Junction temperature is the temperature measured at the hottest point on a GPU’s silicon chip, as opposed to the average temperature across the entire package. It’s the single most important thermal reading for understanding whether your graphics card is running safely, because it reflects the peak heat where the chip is working hardest. Most modern GPUs have multiple temperature sensors embedded in the die, and the junction (or “hotspot”) reading comes from whichever sensor reports the highest value at any given moment.
How It Differs From Other GPU Temps
When you open monitoring software like HWiNFO, GPU-Z, or Radeon Software, you’ll typically see more than one temperature reading. The most common are “GPU Temperature” (sometimes called edge or package temperature) and “GPU Hot Spot” or “Junction Temperature.” These are not the same thing.
The standard GPU temperature is measured by sensors near the edge of the chip. It gives you a general sense of how warm the die is overall. Junction temperature, by contrast, comes from sensors closer to the center of the die, where transistors are packed most densely and heat generation peaks. Under heavy load, the hotspot can run 10 to 25°C hotter than the edge reading. So if your edge temp shows 72°C, the junction might be sitting at 85 to 95°C depending on the workload, cooler design, and airflow.
This distinction matters because the GPU’s own thermal protection system uses the junction temperature, not the edge temperature, to decide when to throttle performance.
Why Junction Temperature Matters for Performance
Every GPU has a built-in thermal limit. When the junction temperature approaches that ceiling, the card automatically reduces the power it draws, which lowers clock speeds and, in turn, frame rates. This process is called thermal throttling, and it’s designed to prevent physical damage to the chip.
In practice, throttling shows up as sudden FPS drops, stuttering, or inconsistent performance during gaming or rendering. You’ll also see the core clock frequency dipping in your monitoring software, sometimes accompanied by a noticeable drop in power consumption. If the card can’t bring the temperature under control even after reducing clocks, it will eventually trigger a critical shutdown to protect itself.
Modern GPUs use aggressive boost algorithms that push clock speeds as high as thermal and power headroom allow. A cooler junction temperature means the card can sustain higher boost clocks for longer, giving you better and more consistent performance without any manual overclocking.
Safe Ranges by Manufacturer
NVIDIA and AMD set different thermal limits for their GPUs, and it’s worth knowing where your card falls.
NVIDIA’s RTX 40-series cards are rated for safe operation up to about 89°C on the package sensor, with the hotspot designed to handle up to around 100°C. In practice, staying below 80°C on the edge sensor during gaming keeps the card comfortably within its boost envelope. Once you’re consistently hitting 85 to 89°C on the edge reading, throttling becomes likely and it’s worth improving your cooling situation.
AMD takes a different approach with its RDNA 3 cards like the Radeon RX 7900 XTX. AMD’s junction temperature limit is 110°C, which looks alarming if you’re used to NVIDIA numbers but reflects a different thermal management philosophy. The card’s fans ramp up aggressively above 100°C, and clock speeds only start dropping if the junction actually hits 110°C and keeps climbing. A critical shutdown occurs around 120°C. Anything below 110°C on an AMD card is technically within spec and not throttling.
Typical Temperatures Under Load
For most modern GPUs regardless of brand, these ranges are a useful baseline:
- Idle: 30 to 45°C
- Gaming load: 65 to 85°C (edge), with junction running 10 to 25°C higher
- Throttling zone: 90 to 110°C junction, depending on the card
Keeping your card under 80°C on the edge sensor is a good target for long-term reliability and consistent boost clocks, even though higher temperatures are technically safe.
Memory Junction Temperature
There’s a separate thermal concern that often gets confused with GPU junction temperature: VRAM junction temperature. High-end cards using GDDR6X memory (like the RTX 3090 and RTX 4090) generate significant heat from the memory chips themselves, and this has its own set of limits.
GDDR6X memory is rated for a recommended operating range up to 95°C, with the thermal junction maximum around 100°C. Above that point, accelerated degradation of the memory begins. At 110°C, the memory enters a self-refresh mode that causes performance loss, instability, and error correction. At 120°C, immediate damage can occur. If you’re monitoring your card during heavy workloads like rendering or mining, keep an eye on the “Memory Temperature” or “VRAM Junction” reading separately from the GPU hotspot.
Common Causes of High Junction Temps
If your junction temperature is unusually high relative to the edge temperature (a gap of 30°C or more), the problem is often physical rather than environmental. The most common culprit is poor contact between the GPU die and the cooler’s heatsink. This happens when thermal pads on surrounding components are too thick, effectively acting as spacers that prevent the heatsink from pressing firmly against the chip itself.
This issue became well-known with the RTX 3080 and 3090, where users replacing thermal pads would accidentally use pads that were too thick (for example, 2mm instead of 1.5mm on the front side). The oversized pads lifted the heatsink just enough to break proper contact with the die, sending hotspot temperatures above 100°C within seconds of loading a game. The fix was compressing the pads or switching to the correct thickness, which immediately brought temperatures back to normal.
Other physical causes include dried-out or poorly applied thermal paste on the GPU die, uneven mounting pressure from the cooler’s screws, and leftover thermal paste around the chip socket interfering with the connection. On the backplate side, some cards have subtle changes in height across different VRAM chip positions, and failing to account for this with slightly different pad thicknesses can raise both GPU and hotspot temperatures by 10 to 15°C.
How to Lower Junction Temperature
The simplest steps don’t require opening the card. Adjusting your case fan curve to increase airflow, ensuring your GPU has at least a small gap of open space below it, and setting a more aggressive GPU fan curve in software like MSI Afterburner can each shave a few degrees. Undervolting the GPU is another effective option: reducing the voltage slightly at a given clock speed lowers heat output without meaningfully affecting performance, and many users report 5 to 10°C drops from a modest undervolt.
If software tweaks aren’t enough, repasting the GPU die with fresh high-quality thermal paste is the next step. On cards that are a year or more old, the factory thermal paste may have dried and lost conductivity. When doing this, check that the thermal pads on the memory chips and voltage regulators are the correct thickness and making firm, even contact. A visual check is easy: after disassembly, each pad should show a clear impression from the component it was touching. If the impression is faint or absent, contact was poor.
For cards with persistent hotspot issues, some users add small copper shims or switch to higher-conductivity thermal pads to improve heat transfer from the memory to the backplate. Mounting pressure also matters. Tightening the cooler screws evenly, in a cross pattern, ensures the heatsink sits flat against the die rather than tilting slightly to one side.