Two CPUs with the same model number can run at noticeably different temperatures and voltages. That difference is the thermal lottery. You test it by recording how your chip behaves at stock settings, then lowering its voltage in small steps and stress testing each one until it stops being stable. The lowest voltage your CPU holds without crashing, and how much that drops your temperature, tells you whether you got a good sample or an average one.
The whole process takes an afternoon, costs nothing, and needs three free tools plus a bit of patience. Here is how to do it properly and how to read what you get.
What the thermal lottery actually means
Every CPU starts as part of a large silicon wafer that gets sliced into many small chips. No two chips come out identical, even from the same wafer, because tiny variations creep in during manufacturing. This natural variation is called process deviation, and it changes how much voltage a chip needs to run at a given speed.
A chip that needs less voltage produces less heat and runs cooler. A chip that needs more voltage runs hotter at the same speed. Since heat and voltage are linked, a “good” sample in the thermal lottery is really a chip that does its job at lower voltage. Testing the thermal lottery means finding out where your chip sits on that scale.
Lower voltage means less heat. So the thermal lottery and the silicon lottery are two views of the same thing: a chip that needs less voltage wins both.
What you need before you start
You only need three pieces of free software and a stable, fully built PC. Get these ready first:
- HWiNFO64 reads your CPU temperature, voltage and power in real time. It is the gauge you watch the whole way through.
- A benchmark is a repeatable load that heats the CPU the same way every run, so you can compare results fairly. Cinebench 2024 or R23 both work.
- A stress test is a tougher torture test for confirming stability once you think you have found a setting. Prime95 (Small FFTs) or OCCT both push the chip hard.
You also need a way to change voltage. On most desktops the cleanest place is the BIOS. Intel users can also use Intel Extreme Tuning Utility (XTU) inside Windows, and AMD users can use Ryzen Master. Write down your numbers as you go, because the test only makes sense if you can compare before and after.
A quick note on laptops: many of them lock voltage control at the firmware level, especially newer models, so this full test is mainly a desktop job. If you are on a laptop and the voltage options are greyed out or missing, your machine is simply not built to allow it, and that is not something you can force safely.
How to test the thermal lottery, step by step
The test has one goal: find the lowest voltage your CPU runs at while staying stable, and see how cool it gets there. Work through these steps in order and change only one thing at a time.
- Set a baseline. Leave everything at stock. Open HWiNFO64, run your benchmark, and note three numbers at full load: the peak temperature, the core voltage (often labelled Vcore or VID), and the package power in watts. This is your reference point.
- Lower the voltage a little. Drop voltage by a small amount. A safe first step is around -50mV as an offset, or roughly -10 to -15 on AMD’s Curve Optimizer scale. Apply it.
- Run the benchmark again. Repeat the same benchmark and watch HWiNFO64. If it finishes without errors or crashes, note your new temperature and power. You should see both fall.
- Repeat in small steps. Keep dropping voltage in 10mV to 25mV steps (or a few more steps of Curve Optimizer), rerunning the benchmark each time. Go slowly. Small steps are how you find the true edge without wasting hours.
- Find the failure point. Eventually the system will hang, crash, throw a benchmark error, or blue screen. That voltage is too low. Step back up to the last setting that passed.
- Confirm with a long stress test. Run Prime95 or OCCT for 30 to 60 minutes at your last stable setting. A benchmark can pass while a harder load fails, so this step catches settings that only looked stable.
- Add a safety margin. Once you have your lowest stable voltage, add a small cushion back, about 10mV to 20mV. This keeps the chip stable on cold mornings and as it ages over the years.
When you are done, compare your final numbers against the baseline. The size of the voltage drop, and the temperature fall that came with it, is your thermal lottery result.
While you test, keep one eye on HWiNFO64’s load readings rather than idle ones. The numbers that matter are the peak temperature, the average effective clock speed, and the package power, all measured during the benchmark. If your clock speed climbs after an undervolt while the temperature drops, that is the best possible sign: your chip is doing more work for less heat, which is the whole reason enthusiasts bother with this.
What a good chip looks like, and what average looks like
There is no official scorecard, so the best way to judge your result is the size of the drop and the temperature you land at. As a rough guide:
| Result | Voltage you could remove | What it tells you |
| A strong sample | Large offset, often beyond -100mV, while staying stable | Your chip needs little voltage and runs clearly cooler. A genuine win. |
| An average sample | A modest offset, roughly -40mV to -70mV | Normal. Most chips land here, and you still gain a few degrees. |
| A weak sample | Little or no room before it crashes | Your chip needs its voltage. Not a fault, just the luck of the draw. |
Real results from this kind of testing show the spread clearly. A modest -50mV undervolt often cuts power by about 10 watts and drops temperatures 3 to 4 degrees Celsius. A chip that takes a deeper offset can shed 15 to 20 degrees and move its throttling point well clear of the danger zone. Two chips with the same model number can sit at opposite ends: one i7-14700K might hold -120mV while another crashes at -60mV. That gap is the lottery in action.
Cooler temperatures at the same speed are the other tell. If your chip stays under load at a lower temperature than other owners of the same model report, you got a good one.
Intel vs AMD: how the test differs
The idea is the same on both platforms. The controls have different names.
On Intel, you lower voltage with a negative Core Voltage Offset. You can set this in the BIOS or apply it from Windows with Intel XTU. A typical run steps down from -50mV to -60mV to -70mV and onward until it fails. Intel publishes the maximum safe temperature for each chip on its product pages, so look yours up before you push hard.
On AMD Ryzen, voltage is handled through Curve Optimizer, part of Precision Boost Overdrive. Instead of millivolts you set negative “counts” per core, and the chip lowers voltage dynamically as it boosts. Many owners settle around -20 to -30 counts. You can run this from the BIOS or with AMD’s Ryzen Master utility. One detail matters for the lottery: AMD’s 3D V-Cache chips, like the X3D models, run cooler by design and respond well to a light undervolt, but they limit how high their voltage can go, so treat their results on their own terms rather than comparing them to a standard chip.
Is your cooling the real bottleneck?
Before you blame a bad chip, make sure your cooler is not hiding a good one. If your CPU hits its temperature limit and throttles almost instantly under load, you are testing your cooling, not your silicon. A chip that throttles at stock cannot show you its true voltage behaviour.
Check the basics first: the cooler is seated firmly, the thermal paste is fresh and spread, the fans are spinning, and your case has enough airflow. Idle temperatures should sit comfortably below 45 degrees Celsius in a normal room. Get the cooling sorted, then test the lottery. Otherwise a great chip can look average simply because heat is piling up faster than the cooler can move it.
Old or badly applied thermal paste is a common culprit. Paste dries out over a few years, and a thin or uneven layer traps heat between the chip and the cooler. If your temperatures look high and the cooler is mounted correctly, a fresh repaste can drop load temperatures by several degrees on its own, which can be the difference between a chip that looks like a dud and one that tests as average.
Ambient room temperature matters too. The same chip will read several degrees higher on a hot afternoon than a cool night, so test in steady conditions and note the room temperature alongside your results. For the exact thermal limit of your processor, Intel’s temperature page explains that most modern chips throttle somewhere between 100 and 110 degrees Celsius, and your goal is to stay well under that during heavy work.
How far is too far?
Lowering voltage is the safe direction. Undervolting does not damage a CPU; the worst it causes is a crash or a failed boot, which you fix by raising voltage again or clearing the BIOS. It is raising voltage too high, not lowering it, that wears a chip out over time.
The risk with going too low is quieter. A setting can pass a stress test yet still crash months later, or fail only in one specific game or task that loads the chip differently. Silicon also drifts slightly as it ages, so a voltage that was fine on day one may wobble a year on. That is exactly why the safety margin in step seven matters. Find your edge, then back off a touch and live there. The few extra degrees you give up are cheap insurance against random crashes down the line.
Undervolt to learn what your chip can do, then keep a small cushion. The point is a cooler, stable PC, not the lowest number you can screenshot.
Frequently asked questions
How long does testing the thermal lottery take?
Plan for an afternoon. Stepping the voltage down and rerunning a benchmark each time takes a couple of hours, plus a 30 to 60 minute stress test at the end to confirm your final setting is truly stable.
Can undervolting damage my CPU?
No. Lowering voltage cannot harm the chip. The worst outcome is a crash or a failed boot, fixed by raising voltage or resetting the BIOS. High voltage causes long-term wear, not low voltage.
What temperature is too hot for a CPU under load?
Most modern CPUs start throttling between 100 and 110 degrees Celsius. Aim to stay below about 85 degrees under heavy load. If you hit the limit at stock, fix cooling before judging the chip.
Do I need to overclock to test the thermal lottery?
No. Undervolting is the easier and safer way to read your chip. You lower voltage rather than raise speed, which shows how little voltage your sample needs without pushing it past its limits.
Why do two identical CPUs run at different temperatures?
Tiny manufacturing variations mean each chip needs a slightly different voltage to hit the same speed. More voltage makes more heat, so one chip can run hotter than another of the exact same model.
Conclusion
Testing the thermal lottery is simple: note your stock voltage and temperature, lower the voltage in small steps, stress test each one, and find the lowest setting that stays stable. A big voltage drop with a clear temperature fall means you won; a modest drop is normal and still useful; almost no room means an average chip. Whatever you find, leave a small safety margin and enjoy a cooler, quieter PC that holds its speed for longer.
