PC Hardware with power limits of CPUs and graphics cards climbing ever higher, and news reports of chips hitting sky-high temperatures appearing almost weekly, it seems that now is as good a time as any to ask a simple question — just how hot is too hot for today’s PC Hardware components?
Below is a reference point for max temperatures you should see on your PC hardware, but answering this question properly isn’t as straightforward as you might think. So let’s peer deep inside the innards of a computer, looking at why things get hot when working, what happens to them if they get too hot, and what can be done about it all.
Summary table of maximum temperature ranges
Before we get started on looking at why the different parts in a PC Hardware get hot, here’s a summary of the typical maximum temperatures you’ll likely to come across in the consumer market.
For specialized industrial situations, the limits can be a lot higher — for example, RAM in automotive applications can be rated to as high as 130°C.
Why do PC components get hot?
All computer chips: CPU, GPU, DRAM, NAND flash, and so on are made out of multiple different materials — semiconductors (e.g. silicon), metals (e.g. copper), ceramics, and plastics. It’s the flow and storage of electrical charge through the semiconductor and metal layers that make the chips function as intended.
Unfortunately, these materials will resist this flow or leak stored charge, and the energy carried by the charge gets dissipated as heat. And since today’s chips contain billions of components, all connected through a dizzying number of traces, the amount of energy lost to heat is quite prolific.
The various materials will absorb this heat and rise in temperature — how much so depends entirely on the material itself. Pure silicon needs less heat energy for its temperature to increase than the equivalent amount of pure copper.
Letting the chip’s temperature rise can generate a number of problems. Metals, for example, will increase in resistance as they get hotter, whereas semiconductors generally decrease in resistance. Overall, the electrical behavior of the device will alter, but engineers test their designs over a range of temperatures to figure out its operating range. Too much or too little, and the chip will run into serious problems.
One important reason for keeping the temperature under a certain limit in CPUs and GPUs is leakage. The processors running inside your PC contain billions of FinFET transistors (other components tend to use MOSFETs) and due to their microscopic dimensions, electric charge ends up going in places where it’s not supposed to go. When the switches are in their ‘off’ state, no current is supposed to be flowing, but leakage results in a tiny amount dribbling through.
Multiply this by a few thousand million and the end result is that the processor uses quite a lot more power than it’s supposed to. And unfortunately, the higher the temperature, the worse this problem gets — exponentially so.
This is why there is a clearly defined maximum temperature limit. You want to keep well away from a temperature that would lead to significant leakage, exacerbate other constant issues (e.g. electromigration) or cause physical problems (contact expansion, package damage).
What’s the maximum temperature for CPUs and GPUs?
When it comes to CPUs, there are a few things to keep in mind when looking at the highest temperatures they are limited to…
Intel has applied two temperature limits to their CPUs, for every model, for at least 10 years. The first is called Tjunction or Tj max. That is the maximum thermal junction temperature that a processor will allow before thermal control systems kick in to bring the heat back under control (also known as “thermal throttling”). This is done by lowering the clock speeds and in some cases, the voltages, too.
This temperature is effectively in the very middle of the chip itself, but modern processors will have multiple sensors dotted about the die to record this.
The second one, Tcase, is more of a target, rather than a hard limit, as it’s the highest temperature that the CPU package (the surface of the metal heat spreader) should ideally reach when used with a suitable cooler. The CPU won’t change anything when running at this temperature.
Measuring this temperature can be tricky, so motherboard manufacturers include one or two sensors in the CPU socket to try and estimate how hot the case is.
If we look at some example CPUs, we can see a clear pattern — despite increases in core count, clock speeds, and power usage, Intel continues to specify a Tcase of 72°C and a maximum Tjunction of 100°C.
AMD uses a similar system for its processors, though the maximum is often a little higher, at 105°C.
GPUs
Today’s GPUs tend to just have a single maximum temperature: Tjunction. This is because the die has direct contact with the cooling system, whereas CPUs and much older GPUs are encapsulated by a heat spreader or other material (hence, the need for a Tcase value).
Just like central processors, graphics cards have a number of sensors dotted around the die, and the reported temperature is usually just an average of the recorded values — the individual sensor reporting the highest value, sometimes 15 degrees or more than the rest of the chip, is declared as a “hot spot” figure.
Depending on the vendor and model, thermal throttling (the reduction of clock speeds and voltages) will normally be based on values from all of the temperature sensors, rather than the single highest value.
Since graphics chips tend to have power levels that are much higher than most CPUs, temperature limits are consequently lower. Nvidia’s GeForce RTX 4090, with its 450W TDP, has a maximum temperature of 88°C, for example. Other models are a little lower than this but they’re all mostly in a window of 80 to 90 degrees.
AMD GPUs often have higher Tjunction values than Nvidia’s; the recently launched Radeon RX 7900 XTX and some of its older models have limits of 110°C. There’s no one specific reason for this, more a case that this is how the engineers have designed the chips to operate.
This raises a simple question: are there any devices inside PCs that can cope with higher temperatures or is this the general limit?
