A 13-Year-Old Intel CPU Hit 6.3 GHz—Don’t Try This at Home

An Intel Core i5-4670K from 2013 has just reached 6.3 GHz and set a new Cinebench 2003 record for that processor. The result is a wonderful piece of competitive PC engineering. It is also a terrible template for tuning the old gaming PC under your desk.

Finnish overclocker Luumi recorded a score of 10.8 cb after running the four-core Haswell chip at 6,301 MHz, according to a fresh BGR report based on the HWBOT submission. The winning margin was only 0.04 points. The test system used 4GB of DDR3 memory in two 2GB modules and, crucially, liquid nitrogen rather than an ordinary air cooler or all-in-one liquid loop.

6.3 GHz is far outside the chip's original job

Intel's specifications put the Core i5-4670K at a 3.4 GHz base frequency and a maximum 3.8 GHz Turbo Boost frequency. Its 6.3 GHz benchmark run was therefore about 66% above the official maximum turbo speed and roughly 85% above base clock. The processor launched in the second quarter of 2013 with four cores, four threads, 6MB of cache, an 84-watt thermal design power, and support for DDR3-1600 memory.

The “K” suffix matters because it identifies an unlocked processor multiplier. That makes frequency tuning possible on a compatible motherboard, but it does not mean every sample can reach the same clock. Manufacturing variation, motherboard power delivery, firmware, memory, cooling, voltage, and even the individual CPU's history all influence the result.

This is also a record within a particular hardware and benchmark category, not evidence that an old i5 can beat a current desktop processor. Cinebench 2003 is a historical test, and a short benchmark pass proves only that the system completed that workload. It does not demonstrate day-long gaming stability, modern application performance, energy efficiency, or years of remaining service life.

Liquid nitrogen turns the run into an exhibition

At approximately minus 196 degrees Celsius, liquid nitrogen can pull heat away quickly enough to make extreme voltage and frequency experiments possible for a brief session. Competitive overclockers mount a metal cooling pot directly over the processor, insulate the surrounding motherboard, and repeatedly manage temperature and condensation while the test runs.

That setup is fundamentally different from a daily PC. Once the nitrogen evaporates and the hardware warms up, moisture can condense around the socket and electrical components. The fluid can cause severe cold burns, while evaporating nitrogen can displace oxygen in a poorly ventilated room. Extreme cooling belongs in the hands of experienced operators with appropriate protective equipment, ventilation, insulation, and recovery procedures.

Even without liquid nitrogen, additional frequency usually requires additional core voltage. Intel warns that changing clock frequency or voltage may reduce stability, security, performance, and component life, and may affect warranty coverage. Higher voltage also increases heat and power consumption. An unstable overclock can look successful in a screenshot yet produce crashes, calculation errors, corrupted work, or intermittent game failures later.

A safer way to tune a daily PC

If you want to experiment with an unlocked processor, begin with a reason. A measurable CPU bottleneck in a game or workload is a better reason than chasing a round number. A 100 MHz increase that cannot improve the task you care about is simply extra heat.

Before changing anything, update the motherboard firmware if the vendor recommends it, record the current BIOS settings, and measure stock performance, temperatures, clock behavior, and power use. Confirm that the cooler is mounted correctly and that case airflow is clean. Old thermal paste, clogged heatsinks, tired fans, and aging voltage-regulator components deserve attention before an old CPU receives more voltage.

Raise the multiplier in small steps and change one major variable at a time. Test after every adjustment. Intel's current tuning guidance describes a five-minute test as a quick screen, around 30 minutes as useful for checking cooling, and three to five hours or longer as a stronger validation for a system intended to run that setting every day. No single stress test covers every workload, so follow synthetic testing with the games and applications you actually use.

Watch for thermal throttling, corrected hardware-error reports, application exits, visual freezes, reboots, blue screens, or performance that falls after a few minutes. Any of those can indicate that the setting is not suitable. Return to the last known-stable configuration rather than adding voltage automatically. The correct limit depends on the exact processor and board, so do not copy a stranger's voltage value from a leaderboard.

The record is valuable for a different reason

Luumi's result shows how much technique remains in a platform long after it leaves store shelves. Extracting another 0.04 points from a record that stood for roughly 12 years required careful preparation, component selection, cooling control, and repeatable execution. That is the achievement—not a recommendation that every Haswell owner should pursue 6.3 GHz.

For a daily PC, quiet cooling, predictable stability, and protected data are worth more than a headline clock. Celebrate the record as motorsport for computer hardware, then tune your own system like something you expect to use tomorrow.