How Much Power Does a Gaming PC Use

I built a rig around an RTX 4070 Ti a while back and did the thing everyone tells you not to do: I looked at the card’s “285W” spec, added it to my CPU’s box number, rounded up a little, and bought a 650W PSU because it was on sale. Ran fine for two weeks. Then I turned on ray tracing in Cyberpunk during a heat wave with three browser tabs and a Discord call open, and the whole system hard-rebooted mid-firefight. Not a blue screen — a full power-off, like someone pulled the plug. It took me an embarrassingly long time to accept that the problem wasn’t my CPU cooler or a driver bug. It was the power supply, and the box number on the graphics card had lied to me by omission.

That’s the part most “how much power does a gaming PC use” articles skip: the number on the spec sheet is an average, not a ceiling, and gaming PCs don’t draw power like a light bulb. They spike.

The Spec-Sheet Number Isn’t the Number That Matters

Every GPU has a rated TDP (Thermal Design Power) — for reference, an RTX 4090 is rated at 450W, an RTX 4070 sits at 200W, and an RTX 4060 at 115W. But TDP describes sustained average draw under a defined test load, not the instantaneous peaks a card actually pulls during a real game. Transient power spikes — millisecond-long surges when the GPU’s clock boosts in response to a sudden workload change, like an explosion filling the frame or a texture stream kicking in — can push draw 30–40% above the rated TDP for fractions of a second. An RTX 4090 rated at 450W has been measured spiking past 600W in these windows. A cheap or undersized PSU that’s running near its limit doesn’t handle that gracefully; it trips its over-current protection and shuts the whole system down, which looks exactly like the crash I had.

CPUs do something similar under a different name. Intel’s higher-end chips run a “PL2” boost state that can pull well above their base rating for short bursts — a chip with a 125W base can hit 250W or more during that window, especially under all-core loads like compiling or a CPU-bound game. AMD’s X3D chips are gentler about this, typically staying closer to their rated TDP, which is one reason the power math is genuinely different depending on which platform you’re building.

The Actual Formula, Not a Rule of Thumb

The number you actually want is: add up your GPU’s rated TDP and your CPU’s peak boost draw (not its base TDP — check the boost/PL2 figure on Intel chips specifically), then add roughly 50–75W for the rest of the system — motherboard, RAM, a couple of NVMe drives, and case fans rarely add up to more than that on a normal build. Then multiply the whole thing by 1.3. That 30% isn’t padding for “just in case” — it’s there specifically to absorb the transient spikes above, and it also keeps the PSU running in the 40–60% load range where modern 80 Plus units hit their best efficiency curve, instead of running near max capacity where they run hotter and degrade faster.

Worked example: an RTX 4070 (200W) paired with a Core i5-14600K, which has a PL2 boost draw around 181W, comes to 381W for the two big components. Add 60W for everything else and you’re at 441W. Multiply by 1.3 and you land at roughly 573W — which is why a “650W should be plenty” recommendation for that exact pairing isn’t overkill, it’s the actual math working out to a standard PSU size. Swap that GPU for an RTX 4090 and the same formula pushes you past 900W before you’ve even accounted for overclocking headroom, which is the real reason Nvidia recommends a 1000W unit for that card and not because the card itself needs a full kilowatt to run.

What It Actually Costs You Every Month

The other half of “how much power does a gaming PC use” is the electric bill, and this is where the math is more forgiving than people expect — because a PC only pulls its peak number while you’re actually gaming, not 24 hours a day. The U.S. residential average was about 18 cents per kWh as of mid-2026 according to EIA data, though it swings a lot by region — New England runs closer to 30 cents, while parts of the middle of the country sit well under 15.

Take a system that averages 350W at the wall while actually gaming (a fairly typical RTX 4070-class build under load, not idle). Three hours a session, five sessions a week, comes to 15 hours a week, or about 65 hours a month. That’s 22.75 kWh a month from gaming specifically — at the 18-cent national average, that’s roughly $4.10 a month, or about $49 a year. Even a heavier RTX 4090 build averaging 550W at the wall under the same 65-hours-a-month usage comes to about $6.44 a month, or $77 a year. The GPU itself isn’t what makes a gaming PC expensive to run — it’s a fairly small line item next to almost anything else in a household budget. What actually changes the number is idle draw if you leave the PC on 24/7 for downloads or a home server role, which at even 60W idle adds another $9–10 a month all by itself, completely independent of how much you actually game.

FAQ

Do I need to size my PSU for the GPU’s rated TDP or its peak spike?

Size for the spike, not the rated TDP — that’s exactly what the 1.3x multiplier in the formula above accounts for. Sizing to the rated number alone is what causes random reboots under load, especially with high-end Nvidia cards that spike well above their box rating.

Does a bigger PSU waste more electricity even when the PC is idle?

No, not meaningfully. A quality PSU’s efficiency curve is about the load percentage, not its rated capacity — a 1000W unit running your system at 300W and a 750W unit running the same 300W load draw almost identical wall power, since efficiency is measured as a percentage of the load actually being drawn, not the unit’s ceiling.

Is it cheaper to undervolt my GPU to save on electricity?

It helps a little on the bill, but the real value of undervolting is lower heat and quieter fans, not savings — cutting 15% off GPU power draw on a build averaging $50/year in gaming electricity saves you roughly $7-8 a year. It’s worth doing for thermals and noise; it’s not a meaningful budget move on its own.

Why did my PC crash under load with a PSU that “should” have been enough on paper?

This is almost always a transient spike tripping the PSU’s over-current or over-power protection, not an actual sustained overload — the same failure mode described above. It’s more common with budget PSUs that have less robust protection circuitry and less capacitor headroom to smooth those millisecond spikes out.