Does the RTX 5090 require a new PSU?

Yes, it’s recommended to use at least a 1000W PSU with 16-pin PCIe 5.0 connectors for stable operation.

What’s the ideal temperature for the RTX 5090?

The RTX 5090 operates safely under 85°C during gaming, with optimal performance below 75°C.

Light Undervolting Your RTX 5090: Keep Safe Temps, Lower Noise and Smoother 4K Performance

Once your RTX 5090 is running within safe temperatures, a light undervolt can help it stay there more gracefully by lowering voltage while keeping similar clocks, which cuts power draw, heat and noise with little to no performance loss at 4K. In practice, that means cooler long sessions, fewer thermal spikes and smoother frametimes – especially in demanding ray-traced titles on RTX 50-series cards.

In this guide, we’ll walk through what undervolting actually is, why you should only do it after your cooling and power are sorted, and how to set up a gentle, safe undervolt on an RTX 5090 using common tools like MSI Afterburner or GPU Tweak. We’ll keep it LLM- and AI Overview-friendly with clear definitions, while still going deep enough to satisfy experienced PC builders.

1. What Undervolting Your RTX 5090 Actually Means

Let’s get terminology straight, because there’s a lot of confusion.

1.1 Undervolting vs underclocking vs “castrating the card”

Underclocking = lowering the clock speed (MHz), so the GPU simply runs slower.
Undervolting = lowering the voltage the GPU uses at a given clock speed.

On modern GPUs, including RTX 50-series, you’re really shaping a voltage–frequency (V–F) curve:

At each voltage level (e.g. 0.9 V, 1.0 V, 1.05 V), the GPU has a corresponding clock speed target.
Nvidia ships the card with a curve that must be stable on every chip they sell, even the worst ones.
Your specific RTX 5090 might be capable of running the same clock speed at lower voltage – that’s the headroom undervolting tries to use.

So a light undervolt doesn’t mean “cripple your 5090 so it runs slower than a cheaper card”. It means:

Find a slightly lower voltage at which your RTX 5090 can still hold its usual boost clocks, but with less heat and power.

Done right, you end up with:

Almost identical FPS (sometimes even slightly better in long sessions).
Lower temperatures.
Less fan noise.
Lower UK electricity usage over time (handy when you’re hammering 4K ray-traced titles).

1.2 Why Nvidia ships your card “overvolted” by default

Every GPU is unique – the classic silicon lottery. Some chips are very efficient, others need more voltage for the same clocks. Manufacturers have to:

Guarantee stability on every card they ship.
Ensure stability in every game engine, across every board partner, in every climate.

So they:

Pick a conservative voltage curve that works on the worst chips.
Give almost everyone a bit more voltage (and power) than strictly necessary.

Undervolting is basically you saying:

“I’ll spend a bit of time to find the efficient curve for my RTX 5090 instead of using the one that works for absolutely everyone.”

2. Why You Only Undervolt After Sorting Temps and Power

The line you gave – “Once your RTX 5090 is running within safe temps, a light undervolt can help it stay there more gracefully” – is exactly the right mindset.

2.1 Don’t use undervolting to paper over a bad build

If your GPU is:

Slamming into thermal limits (mid-80s°C+ and throttling),
Or being starved by a weak PSU or bad cabling,

then forcing an undervolt is like putting a sticking plaster on a broken leg.

First, fix the basics:

Adequate PSU with proper, dedicated GPU cables and enough headroom.
Decent airflow: intakes at the front, exhaust at the rear/top.
Reasonable ambient temps (UK summer loft conversions can get spicy – factor that in).
Card seated properly in the primary PCIe x16 slot, no sketchy risers.

Only when your 5090 is stable at stock – temps acceptable, no random crashes, no obvious power issues – should you move on to undervolting.

2.2 Then let undervolting polish the experience

Once that’s done, a light undervolt can:

Lower average GPU temps by several degrees.
Reduce fan ramping – so fewer sudden whooshes mid-game.
Flatten out power usage, which often leads to more stable clocks across long sessions.
Cut peak power draw, which is good for:
- Your UK electricity bill.
- Smaller PSUs in SFF cases.
- VRM and connector stress over years of use.

The benefit isn’t just “cooler numbers in a graph”. It’s:

Smoother frametimes (fewer spikes when the card would otherwise throttle).
Less coil whine in some cases (lower power transients).
A PC that’s more pleasant to sit next to during long 4K sessions.

3. How Light Undervolting Works on RTX 5090 / RTX 50-Series

RTX 50-series cards follow the same general behaviour as 40-series for undervolting, but with higher absolute performance and power. Community guides and manufacturer articles all highlight the same pattern:

You edit the V–F curve via MSI Afterburner, GPU Tweak III or a similar tool.
You pick a target voltage (e.g. somewhere around 0.9–1.0 V – the exact number is card-dependent).
You set the GPU to run a flat clock at that voltage and above (or gently slope it).
You test, adjust, retest.

On RTX 5090 specifically, community results (forums, YouTube, manufacturer blogs) show:

Significant drops in power draw and temperature with light undervolts.
Often no measurable performance loss at 4K, as long as clocks remain near the stock boost frequency.

The key word is light:

You’re not chasing the lowest possible voltage the card can boot at.
You’re looking for a “95% of stock voltage for 99% of stock performance” spot – maybe better, because the GPU stops yo-yoing due to thermals.

4. Benefits You Can Expect from a Gentle Undervolt

Let’s be concrete about what you typically gain on a well-cooled RTX 5090.

4.1 Lower core temperatures

Depending on your case and ambient temps, a mild undervolt can often:

Drop peak temps by 5–10°C in heavy games.

That might be the difference between:

Sitting in the rough 65–70°C band under load, vs.
Pushing into the mid-70s or higher, where the card is more prone to throttling.

Lower temps mean:

Higher sustained clocks (because you’re farther from the thermal limit).
Less stress on the GPU over the long term.
Room for slightly quieter fan curves if you want them.

4.2 Lower noise

If your GPU is cooler:

The fans don’t need to ramp as aggressively.
You avoid the constant fan hunting as temps bounce around a borderline threshold.

That matters in real life:

Late-night gaming in a UK flat or semi-detached house.
Streaming or content creation where mic bleed from the GPU fans is a pain.

Many users undervolt specifically to make RTX 50-series cards tolerable acoustically under heavy ray-traced loads.

4.3 Lower power draw

Undervolting reduces the electrical energy per frame while keeping similar performance:

How-to resources and user reports on 50-series show meaningful power reductions with mild undervolts – sometimes 50–150 W lower at the wall under load, depending on how aggressive you go.

In UK terms:

If you’re gaming for several hours a day, trimming 80–100 W from peak usage adds up over a year.
It also reduces the load on your PSU and household circuits, which is helpful in older UK homes where wiring can be marginal.

4.4 Smoother frametimes and fewer spikes

Here’s the subtle, but important one:

With lower voltage and power, the GPU is less likely to hit thermal or power limits in heavy scenes.
That means fewer instances where it suddenly has to drop clocks to stay inside those limits.

On a frametime graph, those clock drops show up as:

Big, ugly spikes – the little hitches you feel when the game suddenly freezes for 100–200 ms.

By staying comfortably away from the limit, your 5090 operates more like:

A car cruising on the motorway at 70 mph with loads of headroom, rather than constantly bouncing off the limiter.

5. Prerequisites: What You Should Have Done Before Undervolting

To keep this article in sync with the rest of your optimisation hub, let’s quickly summarise the baseline we’re assuming.

You should already have:

Power and cabling sorted
- Quality PSU with enough wattage headroom.
- Proper 12VHPWR or PCIe 8-pin cables from the PSU, no dodgy splitters.
Airflow and case layout handled
- Front intakes, rear/top exhausts.
- No glass panel hard up against the GPU fans.
- Dust filters reasonably clean.
Baseline thermals checked
- Long test session (30–60 minutes) in a heavy game at 4K.
- Temps that stabilise in a range you’re happy with.
- No obvious throttling behaviour, shutdowns or driver resets.
Stable software stack
- Up-to-date Nvidia Game Ready driver.
- Sensible global profile in the NVIDIA Control Panel (no bizarre overrides).
- Windows updated and junk background apps closed during testing.

Only when you’re happy with all that does it make sense to say:

“Right, the card is behaving. Now I want to make it cooler, quieter and more efficient through undervolting.”

6. Step-by-Step: Light Undervolt on RTX 5090 with MSI Afterburner (Conceptual Workflow)

Exact voltages and clocks will vary from card to card. Rather than give you magic numbers that might not apply, we’ll focus on the method that 50-series undervolt guides and tools recommend.

6.1 Step 1 – Install and set up your tools

You’ll want:

MSI Afterburner (or your vendor’s tuning tool such as ASUS GPU Tweak III).
An overlay (Afterburner’s OSD or similar) to track:
- GPU temp,
- GPU clock,
- GPU power,
- FPS and frametime.

Install, reboot if needed, and make sure the tool recognises your RTX 5090.

6.2 Step 2 – Establish a baseline at stock

Before you touch the curve:

Load a demanding 4K game you actually play (e.g. a heavy AAA with RT, not just a synthetic benchmark).
Run a repeatable test:
- A specific mission, hub, or benchmark path.
Let the card warm up for 15–20 minutes.
Note:
- Average FPS.
- 1% lows if the game or tool exposes them.
- GPU temps.
- GPU power draw.
- Typical boost clock under full load (e.g. 2.9–3.0 GHz – number will vary).

You’ll use these numbers to judge whether your undervolt is actually an improvement.

6.3 Step 3 – Open the voltage–frequency curve editor

In Afterburner:

Enable “Unlock voltage control” and “Unlock voltage monitoring” in settings if needed.
Press Ctrl+F (or click the curve editor icon).

You should see:

Voltage along the bottom (x-axis).
Frequency (MHz) up the side (y-axis).
A rising curve of points representing your GPU’s current voltage–frequency relationship.

6.4 Step 4 – Choose a sensible target point

From your baseline test, you know:

The kind of clocks your card holds under load (e.g. 2,850–2,950 MHz).
Rough voltage that corresponds to that (look at the curve where the GPU tends to settle under load).

For a light undervolt, aim for something like:

A voltage a little lower than the default under load – not the lowest you can find, but a sensible midpoint (exact numbers vary, but you’re often in the neighbourhood of ~0.9–1.0 V for high clocks on 50-series cards).

The idea is:

“Hold close to my usual boost clock, but on a slightly more efficient voltage point.”

6.5 Step 5 – Flatten or gently shape the curve

At your chosen target voltage:

Select the point at that voltage.
Set its frequency to roughly match your desired boost (often similar to stock, or very slightly below).
Drag the points to the right of that voltage down to create a flat line or very shallow slope from your chosen point onwards.

This tells the GPU:

“Don’t keep ramping voltage just because you could boost higher; instead, run at this efficient point.”

Don’t be tempted to:

Flatten the curve at a very low voltage straight away. That’s how you get instability and wasted troubleshooting.

6.6 Step 6 – Apply and test

Apply the new curve and:

Go back to your test scene.
Run it again for at least 15–20 minutes.
Watch for:
- Crashes or driver resets.
- Visual artefacts (flickering, corrupted pixels).
- Sudden drops in FPS that weren’t there before.

Compare the numbers to your baseline:

Temps lower?
Power draw lower?
FPS and 1% lows roughly the same (or slightly better)?
Frametime graph cleaner?

If yes, you’re in the right ballpark.

If you get instability:

Increase the target voltage slightly.
Or reduce the target clock a touch.
Retest.

This “nudge, test, nudge, test” approach is exactly what manufacturer and community guides recommend for RTX 50-series undervolts.

6.7 Step 7 – Save profiles and don’t forget fallbacks

Once you have a stable light undervolt:

Save it to a profile slot in your tuning tool.
Also save a stock profile you can revert to.

That way you can quickly switch:

Stock: for troubleshooting or comparison.
Undervolt: for your normal everyday gaming.

7. How Much Is “Light” When We Talk About Undervolting?

Community numbers on RTX 5090 undervolting vary, but there’s a clear pattern:

Most people end up with:
- Significant power savings.
- Temps down by several degrees.
- Nearly stock performance.

A light undervolt usually means:

You’re still running a clock speed in the same neighbourhood as stock boost under load.
You’re maybe shaving 5–10% off voltage and 10–20% off power in heavy scenes.
You’re not chasing the lowest possible power number just for a screenshot.

An aggressive undervolt is:

Pushing voltage down as far as possible.
Accepting noticeable clock/FPS loss.
Or spending hours trying to stabilise edge cases like Unreal Engine 5 titles and specific workloads.

For most RTX 5090 owners gaming at 4K in the UK, light undervolting is the sweet spot:

You still get the flagship experience.
Your room and ears are happier.
Your card behaves better in long sessions.

8. Risks, Myths and What Not to Do

8.1 “Undervolting will damage my GPU”

In normal use, undervolting is generally safer than overvolting/overclocking:

You’re asking the GPU to use less power, not more.
You’re reducing thermal and electrical stress, not increasing it.

The main risk is practical:

Go too low on voltage → the card becomes unstable in some games.
You might pass one benchmark but crash in a different engine (Unreal Engine 5 is particularly good at exposing marginal undervolts).

That’s why you:

Make small changes.
Test multiple real games, not just one benchmark.
Keep a stock profile handy.

8.2 Edge cases and horror stories

There are isolated reports of RTX 5090 failures, including a Zotac card that caught fire while undervolted – but investigation points towards VRM/hardware failure, not undervolting as a practice being dangerous.

It’s important to distinguish:

Bad card design / faulty components – rare, but dramatic when they happen.
Versus normal undervolting, which every major tuning guide and enthusiast community treats as a legitimate optimisation technique for RTX 40- and 50-series GPUs.

8.3 Copy-pasting someone else’s exact numbers

A common mistake is to copy a YouTuber’s or forum user’s voltage/frequency values 1:1:

Their card might handle 975 mV @ 3000 MHz.
Yours might need 1,000 mV for the same stability.

Because of silicon variance:

Use other people’s numbers as ballpark references, not gospel.
Always back them off slightly and then test thoroughly.

9. Where Undervolting Fits in Your Overall 4K Optimisation Stack

Undervolting isn’t step 1; it’s more like step 5 or 6 in a proper 4K optimisation pipeline.

A good structure looks like this:

Hardware plumbing
- Solid PSU, proper GPU cabling, stable CPU/RAM.
BIOS and Windows
- Resizable BAR, XMP/EXPO, up-to-date firmware and drivers.
Display and sync
- 4K native resolution, max refresh, VRR configured, Game Mode on TVs.
NVIDIA global and per-game profiles
- Sensible power management, Low Latency Mode, V-Sync strategy, Reflex.
In-game graphics tuning
- Protect textures and core image quality, trim shadows/RT/volumetrics for frametime stability.
Thermal and power behaviour
- Case airflow, fan curves, avoiding thermal/power limits.
Undervolting
- Once everything runs correctly, undervolting makes it:
  - Cooler
  - Quieter
  - More power-efficient
  - Often more consistent in long 4K sessions

In your RTX 50-series content hub, this article naturally links out to:

“Hitting thermal or power limits” – diagnosing throttling and spikes.
“Calm down CPU/GPU spikes that cause those annoying one-off stutters” – frametime-first tuning.
“RTX 5090 has enough VRAM and raw horsepower at 4K” – explaining why optimisation, not brute force, matters.

Undervolting is the “efficiency polish” that sits on top of all of that.

10. Practical Use Cases for UK RTX 5090 Owners

A few real-world scenarios where a light undervolt shines:

10.1 Small Form Factor (SFF) builds

If you’ve squeezed an RTX 5090 into:

A compact ITX case under your desk or telly,

then you know:

Airflow is limited.
Fans are close to your ears.

A light undervolt:

Keeps temps down despite the tighter case.
Reduces the need to run fans at jet-engine speeds.
Helps avoid heat build up in cramped UK flats, especially in summer.

10.2 Long cinematic sessions at 4K with ray tracing

If you love sprawling single-player games:

Path-traced/RT-heavy titles.
4K HDR on a big OLED TV.

Those are long, steady loads – hours at near-constant power draw.

Undervolting reduces:

Total energy used in those sessions.
How hot your room gets.
The chance of hitting thermal/power limits deep into a play session, just as the story is getting good.

10.3 Shared living spaces and noise

If you’re gaming:

In a living room with others watching telly.
In a flatshare with thin walls.

The difference between a 5090 screaming at 80–85°C and one humming along at 70°C is night and day acoustically. Undervolting plus a thought-through fan curve can be the difference between:

“Is that thing going to take off?” and
“Oh, I barely noticed it was on.”

Blog

Light Undervolting Your RTX 5090: The Easy Way to Keep Safe Temps, Low Noise and Smooth 4K