GPU Performance Per Watt

How GPU Performance Per Watt Is Calculated

The score is: normalized benchmark score ÷ TDP (Thermal Design Power — the maximum heat a chip generates, in watts) × 100. A GPU scoring 49.60 delivers exactly twice the benchmark performance per watt as one scoring 24.80. Higher is always better.

Benchmark scores are normalized to the RTX 5090 = 100 — the same reference used on the Performance Per Dollar chart. The RTX 5090 scores 100 on the benchmark but draws 575W, landing it near the bottom of this chart. A lower-TDP card with a fraction of that raw score can still rank ahead of it simply by drawing far less power. More power erases the advantage of a higher raw score.

A GPU can rank first on performance per dollar and low on performance per watt. These are complementary metrics, not substitutes. A card may be cheap enough to lead the value chart while drawing so much power that it raises your electricity bill and won't fit an SFF case. Use both charts together for a complete picture before buying. If you're also comparing CPUs, the CPU Performance Per Dollar chart uses the same normalized scoring methodology.

Why Power Efficiency Matters More Than Spec Sheets Suggest

Running a 350W GPU for 4 hours a day — a realistic average gaming session — costs roughly $6.30 per month at $0.15/kWh. A 115W GPU on the same schedule costs $2.07 — a difference of around $51 per year. At $0.30/kWh, common across most of the EU and UK, the annual gap rises to roughly $102. Over a three-year ownership period, the electricity cost difference approaches $305 at European rates before the card is resold.

Small form factor (SFF) and mini-ITX cases often cap GPU TDP at 200–250W. Common limits: Cooler Master NR200 at 180W, NZXT H1 v2 at 200W, Fractal Terra at 250W. A GPU rated above the case limit runs hotter than its cooler can manage, throttles under sustained load, and spins fans faster to compensate. Cards at 115W TDP fit within the thermal limits of virtually every case, including the most restrictive SFF builds. For physical size limits (length and slot width) alongside TDP constraints, see the GPU size and case clearance guide.

Lower TDP directly reduces room temperature. Every watt the GPU draws becomes heat. In a room without central air conditioning, a 350W card adds meaningful heat load during summer gaming sessions — roughly three times more than a 115W card. Silent build builders get the clearest benefit: fewer watts means fewer RPMs needed from case fans, which means a quieter system without a trade-off in cooling safety.

Which GPU Tiers Lead on Efficiency

Budget tier: the formula for a high perf/watt score is low TDP combined with an adequate benchmark score. Budget cards that draw 100–130W while delivering competitive 1080p frame rates consistently rank near the top of this chart. Check the interactive chart for the current leaders — the ranking shifts as new GPUs are released and benchmark scores are updated. Use the table's TDP column to compare exact wattage figures side by side.

Mid-range: efficiency varies significantly within this tier and does not follow brand lines cleanly. Cards from the same generation can differ by 60W or more for similar benchmark scores — which is a meaningful perf/watt difference. AMD's RDNA 4 generation improved efficiency substantially over RDNA 2, but individual card results vary. Check the chart filtered by brand to see which current mid-range cards land above the class average before drawing conclusions about a specific model.

High end: the RTX 5090 has the highest raw benchmark (100/100) but ranks near the bottom on efficiency — 575W of TDP divided into a perfect score produces one of the lowest perf/watt values in the database. Intel Arc B580 at 190W sits mid-pack. Its 12GB VRAM advantage over competing 8GB cards doesn't appear in this chart — check the VRAM guide if VRAM headroom matters for your use case.

Efficiency rankings are stable across resolutions. TDP is fixed — a GPU draws the same power at 1080p as at 4K. Benchmark scores shift slightly at each resolution, but the leaders at 1080p are almost always the same at 1440p and 4K. Use the resolution toggle to confirm for any specific card you're comparing.

When Efficiency Should Drive Your GPU Choice

SFF builds: efficiency is non-negotiable. If your case caps GPU TDP at 200W, a card rated above that limit will throttle under sustained load regardless of its benchmark score. Filter the table by the TDP column and eliminate any card above your case's rated limit before looking at anything else.

Apartments without central air conditioning: every watt the GPU draws becomes heat in the room. A 350W GPU running flat-out adds roughly 4°C (7°F) of heat to a typical bedroom. A 115W card adds roughly 1°C (2°F). In climates that reach 30°C (85°F) or more in summer, that difference affects comfort meaningfully — independent of any air conditioning cost.

Long-term cost in high-electricity markets: if you're paying $0.25/kWh or more, calculate the three-year electricity cost difference before deciding between two cards at the same price point. A $40 price gap can be more than recovered in electricity savings within the first year if one card draws 150W more than the other.

When efficiency should not be the primary metric: large ATX mid-towers with good airflow, regions with cheap electricity (under $0.10/kWh), and situations where the most efficient GPU in your budget scores too low at your target resolution. In those cases, use the GPU Performance Per Dollar chart instead — it identifies the best raw value at your price point without the power constraint. For a full breakdown of how to evaluate a GPU choice beyond a single metric, see the how to choose a GPU guide. Or let the PC Builder handle it — it weighs benchmark score, price, and compatibility simultaneously.

Frequently Asked Questions