Your GPU hits 85°C mid-game. Is that fine or is your hardware about to throttle? You check your CPU temperature and see 95°C during a render. Should you be worried? Most temperature guides either give you a generic “under 80°C is safe” rule that doesn’t apply to modern hardware, or a wall of chip-specific specifications that doesn’t help you understand what you’re actually looking at.this how to monitor gpu temperature guide walks through all the steps you need to know
Here’s the complete picture what to install, what to watch, what the numbers mean for specific hardware, and when to actually act on what you see.
The tools you need — and which one to actually use
Three tools cover temperature monitoring for most builders. You don’t need all three.
HWiNFO64
is the one to install first and the one to keep. Free, no installation bloat, updated regularly, and it reads more sensor data than any competing tool. It shows CPU package temperature, individual core temperatures, CPU hotspot, GPU temperature, GPU hotspot, VRAM temperature, VRM temperatures, fan speeds, power draw, and dozens of other values simultaneously. The sensors-only mode gives you a clean table of everything your hardware is reporting. For monitoring while gaming or during stress testing, HWiNFO64 is the definitive tool.
MSI Afterburner with RivaTuner Statistics Server
is the correct tool when you want temperatures overlaid on screen during gameplay. Configure the overlay in Afterburner add GPU temperature, GPU hotspot, CPU temperature, GPU usage, frame rate and it appears in the corner of your screen while you play. You don’t need to alt-tab. You see the numbers in real time as the game runs. The combination is free and the RTSS overlay works with virtually every game.
GPU-Z
is useful for a quick check of your GPU’s specifications, current temperature, and sensor readings without the full scope of HWiNFO64. If you want to verify your GPU is recognised correctly and check its baseline temperature at idle, GPU-Z is the fastest tool for that specific task.
Install HWiNFO64 for general monitoring. Add MSI Afterburner and RTSS if you want in-game overlays. You don’t need anything else.
CPU temperatures — what’s normal and what isn’t
Modern CPUs are designed to run hot. This is the most important thing to understand before interpreting temperature readings. AMD and Intel both design their chips to boost aggressively right up to their thermal limits the silicon is most efficient there. Seeing high temperatures under load is not automatically a problem.
AMD Ryzen 9000 series (Zen 5): AMD specifies a maximum junction temperature (TJ Max) of 95°C for Ryzen 9000 chips including the 9800X3D. The CPU boosts to its maximum clock speed right up to that temperature. Under sustained all-core workloads like rendering or compiling, hitting 90–95°C is normal and expected. Under gaming loads where not all cores are simultaneously at maximum most Ryzen 9000 chips settle in the 70–85°C range depending on the cooler.
The 9800X3D specifically runs differently. The 3D V-Cache stacked on top of the compute die creates a thermal constraint AMD limits the voltage and clock speed to protect the cache stack. This means the 9800X3D often runs cooler than a non-X3D chip despite its game performance advantage. 70–80°C under gaming load on a 9800X3D with a quality cooler is typical.
AMD Ryzen 7000 series (Zen 4): Same 95°C TJ Max. Same boosting behaviour. Same normal operating range of 70–90°C under gaming, potentially higher under full sustained workloads.
Intel Core Ultra 200S (Arrow Lake Refresh): Intel’s TJ Max on these chips is 100°C. Under gaming loads 60–80°C is typical depending on the cooler. Under sustained all-core workloads where power limits haven’t been set, these chips can run hot quickly check BIOS power limit settings if CPU temperatures are unexpectedly high under productivity workloads.
The number that actually matters — CPU hotspot vs package temperature
HWiNFO64 shows multiple CPU temperature readings. The one that matters most is the CPU package temperature or, on AMD platforms, the Tdie reading. This represents the overall chip temperature as AMD or Intel measures it for thermal protection purposes.
Individual core temperatures vary some cores run hotter than others depending on which workload is assigned to them. The maximum core temperature is worth watching but the package/Tdie reading is the primary thermal indicator.
If temperatures are consistently hitting TJ Max under normal gaming loads 95°C for AMD or 100°C for Intel something is wrong. Check thermal paste application, cooler mounting pressure, and case airflow before assuming the cooler is undersized.
GPU temperatures — what the numbers actually mean
GPUs have two temperature readings that mean different things and both matter.
GPU temperature is the average temperature across the die. This is the number most people look at. Under gaming load on most current cards a GPU temperature of 65–83°C is normal. Each card has a thermal throttle point the temperature above which it reduces clock speed to protect itself. Most current Nvidia and AMD cards throttle at 83–87°C depending on the specific card and AIB cooler design.
GPU hotspot temperature is the highest temperature recorded at any point on the die. It’s always higher than the average GPU temperature typically 15–25°C higher. On an RTX 5070 with an average GPU temperature of 72°C, the hotspot might be reading 92°C. This is normal. Nvidia specifies the RTX 50 series hotspot limit at 110°C if you’re under that, the card is operating within specification regardless of what the hotspot number looks like at first glance.
The hotspot is where thermal throttling actually begins. Watch both readings but use the hotspot to assess whether the card is approaching its thermal limit. Average GPU temperature alone doesn’t tell the complete story.
VRAM temperature is the third GPU reading worth monitoring. On cards with GDDR6 or GDDR7, VRAM has its own thermal sensors. GDDR6 memory is typically rated to 95°C. GDDR7 runs hotter by design Nvidia specifies the RTX 5090’s GDDR7 operating temperature up to 110°C. High VRAM temperatures with normal GPU temperatures suggest restricted airflow specifically over the memory chips which can happen in cases with poor side ventilation or with specific AIB cooler designs.
Normal operating ranges by component
| Component | Idle | Gaming load | Concern threshold |
|---|---|---|---|
| AMD Ryzen 9000 CPU | 35–50°C | 70–90°C | Consistently above 95°C |
| Intel Core Ultra 200S | 30–45°C | 60–80°C | Consistently above 95°C |
| GPU average (Nvidia RTX 50) | 30–45°C | 65–83°C | Above 87°C sustained |
| GPU hotspot (Nvidia RTX 50) | 45–65°C | 80–95°C | Above 110°C |
| GPU average (AMD RX 9000) | 30–45°C | 60–80°C | Above 85°C sustained |
| VRAM (GDDR7) | 40–55°C | 70–95°C | Above 105°C |
Setting up an in-game temperature overlay
Install MSI Afterburner and RivaTuner Statistics Server both install together from the Afterburner download page. Open Afterburner → Settings → Monitoring tab. Find GPU Temperature in the list and tick the Show in On-Screen Display checkbox. Do the same for GPU Hotspot Temperature, CPU Temperature, and GPU Usage. Click OK.
Open RivaTuner Statistics Server from the system tray and ensure On-Screen Display support is set to On. Launch any game. The overlay appears in the corner of your screen by default position and size are adjustable in RTSS settings.
During your first gaming session with the overlay active, note the temperatures at typical load 20–30 minutes into a demanding session rather than at the very start. Initial temperatures are lower because the components haven’t fully stabilised. The settled temperatures after 20 minutes are the representative values for your system.
When to act on what you see
CPU temperature at 90–94°C under gaming normal for Ryzen, no action needed unless it’s throttling (check HWiNFO64 for a CPU Throttling sensor turning true).
CPU temperature consistently hitting 95°C under gaming with frames dropping the CPU is throttling. Check thermal paste, cooler mounting, and case airflow. Refer to the thermal throttling guide on SpecClear for the full diagnostic process.
GPU temperature above 87°C sustained with frame rate inconsistency fans may be thermally limited, dust may be blocking the heatsink, or the GPU may need repasting if it’s over two years old.
GPU hotspot above 105°C investigate. This is approaching the junction limit even if the average temperature looks acceptable. Check case airflow and GPU fan speed.
VRAM temperature above 100°C GPU needs better airflow or the thermal pads between VRAM chips and the heatsink may need replacing. This is more common on cards over two to three years old.
Everything within the normal ranges do nothing. The system is working correctly. Monitoring temperatures regularly prevents problems from becoming expensive by catching the slow creep before it becomes thermal throttling you notice in frame rates.