How we measure performance claims

Why this page exists

Performance figures in the PC optimization category are usually a single impressive number with nothing underneath it: no named hardware, no scene, no method, and no way for you to check. We removed claims like that from this site, and pages here that discuss performance link to this one so that the reasoning behind every remaining claim is written down in one place.

A methodology page is only worth something if it describes what actually happens, including the parts that fall short. So this page says plainly what the current published captures are and are not.

The standard a performance figure has to meet

For a figure to count as a measured result on this site, it needs all of the following:

• Named hardware. Processor, graphics card, memory and resolution stated next to the figure, never somewhere else.
• A repeatable scene. A built-in benchmark, replay or fixed route, so the same work is measured each time.
• Frame times, not a counter. Recorded with a capture tool so uneven frame delivery is visible, rather than read off an average.
• Several runs of each state. Enough to show the spread between identical runs, so a difference smaller than that spread is not reported as a gain.
• One change between states. Everything except the thing being tested held constant, including drivers and background software.
• No generalisation. The figure describes that machine in that scene. It is never restated as something you should expect.

What the site publishes today, and how far it meets that standard

The only frame-rate figures on this site are on the homepage, in the section headed "Same PC. Measured.": three before-and-after captures from three games, each showing the frame counter on screen. The figures beside them are read directly off those captures, and a method line names the test PC as an AMD Ryzen 7 5800X3D with an NVIDIA GeForce RTX 2070 and 32 GB of memory at 1920×1080.

Measured against the standard above, those captures meet the first rule and fall short of others. Each is a screenshot of an on-screen counter in each state, not a set of frame-time recordings across several runs. They show what the counter read on one machine at the moment of capture. That is why they carry their method line, are not restated as expected results anywhere else on the site, and are not quoted on pages like this one.

Why there are no blanket FPS percentages here

A percentage invites you to apply it to your own PC, and it almost never transfers. The effect of any Windows change depends on where your system is limited: a PC held back by its graphics card has little for a Windows setting to release, while one held back by background software or power policy may have more. It also depends on what was running before, which game, which driver, and which patch of that game.

Our own research page, registry tweaks that do nothing, deliberately contains no benchmarks for the same reason: its findings rest on Microsoft's documentation, which holds on any hardware, rather than on a measurement that would only describe ours. If you want to know what a change does for you, measuring it yourself is the only answer that applies, and the optimizer FPS guide sets out a method for doing that at home.

What this means for our own product

JINSHI Tweaks does not advertise an expected frame-rate gain, because we could not support one under the rules above. What we publish instead is exactly what it changes: every change in PC Tweaks is listed with its registry path on the complete tweak list, so you can judge each one yourself and measure the ones you care about. If that approach suits you, the JINSHI Tweaks homepage shows what the four apps include.

For the sceptical case against optimizers in general, see whether PC optimizers really boost FPS, and for choosing a tool on evidence rather than claims, the tweaking app comparison.