Conceptual illustration: Mais FPS, created with AI.
A new graphics card should not come with an automatic motherboard-and-memory tax. If a Ryzen 5 5600X3D or Ryzen 7 5700X3D already runs your games well, installing a GeForce RTX 5070 Ti or RTX 5080 is a reason to measure the result, not immediately retire AM4. The expensive mistake is buying the platform upgrade before discovering which part of your experience it would improve.
Our position is simple: keep a capable processor until the games you actually play give you a good reason to replace it. That is not the same as claiming an older X3D matches every AM5 processor. The game, scene, settings and desired frame rate get a vote.
One powerful GPU, two very different verdicts
In TechSpot’s August 2025 CPU/GPU scaling test, a Ryzen 7 5800X3D, Ryzen 5 7600X and Ryzen 7 9800X3D all delivered 94 FPS with an RTX 5080 in Marvel Rivals at 1440p Ultra. The graphics workload set the limit. Paying for the faster CPU did not buy a faster average there.
In the same review, Space Marine 2 at 1440p Ultra with the same RTX 5080 left the 5800X3D 28% behind the 9800X3D. At 4K Ultra, that gap narrowed to 5%. Those results make a better purchasing guide than either “AM4 is finished” or “the CPU never matters at high resolution.”
These are TechSpot’s measurements, not MaisFPS tests. They used a 5800X3D, not a 5600X3D or 5700X3D; they also do not establish an exact RTX 5070 Ti result. Treat them as evidence that an AM4 X3D can remain useful beside a strong GPU, with meaningful game-dependent limits. Do not paste those percentages onto a different processor.
The six-core 5600X3D and eight-core 5700X3D nevertheless deserve more respect than their platform’s age suggests. In TechSpot’s February 2024 comparison, they were closely matched across its 12-game average. That supports investigating what you own before buying; it does not promise the same relationship in every newer game.
Make your own PC answer the useful question
Choose a repeatable section where performance actually bothers you. A crowded city, a saved battle or a replay is more revealing than a quiet training map if the busy match is your problem. Record the resolution, render scale, preset and upscaling mode.
For diagnosis, temporarily disable frame generation so it does not obscure the underlying rendering rate. Check the frame cap and V-sync. A GPU meeting a deliberate limit is allowed to have a quiet afternoon.
Run the same route several times after normal loading settles. Keep an unusually rough first run separate. Record frame times, average performance, your tool’s low-percentile result, GPU activity, and sustained CPU/GPU clocks and temperatures. PresentMon-based tools can help; use the same tool and settings for every comparison. Different “1% low” calculations are not automatically interchangeable.
Now lower the internal rendering resolution substantially, leaving simulation, crowd density and the route unchanged. A repeatable improvement points to an important graphics-side constraint. Little improvement sends you looking at CPU work, caps and engine limits. It does not, by itself, convict the processor.
Return to your normal resolution and separately reduce a setting the developer identifies as CPU-heavy. If that helps where the resolution cut did not, the diagnosis gets stronger. Changing every slider together produces a result that is wonderfully easy to misread.
Low overall CPU usage is not an alibi
A game may be waiting on one demanding thread, a sequence of instructions, while other cores have spare capacity. A modest total CPU percentage therefore cannot rule out a CPU limit. Equally, low GPU activity can reflect synchronization, loading or a cap. Intel’s analysis guidance stresses that the limiting component can change between scenes.
Investigate hitching separately. A first traversal that stutters and then settles suggests a different problem from a battle that repeatedly slows down. Check known game issues, memory pressure, background work and sustained clocks before assuming a shopping cart will cure it.
Upgrade for the target, not the screenshot
Someone chasing consistent 240 or 360 FPS at competitive settings can justify a CPU upgrade that does little for someone enjoying 4K Ultra around a lower target. Upscaling complicates the shorthand: a 4K output may use a much lower internal resolution and expose more CPU limitation.
If your 5600X3D or 5700X3D meets your actual target beside a new GPU, keeping it is a sensible decision. If repeatable CPU-heavy scenes miss that target, compare independent tests and the entire CPU, board and memory bill. Unused GPU capacity is evidence to interpret, not an invoice you are obliged to pay.
