I almost RMA’d a perfectly fine GPU because I didn’t understand what hotspot temperature actually measures. GPU hotspot temperature is the single hottest point on your graphics card’s die, not the average core temp most people are used to checking, and that distinction matters more than you’d think once you start chasing 90°C readings.
So before you panic and start shopping for a new card — let’s get the monitoring part right first, then figure out whether your number is actually a problem.
Quick Answer
- Hotspot (junction) temperature is the hottest spot on the GPU die, measured separately from the average core/edge temperature.
- A 90°C hotspot reading is normal on many modern GPUs under full load — it’s the delta between hotspot and edge that usually signals a real problem.
- A gap bigger than 15–20°C between hotspot and edge usually means a thermal paste or pad contact issue.
- HWInfo64 and MSI Afterburner (with RivaTuner Statistics Server) are the most reliable free tools for tracking this sensor.
- Repasting and repadding fixes the issue more often than any case fan upgrade does.
Why It Matters (and Why “90°C” Isn’t One Number)
This is where a lot of guides get sloppy, so let’s not repeat that mistake.
Hotspot and edge temperature are not the same sensor. “Edge” or “core” temperature is what GPUs have reported for years — basically an average reading across the die. Hotspot (sometimes labeled “GPU Hot Spot” or “junction temperature”) is a newer sensor that tracks the single hottest point, which is almost always over the GPU core itself rather than the memory or VRM area. Nvidia added this with the Ampere (RTX 30-series) generation; AMD has had something similar on RDNA2 and RDNA3 cards.
Different GPUs have different safe ceilings. This is the part that trips people up. RTX 30 and 40 series cards are generally engineered to handle hotspot temperatures in the 95–110°C range without damage — that’s within spec, not a sign of failure. AMD RDNA2/3 cards have their own junction temp limits, often a bit higher tolerance than you’d expect from the number alone. So seeing “90°C” doesn’t automatically mean something’s wrong. What’s actually informative is the gap between hotspot and edge, not the raw hotspot number by itself.
A wide hotspot-to-edge delta usually points to a contact problem. If your edge temp sits at 65°C but hotspot spikes to 95°C, that ~30°C gap is the real story — it means heat isn’t transferring evenly from the die to the cooler. Thin, dried, or unevenly applied thermal paste is the most common cause. This became a well-known issue on some RTX 3080/3090 Founders Edition and AIB cards a few years back, where factory thermal pads were slightly too thin to make full contact with the VRAM and VRM, and a similar effect happens with paste that’s degraded after a couple years of thermal cycling.
Airflow and dust matter, but less than people assume. A dusty case fan or blocked intake will raise both edge and hotspot temps somewhat evenly. If only hotspot is spiking while edge stays reasonable, dust usually isn’t the main culprit — that’s a paste/pad story, not an airflow story.
Common Scenarios
- Newly built PC, high hotspot from day one: often a factory paste application issue, especially on cards bought used or from a batch with known QC problems.
- 2–3 year old card, hotspot climbing over time: classic paste degradation — pump-out or dry-out is normal wear, not a defect.
- Laptop GPUs: hotspot readings run higher in general due to compact cooling, and the “normal” ceiling is often higher than desktop cards of the same chip.
- Undervolted or factory-overclocked cards: aggressive factory clocks can push voltage and heat density up specifically at the hotspot, even when the cooler itself is fine.
Hotspot vs Edge: What the Numbers Are Telling You
| Reading Pattern | Likely Meaning | Action |
|---|---|---|
| Edge 60–70°C, Hotspot 85–95°C (normal delta ~15-25°C) | Normal for many modern GPUs under load | No action needed |
| Edge 65°C, Hotspot 100°C+ (delta 30°C+) | Thermal paste/pad contact issue | Repaste/repad |
| Edge 80°C+, Hotspot also high, small delta | Airflow/dust/ambient temp issue | Clean case, improve airflow |
| Hotspot fluctuating wildly second to second | Sensor glitch or unstable VRM, less commonly seen | Update drivers/BIOS first, monitor again |
Not every card behaves identically here — laptop GPUs especially don’t fit neatly into the desktop delta guidelines above, so treat that table as a desktop-card starting point, not gospel.
How to Actually Monitor It
Step 1: Install HWInfo64
This is the most complete free option and the one I keep coming back to. Run the Sensors-only mode, scroll to your GPU section, and look specifically for “GPU Hot Spot” or “Hot Spot Temperature” — not just “GPU Temperature,” which is the edge reading.
Step 2: Set up MSI Afterburner with RTSS for an on-screen overlay
Afterburner alone often shows edge temp by default. You’ll need to dig into the monitoring tab, find the hotspot sensor (sometimes buried under “show in On-Screen Display” checkboxes), and enable it specifically. It’s not always obvious which checkbox is which — I had to hover over half a dozen sensor names before finding the right one.
Step 3: Run a sustained load test, not a quick benchmark
Hotspot issues often don’t show up in a 2-minute benchmark run. Use something like a 15–20 minute loop of a demanding game or a stress test (FurMark, OCCT) and watch the trend, not just the peak. Paste/pad issues tend to get worse the longer the card runs hot.
Step 4: Log the session and check the delta after, not just live numbers
Both HWInfo and Afterburner can log to CSV. Watching live numbers while gaming is distracting and you’ll miss spikes — log it, then open the file afterward and check max hotspot, max edge, and the gap between them.
What Actually Worked For Me
My case had decent airflow, decent fan curve, nothing obviously wrong — but my hotspot was sitting around 98°C while edge stayed at a calm 68°C. That’s a big delta, and I knew enough to be suspicious, but my first move was dumb: I reseated the GPU in the PCIe slot, thinking maybe it wasn’t fully seated. Didn’t change a thing, which in hindsight makes sense since seating doesn’t affect the cooler’s contact with the die.
And then I tried updating drivers, on the off chance it was a sensor reporting bug. No change there either. What actually fixed it was repasting — opened the cooler, found the factory paste dried out and unevenly spread (more on one side of the die than the other, which is a more common factory flaw than people realize). Cleaned it off, applied a proper X-pattern of decent thermal paste, and the delta dropped to about 18°C on the next stress test. Not perfect, but a lot more reasonable.
That’s not entirely the full story though — I’d also read a forum comment suggesting some cards need the mounting screws tightened in a specific diagonal order for even pressure, which I hadn’t done the first time around. Did that on the reseat, and it’s hard to say how much that contributed versus the paste itself, but the combination got me there.
Advanced Fixes and Edge Cases
Checking VRM and VRAM pad contact, not just the GPU die paste. Some cards use thermal pads (not paste) over the memory chips and VRM, and these degrade or were undersized from the factory on certain models. If hotspot stays high after a clean repaste on the core, pull the cooler again and check pad thickness and coverage on the VRAM specifically.
BIOS/vBIOS power limit and fan curve checks. A locked or overly conservative fan curve in the card’s vBIOS can leave the fan ramping too slowly relative to hotspot spikes — by the time the fan responds, the hotspot has already overshot. Afterburner lets you set a custom fan curve tied more aggressively to temperature, which can mask (not fix) a marginal paste job.
Event Viewer / driver crash logs for thermal throttling correlation. If you’re also seeing random frame drops or driver resets, check Windows Event Viewer for “Display driver stopped responding” events and cross-reference the timestamps against your logged hotspot spikes. Sometimes what looks like a driver crash is actually thermal protection kicking in.
Repaste vs repad cost-benefit. Repasting is straightforward on most cards and low-risk if you’re reasonably careful. Repadding requires matching pad thickness precisely (too thick and you risk insufficient contact pressure on neighboring components, too thin and you’re back to the original problem) — this is the step people most often get wrong on a first attempt, so measure twice.
Prevention Tips
- Repaste every 2–3 years on cards you keep long-term, even if temps seem fine — paste degrades gradually before it becomes obvious.
- Don’t judge GPU health from edge temperature alone; check the delta periodically, not just once.
- Keep case intake filters clean — this helps overall thermals even though it’s rarely the root cause of a hotspot-specific spike.
- Avoid extreme thermal paste application techniques (too much paste pooling at the edges is almost as bad as too little in the center).
FAQ
Is 90°C hotspot dangerous on an RTX 4070 or similar? Not by itself, generally. Most RTX 40-series cards are rated for hotspot temps into the low-100s under sustained load. Check the delta against edge temp before assuming a problem.
Why does HWInfo show “Hot Spot” but my old GPU-Z doesn’t? Older GPU-Z versions or older GPUs without the junction sensor won’t expose this reading at all — update GPU-Z, and if it’s still missing, your specific GPU may not report it.
Will undervolting lower my hotspot temperature? Often yes, and it’s one of the lower-effort fixes worth trying before opening the card up. Lower voltage at similar clocks usually reduces heat density at the hotspot specifically.
Can a bad PSU cause high hotspot readings? Not directly — hotspot is a thermal/contact issue, not a power delivery issue, though a struggling PSU under load can cause unrelated stability problems that get confused with thermal symptoms.
Does repasting void my warranty? On most cards, yes, opening the cooler voids warranty coverage unless it’s a card explicitly designed for easy user maintenance. Check your specific manufacturer’s policy first.
Editor’s Opinion
People freak out over hotspot numbers way more than they need to, and honestly the marketing around “90°C bad!!” videos online doesn’t help. Check the delta, not just the raw number, before you start tearing your card apart. That said, if you do have a real delta problem, don’t just throw more fans at it — that’s the fix that gets recommended constantly and rarely actually solves anything, since the bottleneck is contact, not airflow.
