How to Check Your Laptop GPU's Power Limit (TGP) in Windows

You bought a laptop with an RTX 4070, RTX 5070 Ti, or RTX 4090. The box says the right GPU name. But how hard is that GPU actually allowed to work? The answer is a wattage number called TGP — Total Graphics Power — and the laptop maker, not NVIDIA, sets it. Two laptops with the same GPU on the sticker can ship with very different limits, and the difference shows up directly in frame rates.

The good news: finding your laptop's real power limit takes about five minutes, and every tool you need is free. Here is how to check it in Windows, what each number means, and what to do with the result.

First, what TGP actually is

TGP is the maximum power the GPU subsystem is allowed to draw. A GPU's clock speed is not fixed: under load, the card boosts its clocks as high as it can within its power and temperature limits. When the power limit is the binding constraint — which it usually is in a laptop — the card settles at whatever clock speed fits inside the wattage budget.

Give the same chip more watts and it can hold higher clocks for longer. That is why TGP matters: the GPU name tells you the silicon, but the TGP tells you how hard that silicon is allowed to work.

Some stock figures, as set by NVIDIA's power policy and validated by driver-level tools: the RTX 5090 and 5080 Laptop ship at 175 W and have been validated up to 250 W; the RTX 4090 Laptop ships at 150 W and up to 250 W; the RTX 4080 Laptop ships at 150 W and up to 225 W; the RTX 4070 and 4060 Laptop sit around 120 W and up to 150 W; the RTX 4050 Laptop around 115 W and up to 140 W. RTX 30 and RTX 20 series laptop GPUs are in the same situation, but raising their limits is experimental rather than validated.

One more wrinkle: Dynamic Boost. NVIDIA shares power between the CPU and GPU, so when the CPU is idle, some of its power budget can be lent to the GPU. That means the limit you see can shift depending on what the machine is doing — a reading at idle and a reading under load are not always the same number.

Method 1: GPU-Z — the fastest check

GPU-Z is the simplest way to see your GPU's power limit, and it is free.

1. Download GPU-Z from TechPowerUp and run it — no installation needed.
2. Look at the main tab first. It confirms the GPU model, the memory, and the driver version, so you know you are reading the right card.
3. Open the Advanced tab and select NVIDIA BIOS (or the power-related entry) from the dropdown. This is where GPU-Z shows the board power limit — the wattage ceiling the driver enforces.
4. Switch to the Sensors tab and look at GPU Power Draw (sometimes listed as Board Power Draw). At idle this will be a few watts. The interesting reading comes under load, so leave GPU-Z open, run a game or a benchmark, and watch the number. If it climbs to the board power limit and stays pinned there, your GPU is power-limited: the TGP is the bottleneck.

GPU-Z also has a "Log to file" checkbox on the Sensors tab. Turn it on before a gaming session and you get a timestamped record of power draw, clocks and temperatures to review afterwards — much more useful than staring at a live graph.

Method 2: HWiNFO — live power, temperatures and clocks together

HWiNFO is the tool to use when you want the full picture, because it shows power draw alongside the things power affects: clocks, temperatures and throttling.

1. Download HWiNFO, run it in Sensors-only mode, and scroll to the GPU section.
2. Look for GPU Power (total board power), GPU Temperature, GPU Hotspot Temperature, and GPU Clock. HWiNFO also reports P-states, which tell you how hard the GPU is working at any moment.
3. Start a game or a benchmark and watch the readings. The useful comparison is simple: if GPU Power is sitting at the limit while the hotspot temperature is well below its ceiling, you are power-limited — more watts would mean more clocks. If the temperature is at its ceiling and the clocks are dropping, you are thermally limited, and more power would mostly make more heat.

HWiNFO can log everything to a CSV file, which makes before-and-after comparisons easy. Run the same benchmark twice and compare the logs: power draw, average clock and hotspot temperature tell you more than any single number.

Method 3: nvidia-smi — the command-line check

If you would rather not install anything, NVIDIA's own command-line tool is already on your machine with the driver.

1. Open a Command Prompt or PowerShell window.
2. Type nvidia-smi -q -d POWER and press Enter.
3. Look for the Power Limit entry. It shows the enforced limit in watts, and on some drivers the enforced limit alongside the default limit.

This is the same number GPU-Z reads from the driver, just without the graphical interface. It is handy for a quick sanity check, and it is the tool to reach for if you want to script a reading or check the limit on a machine you are remoting into. What nvidia-smi cannot do on a laptop is change the limit — it can read the power policy, but laptop GPUs do not expose the write side of it. That is a deliberate limitation, not a bug.

Method 4: the laptop maker's spec sheet

Before you trust any on-machine reading, it is worth knowing what the laptop was sold as. The manufacturer's product page usually lists the graphics TGP next to the GPU model — sometimes plainly ("RTX 4070 Laptop GPU, 140 W TGP"), sometimes buried in a spec table or a footnote.

The spec sheet number and the on-machine number do not always match, and there are honest reasons for that:

• Dynamic Boost headroom. The advertised figure may include watts borrowed from the CPU when it is idle. The base limit GPU-Z reports can be lower, with the extra watts only available when the CPU is not using them.
• Model variants. The same laptop chassis can be sold with several GPU and TGP combinations. Make sure the spec page you are reading matches your exact configuration, not just the model family.
• Silent revisions. Laptop makers occasionally change the power policy in a BIOS or driver update. The spec sheet is a snapshot from launch; the machine in front of you is the truth.

Use the spec sheet as the expectation and the on-machine tools as the reality. When they disagree, believe the machine.

What the number means for your games

Once you know your limit, the question is whether it is holding you back. The honest answer is: it depends on what is limiting you right now.

Run a GPU-heavy benchmark — 3DMark Time Spy or Steel Nomad are good choices because they are repeatable — with HWiNFO logging in the background. Then look at the log and ask three questions:

1. Was power draw pinned at the limit? If yes, the GPU wanted more power than it was allowed. Extra TGP would have bought higher clocks.
2. Was the hotspot temperature at its ceiling? If yes, the cooling system is the bottleneck. More power would mostly turn into more heat, and the card would throttle anyway.
3. Were the clocks steady or dropping over time? Steady clocks with pinned power means the limit is doing its job. Clocks that sag as temperatures climb mean thermals are winning.

This is also the baseline you need before changing anything. If you later raise the limit, run the same benchmark again and compare the logs. The difference in average clock speed and frame rate is the real answer to "was it worth it" — not a marketing number.

If the number looks low, what can you do about it?

If your GPU is power-limited and your temperatures have headroom, raising the limit is a real option. MSI Afterburner is the tool most people try first, and it is worth knowing its limit: it can shift clocks, but it cannot raise a laptop GPU's power ceiling. The ceiling itself is set in the driver power policy, and changing it needs a tool that works with the driver directly.

NvpwrControl is built for exactly that. It is a portable Windows app — extract the zip, run NvpwrControl.exe as administrator, no install — that raises NVIDIA laptop GPU power limits by working with the NVIDIA driver directly. It never flashes VBIOS. Before changing anything it backs up the driver settings and checks the GPU, PCI IDs, driver and VBIOS against the live power policy, so a write only happens when the card matches what the tool expects.

The flow is simple: set the maximum limit, save, reboot, then set the current runtime limit while the GPU is idle. Limits adjust in 5 W steps, so you can creep up rather than jumping straight to the top. Core clock offset goes up to ±1000 MHz and memory offset up to ±3000 MHz, and there is experimental V/F curve voltage control for those who want it. Telemetry covers clocks, voltage, power, temperature, hotspot, P-states and rails, so you can watch exactly what the extra watts are doing.

An auto-reapply option re-sets the limit after MSI Center mode switches — a kernel event with a 1–10 second delay — and Reset All returns everything to stock and removes the override and autostart. The published 3DMark results give a sense of what the headroom is worth: an RTX 4090 Laptop scored 25,564 in Time Spy Graphics unlocked versus a 21,007 stock average, and an RTX 5090 Laptop scored 9,072 in Steel Nomad versus 5,899 stock — around 30% more GPU performance in 3DMark across the validated range.

The honest risks, before you touch anything

Raising a power limit is not free performance. More watts mean more heat — on the GPU, the VRAM, the voltage regulators and the cooling system. Your power adapter has to supply the GPU and the CPU together; if it cannot, the laptop throttles or drains the battery even while plugged in. And warranty depends on the laptop maker: some treat a raised power limit as a modification, others do not.

So the rules are simple: raise in small steps, watch temperatures with HWiNFO or GPU-Z the whole time, and stop if the hotspot climbs faster than the clocks do. A 5 W step that buys you a stable clock increase is worth more than a 50 W jump that ends in thermal throttling. Check your limit first, measure the baseline, and only then decide whether the extra watts are worth it.

---

NvpwrControl is independent software, not affiliated with NVIDIA. Support: timmy@nvpwrcontrol.site.