01 — What Is GPU Overclocking?
GPU overclocking is the process of pushing your graphics card's core clock and memory clock speeds beyond their factory defaults. Modern GPUs boost dynamically, meaning they already run faster than their base clock in real-time — overclocking extends this ceiling further.
The two primary levers are:
- Core Clock (GPU clock): Controls how fast the shader processors, rasterizers, and compute units run. A higher core clock directly increases frame rates in GPU-bound scenarios.
- Memory Clock (VRAM clock): Controls the speed of the video memory. Faster memory increases bandwidth, which matters significantly at higher resolutions (1440p, 4K) and in bandwidth-limited games and applications.
A well-executed overclock typically yields 5–15% more performance with no financial cost — just time and some patience. On a high-end GPU, that can translate to 10–20 extra frames per second in demanding titles.
02 — Tools You Need
You need three things: an overclocking utility, a temperature/stats monitor, and a stress testing tool. All of the following are free.
| Tool | Purpose | Works With |
|---|---|---|
| MSI Afterburner | Overclock utility — adjust core/memory clock, power limit, fan curve, voltage | NVIDIA + AMD |
| AMD Adrenalin | AMD's native OC tool — clean, built-in, good for AMD-only users | AMD only |
| HWiNFO64 | Real-time monitoring — temps, clock speeds, power draw, fan RPM | NVIDIA + AMD |
| 3DMark (Free) | GPU benchmark — use TimeSpy or FireStrike to stress test and measure gains | NVIDIA + AMD |
| Unigine Superposition | Intense GPU stress test — pushes harder than most games | NVIDIA + AMD |
| GPU-Z | Detailed GPU info — verify clock speeds, TDP, VRAM details | NVIDIA + AMD |
03 — Safe Overclocking Temperature Targets
The most important rule in GPU overclocking: the goal is performance, not heat. If your overclock produces temperatures that are uncomfortably close to TJ Max, it's not a good overclock — it's a throttling overclock.
| GPU Family | TJ Max | Target OC Temp (Core) | Memory Temp Limit |
|---|---|---|---|
| RTX 40 Series (Desktop) | 89°C | ≤ 80°C | ≤ 90°C (GDDR6X) |
| RTX 30 Series (Desktop) | 93°C | ≤ 83°C | ≤ 90°C (GDDR6X) / ≤ 95°C (GDDR6) |
| GTX 10 / 16 Series | 91–97°C | ≤ 83°C | Not typically reported |
| AMD RX 7000 Series | 110°C | Edge ≤ 85°C | Hotspot ≤ 100°C |
| AMD RX 6000 Series | 110°C | Edge ≤ 85°C | Hotspot ≤ 100°C |
GDDR6X Memory Junction Temperature: On cards with GDDR6X VRAM (RTX 3080, 3080 Ti, 3090, 3090 Ti, 4080, 4090), HWiNFO64 will show a "GPU Memory Junction Temperature" sensor. This can run 20–30°C above the core temperature. Keep it below 100°C during overclocked loads. This is the primary limiting factor for memory overclocks on these cards.
04 — Step-by-Step Overclocking Process
Baseline Your GPU
Before touching any settings, run a benchmark (3DMark TimeSpy) and record the score, peak GPU temperature, and peak power draw. This is your baseline. Open HWiNFO64 → Sensors to monitor temps in real time during the run.
Increase Power Limit First
In MSI Afterburner, raise the Power Limit slider to its maximum (usually +10 to +25% depending on card). This alone can free up several MHz of boost clock that the card was previously throttling to stay within TDP. Run your benchmark again — some cards gain 2–5% here with zero other changes.
Set Fan Curve
Before any clock changes, set a more aggressive fan curve. In Afterburner, click the Fan icon and enable "Define your own fan profile." A good starting curve: reach 70% fan speed at 70°C, 85% at 80°C. This ensures the card can sustain higher clocks without thermal throttling.
Overclock Core Clock — Incrementally
Start with a +100 MHz core clock offset in Afterburner. Click Apply, then run Superposition for 10 minutes or play a demanding game for 20–30 minutes. If stable, increase by +25 MHz and repeat. Keep going until you hit a crash, artifact (visual glitch), or driver timeout. Then back off by 25–50 MHz — that's your stable core overclock.
Overclock Memory Clock — Separately
With core stable, reset core to stock and start pushing memory. Increase in +100 MHz steps for GDDR6X or +200 MHz steps for GDDR6. Artifacts often appear as texture flickering or coloured pixels before a crash — these are your warning signs. Back off 100–200 MHz from the point of artifacts.
Combine and Final Stability Test
Apply your stable core + memory overclock together with the raised power limit. Run a 30-minute Superposition stress test at maximum settings, then run 3DMark TimeSpy twice to confirm the score is repeatable. Check that peak temperatures stayed within safe targets. If stable through all of this, your overclock is solid.
Save the Profile
In MSI Afterburner, save your final settings to a profile slot (1–5 in the interface). Enable "Apply overclocking at system startup" so the OC loads automatically. Use the hardware buttons on the Afterburner interface or a keyboard shortcut to toggle profiles on/off.
05 — Stress Testing Your Overclock
An overclock that crashes once in 100 hours isn't stable — it will crash at the worst time. Proper stress testing gives confidence before relying on an OC daily.
| Test | Duration | What It Catches | Severity |
|---|---|---|---|
| Superposition (1080p Extreme) | 20–30 min | Core and memory instability, sustained temps | High |
| 3DMark TimeSpy Loop | 5 runs | Score consistency, crash under repeated load | Medium |
| FurMark (burn-in) | 10–15 min | Extreme thermal stress — unrealistically harsh | Very High |
| In-game (demanding title) | 45–60 min | Real-world stability in actual usage patterns | Realistic |
06 — Signs of an Unstable Overclock
- Driver timeout / black screen recovery — Windows "Display driver stopped responding and has recovered" — your most common instability signal. Reduce core or memory by 25–50 MHz.
- Texture artifacts / pixel flickering — coloured dots, striped patterns, or texture errors. Usually a sign of memory OC being too high. Back off memory immediately.
- Hard crash / BSOD — blue screen or complete system freeze. Usually indicates too aggressive a core clock or power limit. Reduce core offset by 50 MHz.
- Lower benchmark scores under OC — if your score drops compared to stock, the GPU is likely thermal throttling under the overclock. Check peak temperatures in HWiNFO64.
- Random game crashes that weren't happening before — the most common user-reported sign. Don't ignore these; track when they started relative to OC changes.
07 — Per-GPU Recommended Settings
These settings represent well-tested, conservative starting points based on community benchmarks and hardware reviews. Due to the silicon lottery, individual results will vary — some cards can push further, others less. Use these as a safe baseline, not a guarantee.