GPU TFLOPS calculator
TFLOPS is a headline figure for graphics card compute power. Enter the number of shader cores and the boost clock to get the theoretical value, and add a second card to compare.
What TFLOPS measures
A teraflop is one trillion floating-point operations per second, and TFLOPS is the unit graphics card makers quote for theoretical compute. The figure comes from a simple multiplication: the number of shader cores, times the clock speed, times two, because each core can complete a fused multiply-add, which counts as two operations, on every clock. Divide by a million when the clock is in megahertz and the result is in TFLOPS.
For example, 10,752 shader cores running at 2,617 megahertz give about 56.3 TFLOPS. The result is a ceiling, not a promise. Games rarely keep every core busy on every clock, so real throughput is lower, and the gap changes from game to game.
TFLOPS for common shader counts
The arithmetic is cores times clock times two. These examples show the result at a 2,500 MHz clock.
| Shader cores | Clock | TFLOPS |
|---|---|---|
| 2,560 | 2,500 MHz | 12.8 |
| 4,608 | 2,500 MHz | 23.0 |
| 8,960 | 2,500 MHz | 44.8 |
| 10,752 | 2,500 MHz | 53.8 |
| 21,760 | 2,500 MHz | 108.8 |
Why TFLOPS is a rough guide
TFLOPS works well for comparing two cards from the same family, where the architecture is the same and a bigger number genuinely means more compute. It is a poor way to compare across generations or brands. Different designs count cores differently, use their execution units differently, and rely on caches, memory systems and specialised hardware for ray tracing and AI that TFLOPS does not capture.
That is why a card with a lower TFLOPS figure can beat one with a higher figure in games. Treat the number as an indicator of raw shader power and use benchmarks or an estimator, such as the GPU comparison tool, for expected game performance.
Reading TFLOPS alongside other specs
A graphics card is more than its compute figure. Memory size and bandwidth decide how well it handles high resolutions and large textures, cache design changes how much of that bandwidth is needed, and dedicated hardware for ray tracing and AI upscaling is not counted in TFLOPS at all. When you read a review, look for how the card performs in games you play at your resolution, then use TFLOPS only to understand the raw shader power underneath.
It also helps to remember that vendors count shader units in different ways. One brand may double its core count by widening each unit, which doubles TFLOPS without doubling game performance. That is another reason to compare within a family and treat cross-brand comparisons with care.
Using the two-card comparison
Enter the shader cores and boost clock for each card from its official specification. The tool shows each card in TFLOPS and the ratio between them. Use the clock table to see how a small clock change affects the figure; a 5 percent higher clock raises TFLOPS by exactly 5 percent, because the relationship is linear.
If you are building a PC, remember to check the rest of the system as well. A fast graphics card needs a capable processor and enough power, so try the bottleneck checker and the PSU wattage calculator before you buy.
How to use the gPU TFLOPS calculator
- Look up the number of shader cores on the official spec page.
- Enter the boost clock for Card A in megahertz.
- Add a second card to compare the two on paper.
- Use the clock table to see how a clock change moves the result.
- Use game benchmarks, not TFLOPS, for the final decision.
Common mistakes to avoid
- Comparing TFLOPS across different brands or generations.
- Using the base clock when a source quotes a boost-clock figure.
- Forgetting that two operations per core per clock is built into the formula.
Related tools and guides
Keep going with these: GPU comparison tool, GPU memory bandwidth calculator, GPU fill rate calculator, RTX 50 GPU picker quiz, graphics card comparison guide.
Frequently asked questions
How do you calculate TFLOPS?
Multiply the number of shader cores by the clock speed in megahertz and by two, then divide by one million. Ten thousand cores at 2,500 MHz gives 10,000 times 2,500 times 2 divided by 1,000,000, which is 50 TFLOPS. Use the same steps for any card.
Does a higher TFLOPS card always get more FPS?
Not always. TFLOPS only measures peak shader compute. Frame rate also depends on memory bandwidth, cache, drivers, ray tracing hardware and the game. Within one architecture it is a decent guide, across different ones it is not, so always check real game benchmarks too.
Why do TFLOPS differ between boost and base clock?
TFLOPS scales directly with clock speed, so the base clock gives a lower number than the boost clock. Marketing figures usually use the boost clock. Real games may run higher or lower than the rated boost depending on temperature and power limits.
Is TFLOPS the same as GFLOPS?
GFLOPS is a thousand times smaller. One TFLOPS equals 1,000 GFLOPS. Older cards and integrated graphics are often quoted in GFLOPS, while modern graphics cards are quoted in TFLOPS. The calculator shows both numbers side by side so you can compare older and newer cards easily.
What is FP32 versus FP16 TFLOPS?
FP32 means 32-bit floating point, the standard for game shaders, and the figure this calculator gives. Some hardware can run 16-bit operations faster, which gives a larger FP16 number. Always check which precision a quoted TFLOPS figure refers to before you compare two numbers from different sources.
Independent site. Not affiliated with or endorsed by NVIDIA, AMD or Intel. Product names are trademarks of their owners.