frame Generation, and massive efficiency improvements. Built on TSMC 4N process, delivering up to 2x performance per watt versus previous Ampere generation. Supports DLSS 3, AV1 encoding, and DisplayPort 2.1 for next-gen displays.
RDNA 3 (AMD Radeon RX 7000 Series)
AMD's current architecture featuring chiplet design, improved ray tracing over RDNA 2, and strong rasterization performance. Built on TSMC 5nm/6nm process. Supports FSR 3, DisplayPort 2.1, and AV1 encoding. Generally offers better rasterization price-to-performance than NVIDIA while trailing in ray tracing and lacking Frame Generation in some titles.
Alchemist (Intel Arc)
Intel's first generation discrete gaming GPUs featuring Xe-HPG architecture with dedicated ray tracing hardware, XeSS upscaling, and excellent video encoding. Competitive pricing and strong ray tracing performance per dollar, but faced driver maturity issues at launch. Ongoing driver updates continue improving performance. Strong value proposition for budget 1080p-1440p gaming.
Pascal (NVIDIA GTX 10 Series)
NVIDIA's legendary 2016 architecture delivering massive generational leap in performance and efficiency. Built on 16nm process, introduced modern power efficiency standards. GTX 1080 Ti became one of most beloved GPUs in history. Final generation before ray tracing introduction, focusing purely on rasterization performance. Many Pascal GPUs remain functional for 1080p gaming today.