The two-minute bottleneck test, with no made-up percentage. Measure how hard your graphics card works in the game you actually play, type in what you saw and get what to do, in order.
| GPU usage in the bad moment | What it means | What you do |
|---|---|---|
| 95-100% | Your processor is fine. The graphics card is the one maxed out. | Nothing on the CPU side. For more frames, lower settings or change the GPU. |
| 85-95% | Healthy zone. A bit of both, which is what a balanced PC looks like. | Nothing. Really, nothing. |
| 60-85% with frames below your monitor | Something other than the GPU is in charge. Usually the CPU or the memory. | The three free fixes first. If nothing changes, processor. |
| Under 60% | Clear limit outside the graphics card. | Check memory and settings before buying anything: it is rarely just the CPU. |
Low GPU usage but frames already above your monitor refresh rate? No problem: the game is capping itself.
The socket decides whether this costs a chip or a whole platform
| Your socket | What to buy | Why |
|---|---|---|
| AM4 (Ryzen 1000 to 5000) | AMD Ryzen 5 5600 | For a Ryzen 1000, 2000 or 3000 (or a 5500) already in your case: swap the chip, keep the board. It drops into almost any AM4 board with an updated BIOS. If you already have a 5600X, 5700X, 5800X or an X3D, this one is not an upgrade: your next step is AM5 (pick New board). |
| LGA 1700 (Intel 12th to 14th gen) | Intel Core i5-14400F | For an i3 or an i5 up to the 12400F: swap the chip, keep the board. Ten cores, no integrated graphics. If you already have a 13400F, a 12600K or anything above, this one is not an upgrade: your next step is a new platform (pick New board). |
| AM5 (Ryzen 7000 and 9000) | AMD Ryzen 7 7800X3D | If your games stutter in simulators, strategy games or crowded open worlds: the stacked cache is what puts X3D chips ahead where the processor is in charge. |
| AM5 (Ryzen 7000 and 9000) | AMD Ryzen 7 9800X3D | The step above, for competitive 1080p with a top-tier card. In TechSpot's 12-game test at 1080p with an RTX 5090 it led the 7800X3D by about 8%; at 1440p with a mid-range card that gap mostly disappears. |
| New board (I am replacing it) | AMD Ryzen 5 9600X | The way into AM5, the platform that is still alive: six Zen 5 cores today, and room to drop an X3D chip on the same board later. |
| New board (I am replacing it) | AMD Ryzen 7 7800X3D | If your games stutter in simulators, strategy games or crowded open worlds: the stacked cache is what puts X3D chips ahead where the processor is in charge. |
Bottleneck calculators ask for your CPU, GPU and resolution and hand back something like "23.7% bottleneck". That number has no unit and no measurement behind it: it is a subtraction between two score lists dressed up as a diagnosis.
A bottleneck is not a fixed property of your PC. It changes per game and even per scene: the same machine can be CPU-limited in a city full of characters and GPU-limited five minutes later in an empty hallway full of reflections. The only answer worth having comes from your game, on your screen.
No. They return a percentage that matches no real measurement and they have no idea what you play or at what resolution. The GPU usage test takes the same two minutes and tells you the truth.
Not at all. It only means one part is waiting on another. Every PC is limited by something; what matters is whether it costs you frames you care about.
High GPU usage and low frames: graphics card. Low GPU usage and low frames: processor or memory. That is the whole decision, and it comes from the same meter.
Once it fills up, yes: what drops is not your average frame rate but smoothness, and you get stutters. Watch memory in use in Task Manager during a match.
No. A Ryzen 5 9600X or a Core i5-14400F handle 1440p fine with a mid-range card. X3D earns its price where the processor is in charge: competitive 1080p, simulators, strategy and heavy open worlds.