See what the GPU and the Neural Engine are doing.
GPU shows the overall load with renderer, tiler and GPU memory beside it, then the top apps by GPU time. Neural Engine load has its own panel underneath, because local models have made it worth watching.
Skip the live demo
macometerApple M4 Pro, 48 GB
GPUFirefox4%
6%Cores20
Inside the GPU
Renderer5%
Tiler3%
GPU memory2.7GB of 48 GB
Top apps by GPU
Neural Engine
6%Power–Memory traffic159MB/sCores16
Firefox is using the most GPU time, 4%.
| 4% | 1.5W | 0.6% | 6.4GB | 7 | |
| 1% | 0.7W | 0.9% | 2.4GB | 4 | |
| 1% | 1.3W | 2.1% | 4.8GB | 5 | |
| <1% | 0.8W | 1.2% | 5.1GB | 14 | |
| <1% | 0.3W | 0.3% | 1.2GB | 4 | |
| <1% | 0.1W | 0.2% | 305MB | 1 | |
| <1% | 0.1W | 0.2% | 442MB | 2 | |
| <1% | 0.1W | 0.2% | 949MB | 3 | |
| <1% | 0.1W | 0.2% | 539MB | 2 | |
| <1% | <0.1W | 0.1% | 252MB | 1 | |
| <1% | <0.1W | 0.1% | 197MB | 3 | |
| 0% | 0.3W | 0.5% | 4.1GB | 4 |
Memory
31.6GB
Available16.4GBWired4.9GBSwap659MBCompressed2.4GB
Top apps by memory
Firefox6.4GB
macOS5.1GB
Xcode4.8GB
Docker4.1GB
GPU
6%
Memory2.7GBRenderer5%Tiler3%
Top apps by GPU
Firefox4%
Figma1%
Xcode1%
macOS<1%
Storage
1.1MB/s
Read467KB/sWrite621KB/sFree412GB
Top apps by disk
Xcode604KB/s
Docker157KB/s
macOS61KB/s
Firefox62KB/s
Simulated data, running in your browser.
Neural Engine load comes from the chip.
It reads hardware counters that differ by chip, and where one is missing it reads a different one. Checked against real Vision inference on an M1 Pro, the load rose and fell with the work.
Top apps by GPU.
The same top apps list as elsewhere, and it moves with the timeline. A spike in GPU use has a name.
Nothing is guessed.
If this Mac does not report a counter, the panel says "Not available on this Mac". It is never a zero.
$7.99 for the first 25, then $14.99, once. No subscription.