This guide covers what the two readings mean, what counts as a worrying gap, and what to check before you open the card.
Two numbers, one die
Rozario describes the regular GPU temperature as being derived from thermal sensors spread across the die. By his account, Nvidia cards report an average of those sensors and AMD cards report an edge reading. The hotspot is the highest value any sensor on the die detects. Vendors name and calculate these sensors differently, so treat the Nvidia-versus-AMD split as the author's description, not a universal rule. Other guides use similar wording. One describes the core or edge reading as an average and the hotspot, also called junction, as the single hottest point the card's internal sensors can find.
A hotspot that reads higher than the core is normal, because some parts of the die generate more heat than others. Rozario says a hotspot sitting 10–15°C above the regular reading is nothing to worry about. His example of a worrying case is 70°C regular against 95°C hotspot, a 25°C gap, as seen in HWiNFO64. He says anything above 20°C deserves monitoring, especially if it has grown over time.
Those figures are the author's rules of thumb, not manufacturer specifications. Other sources draw the line in different places:
| Source | Delta it treats as a concern |
|---|---|
| XDA Developers | Above about 20°C, especially if growing |
| Steam community thread | Above 20°C |
| TechFuelHQ | Over 20°C on Nvidia or 30°C on AMD |
| How-To Geek (RX 9070 XT owner) | Consistently around 30°C or higher, varying by GPU |
One How-To Geek writer says a consistent gap around 30°C or higher, which varies from GPU to GPU, is a strong reason to look at the thermal interface. TechFuelHQ's guide goes further and says that a delta over 20 °C on NVIDIA or 30 °C on AMD points to dried, pumped-out paste. That is a firm claim from a secondary source, and I wouldn't treat it as settled. The sensible takeaway is that no single universal threshold exists. Your card's own baseline, and how the gap changes, tell you more than any one number.
Why paste can cause a wide gap
Thermal paste fills microscopic gaps between the GPU die and the cooler's cold plate. Rozario names two ways it deteriorates: it can dry out, and it can suffer "pump-out". In pump-out, repeated heating and cooling cycles gradually push paste away from the center of the die toward the edges. That leaves patches with poor contact.
How-To Geek describes the mechanism in more detail. The die and heatsink expand and contract at different rates, which can gradually push conventional paste out from between the surfaces. The author of that piece saw a gap of roughly 45°C on an RX 9070 XT. That is one user's experience, not a statistical finding.
The logic of the symptom is that a localized contact problem hurts the hottest spot first. The broader reading may rise only a few degrees, while the hotspot climbs sharply. Rozario's point about trends follows from this. A gap that has widened over months or years is more telling than a single snapshot.
If you never recorded a baseline when the card was new, you can start one now.
Limits of the diagnosis
A big delta is a clue, not proof. Rozario himself says mounting pressure and uneven contact can produce the same picture, especially on a card that has already been repasted. Some guides blame mounting pressure loss or pad problems alongside paste.
Absolute temperature matters too:
- Thermal limits differ by model. Rozario says an alarming reading on his RTX 4090 could be fine on another card. He cites the AMD RX 9070 XT as able to reach 110°C hotspot before throttling. I could not confirm that figure against an official AMD specification, so check your own card's documented limits.
- Cooler design matters. A large triple-fan cooler and a compact dual-fan one will behave very differently.
- Symptoms count. Throttling, louder fans, crashes, or falling performance matter more than a delta in isolation. Rozario notes that a hotspot near the limit can cause louder fans or throttling even when the regular reading looks fine.
A sensible troubleshooting order
This sequence follows Rozario's advice, with some practical measurement guidance added.
- Get a tool that shows the hotspot. Rozario says MSI Afterburner doesn't show it and uses HWiNFO64 for the reading. Software support can vary by GPU generation, so confirm that the sensor your tool shows is valid for your card.
- Measure like for like. Log the regular and hotspot readings during the same sustained workload, with similar room and case conditions. A single spike, or readings from different games, won't give you a reliable comparison. This is general measurement practice, not a vendor-prescribed test.
- Note the baseline and recheck periodically. Look for a widening gap over time.
- Rule out the basics. Rozario recommends checking dust buildup, fan operation, and case airflow first. Dust or poor airflow typically raises both readings together, not just the hotspot.
- Consider a mild undervolt. He suggests it as an easier step than disassembly. Treat it as a way to reduce heat, not a repair for poor contact.
- Check warranty terms before opening the card. Rozario warns that prying open a GPU can void your warranty. Policies vary by maker, so check yours.
- If you repaste, choose carefully. Rozario advises using a paste resistant to pump-out, or the problem can return. How-To Geek's author reported a lower maximum gap of 27°C after switching to Arctic MX-7, a paste formulated to resist pump-out. That is one anecdotal result, not a guarantee.
Do you need to repaste?
Probably not just because the hotspot looks high. Plenty of modern cards run hot hotspots by design. Rozario says not to obsess over the number the way you might with core temperature, but to check now and then as the card ages. A hotspot that creeps up while the regular reading stays put is the pattern worth acting on.
If the gap is large, growing, and accompanied by throttling after you've cleaned and checked airflow, then contact or paste becomes a reasonable suspect. At that point, consult the card maker or board partner, especially if the card is under warranty.
Bottom line
- The regular GPU temperature and the hotspot measure different things, and a gap is expected.
- A persistent or growing gap is a useful early warning, but the thresholds quoted online vary and aren't official.
- Paste pump-out is a plausible cause, not the only one.
- Rule out dust, fans, and airflow first, and check your card's own limits before reaching for a screwdriver.