Your browser is not supported or outdated so some features of the site might not be available.
52
Laptops store-bought and tested, supported by you via insider access, and when you purchase through links on our site, we may earn an affiliate commission.
Driven by data, run by a passionate team of engineers, testers, technical writers, developers, and more. We are hiring!

Our Laptop Performance Tests
Performance Over Time

Updated
What it is: How the performance of the laptop decreases as it gets hot from working hard over a long time. Performed in a temperature controlled environment (22℃).
When it matters: When running a heavy workload over a long time, such as when gaming or running large 3D renders.
Score distribution

Laptop manufacturers often talk about their products' CPU and GPU peak performance but generally don't disclose whether this performance can be maintained when the laptop has been under a heavy load for an extended period. This can be frustrating as you might pay for a laptop with high-end parts, only to find that it's only marginally better than a lower-tier configuration, leaving you feeling like you aren't getting your money's worth. As such, we have included a performance over time test to see whether a laptop can maintain reasonable temperatures and performance over an hour so that you can make a more informed buying decision. This article details how we conduct the performance over time test.

Test results

Why It Matters

Heat is often an issue for laptops, so to prevent any damage to the internal components, manufacturers put in safeguards to limit a CPU or GPU's performance. It's known as thermal throttling. For most people doing simple tasks like checking emails, web browsing, or watching online videos, performance loss isn't an issue because these tasks don't stress the CPU or GPU enough to cause thermal throttling. However, any performance loss is immediately noticeable for those with heavy workloads, like gamers and content creators. Our performance over time test shows you how hot the CPU and GPU get under load and whether there's any performance loss over an hour.

Our tests

Setup

Our performance over time test doesn't require any special equipment as we can obtain the results through software. We perform the test in a temperature-controlled room set to 22°C (+/- 0.5°C tolerance), with the laptop plugged in using its included charger.

CPU Temperature (Cinebench)

What it is: How hot the CPU gets when it's working really hard for a long period. Measured as the average temperature of the CPU after running the Cinebench R23 Multi-Core workload for an hour. Tested in a temperature controlled environment (22 °C).
When it matters: When gaming, or when running a CPU heavy workload like image processing or video rendering without hardware acceleration.
Good value: < 80 °C
Noticeable difference: 5 °C

The first test is to determine the CPU's temperature under load. It's an important metric because, like most electronics, extreme heat can damage the components or shorten their lifespan. It can also cause a hotspot that might make using the laptop uncomfortable. Although we consider anything below 80°C as good, it isn't uncommon to see CPUs get much hotter, especially with H-series processors in gaming or 'workstation' laptops. The consensus is that the temperature shouldn't exceed 100°C, but may vary depending on the individual CPU's specifications.

We perform this test by running Cinebench R23 on a loop for an hour. We collect the temperature data using 'HWinfo64' on Windows machines and 'Intel Power Gadgets' on Intel-based MacBooks. Unfortunately, we don't have a way to log the temperature on M1 Macs at this time, and we can't perform this test on Chromebooks because there isn't a Chrome OS version of Cinebench R23. The number we post in the review is the average temperature recorded during the last minute of the hour-long loop test.

CPU Performance Loss (Cinebench)

What it is: How the CPU performance decreases over time as the laptop gets hot, from running the intensive Cinebench R23 benchmark over a long period. Measured as the difference between the first score and the average of every score afterwards.
When it matters: When running a CPU-intensive workload over a long time, such as 3D rendering or video encoding.
Good value: < 10%
Noticeable difference: 5%
Score distribution

We obtain the CPU performance loss result by recording the benchmark score of each Cinebench run during the hour-long loop. It's common to see some performance loss on fanless devices because they're thermally constrained. For laptops with active cooling, it depends entirely on the cooling system (number of fans, vapor chamber, etc.) and the CPU's power/thermal configuration. It's worth mentioning that benchmark scores can differ from one run to another, even on the same machine. That said, performance variation of less than 5% isn't noticeable to most people in real-world usage. Also, the method we use to detect performance loss has limitations, as the CPU might have already throttled during the first run of the benchmark, and all subsequent results are already in the 'throttled' state, making it seem like there isn't any performance loss.

GPU Temperature (Heaven)

What it is: How hot the GPU (video card) gets when it's working really hard for a long period. Measured as the average temperature of the GPU after running the Unigine Heaven benchmark for an hour. Tested in a temperature controlled environment (22 °C).
When it matters: When gaming, or when running a GPU-heavy workload like video rendering.
Good value: < 80 °C
Noticeable difference: 5 °C

For the GPU temperature test, we run the UNIGINE Heaven benchmark for an hour with the following settings:

  • API: OpenGL
  • Quality: Ultra
  • Tessellation: Moderate
  • Anti-Aliasing: 2x
  • Fullscreen: Enabled
  • Resolution: 1920 x 1080 (If the laptop's screen isn't 16:9, we connect it to an external monitor with the image showing on the monitor only)

The Heaven benchmark logs the GPU's temperature. The result in the review is the average temperature recorded during the last minute of the benchmark loop. Like the CPU, GPU temperature can vary greatly depending on the cooling solution and power configuration, which means you might see different results for the same GPU in two different laptops. We can't run this test on Chromebooks because there isn't a Chrome OS version of Unigine Heaven at this time.

GPU Performance Loss (Heaven)

What it is: How the GPU/gaming performance decreases over time as the laptop gets hot, from running the intensive Unigen Heaven gaming benchmark over a long period. Measured as the difference between the first score and the average of every score afterwards.
When it matters: When playing video games for long periods, or doing other intensive workloads that use the video card, such as hardware-accelerated video encoding.
Good value: < 10%
Noticeable difference: 5%
Score distribution

We record GPU performance loss through the multiple runs of UNIGINE Heaven. Each run takes about 4.5 minutes, which means we can run the benchmark 13-14 times during the hour-long loop and see whether there's any performance loss after each run. Like CPUs, GPU performance depends on the cooling system and the GPU's power/thermal configuration. Again, benchmark scores vary from one run to another, and any performance loss of less than 5% isn't likely to be noticeable to most people. Similar to our CPU performance loss test, there are limitations to our methodology as the GPU might have already entered its throttled state during the first run.

Additional Information

If you're concerned about performance loss on your laptop, there are a few things that you can do to prevent or minimize it. Here are some tips on getting the best performance out of your laptop.

Best Practices

  1. Avoid blocking the exhaust vents. Although this may seem obvious, we sometimes forget when we're busy or caught up in a gaming session. Know where the exhaust vents are on your laptop. If they're on the bottom, make sure not to place the laptop on a carpet, bed, or any fabric that may impede airflow. If you must, put something hard underneath to support it, like a book or laptop cooling mat. You don't need to worry about this as much on a fanless device, but if the hottest spot under load is on the bottom of the laptop, it may still be helpful to leave a little room for some airflow.
  2. If possible, avoid very hot environments, like direct sunlight or poorly ventilated areas. Warm surroundings raise the base operating temperature and make it harder for the laptop to cool off.
  3. Avoid dusty environments. This applies to laptops with active cooling only because dust can get into the device, reducing airflow and potentially clogging the fans.

Maintenance & Optimization

  1. Clean the fans and internals. If you have a laptop with airflow vents, chances are that the fans might clog with dust after a year or two. You might even hear the fan noise increase over time. To clean the fans, you only have to open the laptop and use a can of pressurized air to get rid of the dust but make sure to hold the fans while you're doing it because spinning the fans in the wrong direction can damage them.
  2. If you're comfortable going further, you can reapply the CPU/GPU thermal paste, as the old one might have dried up over time, becoming less effective.
  3. Undervolting. Reducing the CPU's voltage can be beneficial in some cases as it lowers the temperature with minimal loss to performance. It might even boost performance on some laptops because the CPU can run at higher clock speeds for longer. You don't need to access the internals for undervolting as you can do everything through software.

 

Conclusion

Generally speaking, performance loss is fairly common on fanless devices because they're thermally constrained. If you plan on performing demanding tasks for extended periods, it's best to check this test in our review to see whether the laptop throttles under load, especially for laptops with power-hungry parts. Do note, though, that our performance over time test is only relevant for the configuration we tested, and we can only make educated guesses about how the other models and variants will perform.

Discussions