RTX 30/40 Series GPU Repasting: Why Your Card Runs Hot and What to Do About It

Written by Andrei Zamiralov, Electronics Repair Technician

Published: April 8, 2025  |  Updated: March 15, 2026

Your RTX 3080 hits 95°C on the hotspot within ten minutes of gaming. The fans sound like a desk fan aimed at your face. You haven't overclocked anything. So what's going on? For most RTX 30 and 40 series owners, the answer is the same: the factory thermal paste has either dried out, pumped out from the die, or was applied poorly from day one. A GPU repaste typically brings hotspot temperatures down by 15–20°C — and that difference is audible.

TL;DR: RTX 30/40 series GPUs are known to run hot due to dried or poorly applied factory thermal paste and undersized thermal pads. A professional repaste — using materials like Honeywell PTM7950 — typically drops GPU hotspot temperatures by 15–20°C (Igor's Lab, 2021). In Helsinki, the service costs €49–€89.

Why Do RTX 30/40 Series GPUs Run So Hot?

The RTX 30 series launched with a documented hotspot problem. Independent testing by Igor's Lab in 2021 showed that factory thermal paste pump-out on the GA102 die (used in RTX 3080, 3090, and 3090 Ti) could raise hotspot readings by up to 20°C within 12–18 months of normal use. The RTX 40 series improved cooler designs but did not fully resolve paste quality issues on flagship models.

There's an important distinction most guides skip. GPU "Core temperature" and "Hotspot temperature" are not the same sensor. The core reading is an average across the entire die. The hotspot is the single highest-temperature point, typically the center of the die where power density peaks. A card showing 72°C core can simultaneously show 104°C hotspot — and that hotspot is what triggers fan ramp-up and boost clock reduction.

GDDR6X memory compounds the problem. Cards like the RTX 3080, 3090, 4080 Super, and 4090 use GDDR6X chips that operate at significantly higher voltages than standard GDDR6. According to Micron's GDDR6X technical documentation, these chips are rated for junction temperatures up to 95°C — but sustained operation near that ceiling accelerates long-term degradation. Poor or dried thermal pads between the VRAM chips and heatsink push memory temps toward that limit during extended gaming sessions.

How Do You Know When Your GPU Needs Repasting?

A general rule used in professional repair: any GPU over 3 years old is a repaste candidate, and any card showing hotspot temperatures above 85°C at stock clocks should be assessed immediately. A 2022 community survey by TechPowerUp forum members found that 68% of RTX 3080/3090 owners reported hotspot temps above 90°C after 18 months of regular use — well before the typical 3-year threshold.

Practical signs to watch for:

• Hotspot temperature consistently above 85°C during gaming
• GPU fans ramping to 80–100% speed within the first few minutes of load
• Boost clocks dropping below the card's rated speed (visible in GPU-Z or HWiNFO)
• The card running noticeably louder than it did when new
• System crashing or throttling during rendering or gaming sessions

Boost clock reduction is the most financially relevant symptom. Your RTX 4090 cost over €1,000. If it's throttling from 2520 MHz down to 2100 MHz because the hotspot is too high, you're losing meaningful performance you paid for. A repaste restores what the card was always capable of.

What Thermal Paste Actually Works for GPU Repasting?

Paste choice matters more for GPUs than for CPUs, because GPU dies have less uniform contact surfaces and far more adjacent components to protect. Thermal Grizzly's published conductivity data shows Conductonaut (liquid metal) at 73 W/m·K, Kryonaut Extreme at 14.2 W/m·K, and Honeywell PTM7950 at approximately 10.5 W/m·K — but real-world GPU temperature deltas between Kryonaut Extreme and PTM7950 are typically under 3°C.

We've found that PTM7950 is the most practical choice for the majority of RTX 30/40 series cards. It's a phase-change material — solid at room temperature, it liquefies slightly under heat and fills microscopic surface irregularities extremely well. Unlike liquid metal, it carries no electrical conductivity risk if it migrates beyond the die edge. That matters on cards where exposed copper traces run close to the GPU package.

Liquid metal (Conductonaut, Thermal Grizzly) is worth considering for RTX 3090 and 4090 owners chasing maximum temperature reduction, but only when the technician has confirmed there are no unprotected SMD components near the die edge and has applied appropriate masking. The conductivity risk is real. One stray droplet bridging two pads is enough to destroy the card.

What About Thermal Pads — Why Do They Matter?

Thermal paste covers only the GPU die itself. Everything else — VRAM chips, VRM MOSFETs, inductors — relies on thermal pads to transfer heat to the heatsink baseplate. According to Thermal Grizzly's product specifications, stock thermal pads on many AIB RTX 3080/3090 designs measure 0.5–1.0 mm but the actual gap may require 1.0–1.5 mm pads for full contact, leaving a measurable air gap from the factory.

Getting pad thickness wrong during a DIY repaste is the most common and costly mistake. Too thin, and the pad compresses but doesn't bridge the gap — the chip still runs hot. Too thick, and the pad pushes the heatsink away from the GPU die itself, making the main paste contact worse. Each card model has specific pad thickness requirements; an RTX 3080 Founders Edition uses different measurements than an ASUS TUF RTX 3080.

RTX 3080 and 3090 Founders Edition cards shipped with VRAM thermal pads measured at 0.5 mm, while the physical gap between VRAM and heatsink on some units measured 1.0–1.5 mm, creating inadequate contact and elevated memory temperatures. Replacing stock pads with correctly-specced alternatives (Gelid GP-Ultimate or Thermal Grizzly Minus Pad 8) typically reduces VRAM temperatures by 8–12°C. (Igor's Lab, 2021)

DIY vs. Professional Repaste — Where Does the Risk Lie?

A CPU repaste is low-risk — remove cooler, clean, apply paste, reseat. GPU repasting involves far more variables, and mistakes can be expensive. The table below compares the main risk areas between DIY and professional GPU repasting for RTX 30/40 series cards.

Risk Factor	DIY Repaste	Professional Repaste
Correct thermal pad thickness	Requires card-specific research; easy to get wrong	Verified per card model before disassembly
Liquid metal migration risk	High — exposed pads easily bridged	Mitigated with masking; liquid metal only where appropriate
SMD component damage	Moderate — small capacitors near cooler screws	Low — correct tools and experience reduce risk
Warranty status	Voided on opening	Voided on opening (same outcome)
Post-service temperature verification	Self-monitored; no baseline comparison	Before/after temperature report provided
Typical cost	€15–€30 (paste + pads)	€49–€89 (Helsinki)

The warranty point is worth emphasising: opening a GPU cooler voids the manufacturer warranty regardless of who does it. If your card is still within its 2–3 year warranty window and is genuinely defective, contact the retailer or manufacturer first. For out-of-warranty cards, the repaste cost is minimal compared to the performance and longevity gains.

What Does a Professional GPU Repaste Include?

A thorough GPU repaste is more than swapping paste. The process should also address the heatsink itself, which accumulates dust in the fin stack that paste replacement alone won't fix. Professional services in Helsinki typically run €49–€89 for RTX 30/40 series cards, with the higher end covering complex flagship cards like the RTX 3090 Ti or 4090 that have more VRAM chips and VRM stages requiring individual pad sizing.

A complete service covers:

1. Full disassembly — cooler, backplate, and shroud removed; all screws and bracket hardware catalogued
2. Heatsink cleaning — compressed air and brushing to clear fin-stack dust, the primary source of ambient temperature rise in older cards
3. Old paste removal — isopropyl alcohol (90%+) and lint-free applicators; no abrasives near the die surface
4. Pad measurement and replacement — each VRAM and VRM pad measured and matched to card-specific specs before replacement
5. Paste application — PTM7950, Kryonaut Extreme, or liquid metal (where appropriate) applied in the correct quantity and pattern
6. Reassembly and testing — card run under load for 30 minutes; before/after hotspot and VRAM temps recorded and shared with the customer

Frequently Asked Questions About GPU Repasting

If your RTX 30 or 40 series card is running hot, noisy, or throttling, the fix is usually straightforward. Dried paste and wrong-spec thermal pads are addressable problems — not signs of a failing GPU. A repaste at the right time extends the card's usable life and restores the performance you paid for.