Many laptops run above their sustainable level for the first few seconds of heavy work, then step down to a level they hold.
- On SystemCheck's validation laptop the step came 4.5 to 5.5 seconds into the CPU stage, and the settled level sat about 25% below the boost level, across ten runs.
- Since measurement revision 2026.09.15 the report shows that boost step on its own and does not score it. Decay is what the machine loses after it settles.
- A short benchmark that ends inside the boost reports the boost. A second run, started warm, may never reach it, which is one reason a second score can read lower.
The short answer
The speed a machine shows in its first seconds of heavy work is often not the speed it keeps. Many laptops are allowed to run above their sustained level briefly, then come down to a level they can hold for as long as the work lasts. The second figure is the one that describes everyday heavy work, so SystemCheck reports the two apart.
What ten runs on one laptop showed
On 5 October 2026 SystemCheck ran ten Standard runs back to back on its validation laptop, an Intel Core i5-1245U on mains power. Read with the curve classifier of revision 2026.09.15, eight of the ten CPU stages had the same shape: a short boost, one step down, then a level that held. The classifier in use when the runs were made split the same ten runs five ways.
Across the ten runs the boost ended between 4.5 and 5.5 seconds into the stage, and the step from the boost level to the settled level averaged 24.6% of the boost level. After the step the curve was close to flat. That is one laptop, and a machine with a longer boost, a desktop that does not boost, or a laptop on battery can look different; those have not been validated yet.
Why the boost is shown and not scored
Before revision 2026.09.15, decay compared the start of the stage with its end, so the boost step was counted as lost performance. On these ten runs that old figure ranged from 0 to 29% and the curve label changed from run to run, because the step sat right where the comparison began.
Now the boost step is measured as the best one-second level before 10 seconds against the level from 10 to 20 seconds, and it is shown as information. Settled decay compares that 10 to 20 second level with the end of the stage, and that is the figure that counts. On the same ten runs settled decay passed the noise gate in only one run, which had a real late fall.
Why a second run can score lower
If the first run starts cold and the second starts warm, the second may get a shorter boost or none. A benchmark that only measures a few seconds then reports a lower figure, even though the machine's sustained level has not changed. Comparing settled levels, not first-seconds peaks, avoids that trap. The report also names other conditions that move a figure between runs, such as battery power or a hidden tab.
Terms used here
Questions
- Is a boost step a fault?
- No. Running above the sustained level for a short time is ordinary behaviour for many laptops. SystemCheck shows the step so you can see it, and leaves it out of the score.
- Can SystemCheck tell why the step happened?
- Not from a browser. A web page sees no clock, power or temperature reading, so the report describes the shape and leaves the cause to a sensor tool.