You’ve read articles about undervolting and power limiting, but what are we actually trying to accomplish? I’ve seen those two terms used like they are interchangeable but they are not.
Undervolting a CPU or GPU does not necessarily reduce performance. The CPU manufacturer specifies operating conditions that guarantee reliable operation, but modern CPUs and GPUs dynamically adjust voltage based on frequency, workload, temperature, and the quality of the individual chip. In many cases, the chip can maintain the same clock speeds at a lower voltage than the default settings provide. The result can be equal or greater performance with less heat generated. Less heat means that higher clock speeds can be maintained for longer before thermal throttling becomes an issue.
Power limiting means restricting the overall wattage available. This CAN limit performance when there simply is not enough power available to maintain boost clock speeds under certain scenarios. While the intentions behind undervolting (potentially improved performance with reduced heat) are obvious to me, power limiting does not make sense.
To me it is like buying an RTX 5080 with a TDP of 360 watts – a number that was not just pulled out of thin air – and limiting the power to say 250 watts (that of my RTX 5070) or 180 watts (RTX 5060 Ti). The RTX 5080 has a 360-watt power target because NVIDIA designed the card to deliver its intended level of performance within that power and thermal envelope. That doesn’t mean the GPU needs 360 watts simply to function. It means you’re giving the GPU the power budget it was designed to use when you’re asking it to deliver its full performance.
If you choke your RTX 5080 down to 180 watts, would you have been better off just buying an RTX 5060 Ti and saving a ton of cash? Not exactly. The RTX 5080 is still superior in many ways and limiting the power draw to that of a 5060 Ti will not completely drop the performance to that of a 5060 Ti, but it will not perform like an RTX 5080 was designed to.
So again, why are people setting power limits on their GPUs? Based on my personal experience, I can only assume it’s noise and heat. (Possibly the power bill, but if that was a concern you probably wouldn’t have an RTX 5080…) My only complaint that I have with my gaming laptop is the amount of heat that it generates while gaming. Part of that is due to the 140-watt RTX 3060 laptop GPU. I appreciate that my laptop has one of the most powerful versions of the mobile RTX 3060, but the tradeoff is heat. I have tackled this in 2 ways. Either put it in turbo mode and let the fans sound like a jet taking off, or if I am going to use a less aggressive fan profile, I limit the thermal target of the GPU. This means it thermal throttles at 82c instead of 87c but there is that much less heat to soak into the rest of the laptop.
There are, of course, legitimate reasons to power-limit a GPU. If you’re dealing with a small-form-factor case, inadequate cooling, a laptop chassis, a limited power supply, or some other physical constraint, a power limit can be a useful tool. In that situation, you’re making a deliberate tradeoff: the hardware is capable of more, but the system it’s installed in can’t comfortably handle its full power envelope. You’re sacrificing some performance to keep temperatures, noise, or power consumption within an acceptable range. My argument isn’t that power limiting is useless. It’s that if your hardware has plenty of performance to spare, why restrict the power available to the GPU when you can simply reduce the amount of work you’re asking it to do?
Today we are not specifically talking about laptops, but the principle is the same. If you lower the max thermal target, the CPU or GPU will still perform at maximum efficiency until it hits that temperature. With a reduced power limit, you lose all the performance under that temperature curve.
Now so far we have talked about managing temperature. Now let’s talk about how to reduce temperature and power consumption at the same time – all with easier to understand settings than undervolting or power limiting.
I have since upgraded from my gaming laptop both for the heat concerns and because the 6GB VRAM was no longer enough. My newest system has an AMD Ryzen 7 7800X3D, 32GB DDR5-6000, and an NVIDIA GeForce RTX 5070 12GB. This is by no means a top-of-the-line system. Both the 7800X3D and the RTX 5070 are very efficient for the performance that they provide, however extended gaming sessions can still be felt in both the temperature of the room and when the power bill arrives. Again this is an efficient system. Unlike an i9 / RTX 5090 system, it is not competing with my A/C for power consumption. But I also think it’s silly to be burning away 400+ watts while watching a cutscene.
The power figures I quote here are the total system including the PC and 1440p 180hz display. I also have the total GPU power as reported by the NVIDIA overlay, but the whole system power tells a larger story.
While I have personally applied a slight overclock and undervolt to my CPU and used NVIDIA’s automatic overclock for the GPU, these will not significantly affect power draw under normal gaming. So while playing something such as RDR2, which is a slow paced, cinematic game, I hear the AIO fans and GPU fans running just enough to hear them over the game. I realized the computer is putting off a lot of heat. Pull up some stats. Temperatures are normal, GPU 96% utilization, roughly 120 FPS. And then it hit me. My poor GPU is running a marathon to pump out 120 FPS while I feed my horse and talk to people in camp. I play RDR2 on my HTPC at 60 FPS and it is perfectly enjoyable.
So I set refresh rate to 60hz…. Game still running at 120 FPS and now I have tearing…. Turn on VSYNC and success! 60 FPS locked in, GPU at 48% utilization, 63c, and 129 watts. Total system power dropped to 360 watts measured at the wall.
At 121 FPS, the GPU was running at 96% utilization, 73c, and 217 watts. Total system power was 480 watts.
| 60 FPS | 121 FPS | Difference | |
| GPU utilization | 48% | 96% | +48 percentage points |
| GPU temperature | 63°C | 73°C | +10°C |
| GPU power | 129 W | 217 W | +88 W |
| Total system power | 360 W | 480 W | +120 W |
That’s a 120-watt difference in total system power just to produce another 61 frames per second.
That’s more than I expected. I knew the GPU would use less power, but I wasn’t expecting the entire system to drop by 120 watts simply by telling the GPU to stop rendering frames I don’t need.
And you don’t need to dig through individual game settings to do this. Both NVIDIA and AMD provide tools for controlling graphics settings globally or on a per-game basis, so the process is simple whether you are on team red or team green. In the NVIDIA app, graphics and driver settings can be configured globally or overridden for individual games. On AMD systems, AMD Software: Adrenalin Edition provides global graphics settings as well as individual game profiles.
AMD also has a particularly interesting feature called Radeon Chill. Unlike a simple FPS cap, Chill can dynamically regulate the frame rate based on what’s happening on screen. You can set a minimum and maximum FPS, and it will lower the frame rate when there is little movement and increase it when the action picks up. AMD specifically describes Radeon Chill as a power-saving feature intended to reduce power consumption and GPU temperature. It can be enabled globally or configured separately for individual games.
This is exactly the kind of setting I am talking about. Instead of telling the GPU, “You are only allowed to use 250 watts,” you’re telling it, “You don’t need to render 200 frames per second right now.” The hardware can still use whatever power it needs to produce the frames you actually want, but it isn’t wasting power producing frames that provide no benefit.
What I find interesting about this specific method is that it is one setting that makes the entire system more efficient. Lowering the FPS from what the system CAN do to what you actually need reduces the load on the CPU and GPU, which in turn reduces the load on the power supply, VRMs, etc. It also reduces power draw and heat generated by the display while operating at a lower refresh rate. And the best part is all the performance you need is still on tap when you play a game that needs all of the FPS your system can crank out.