This article may contain affiliate links. Learn more on our Partner Program page.
Six days before the keynote, an image allegedly showing the bare chip of the upcoming flagship model is circulating. It appears to show a graphics unit with seven cores – one more than in the current flagship. However, the image itself bears no labels.
The A20 Pro is expected to power the iPhone 18 Pro and iPhone Ultra and would be Apple's first chip manufactured using a 2-nanometer process. Until now, information about its performance has been limited to percentages – most recently, figures of 18 percent faster and 30 percent more efficient compared to the A19 Pro were circulating. Now, for the first time, an image has surfaced that purports to show the chip's structure itself, providing a completely different kind of information: not a performance forecast, but a component count.
Key Facts at a Glance
- A post on the Chinese platform Weibo shows a picture that is supposed to depict the exposed A20 Pro.
- This results in seven graphics cores. The A19 Pro in the iPhone 17 Pro has six.
- The number of processing cores remains unchanged at two performance cores and four efficiency cores.
- The cache of the efficiency cores is expected to increase from 6 to 8 MB, while that of the performance cores will remain unchanged.
- Neither the Neural Engine nor the image processor are assigned to the image – precisely the units to which the AI functions depend.
What can be read from the recording
The block assignment comes from the author of the post, not from a label on the chip. Apple itself does not publish figures for cache sizes and bus widths; the only confirmed information on Apple's website is what is stated in the announcement for the iPhone 17 Pro regarding the A19 Pro: six graphics cores, Neural Accelerators in each of them, and a 16-core Neural Engine.
| Feature | A19 Pro in the iPhone 17 Pro | A20 Pro according to recording |
|---|---|---|
| Graphics cores | 6 | 7 |
| Processing cores | 2 Performance, 4 Efficiency | 2 Performance, 4 Efficiency |
| Cache performance cores | 16 MB | 16 MB |
| Intermediate storage efficiency cores | 6 MB | 8 MB |
| Storage connection | Apple has not specified a figure. | 96-bit LPDDR5X |
| Production | 3 nanometers | 2 nanometers |
The seventh graphics core would be the most noticeable change. The current Pro model has six, and Apple achieved this increase through its architecture: Neural Accelerators are integrated into each individual graphics core and accelerate local AI models. An additional core would simply multiply these processing units.
The second noteworthy feature is the memory interface. A wider connection between the chip and main memory increases the amount of data that can be transferred without requiring the memory to run at a higher clock speed. For locally executed language models, this is precisely the limiting factor: such models read large amounts of data from main memory with each individual computation step, and the speed then depends on the interface, not on the processing power.
Why the figures should be read with caution
The image is unlabeled. The creator assigned each block to a specific unit himself, after, according to his own statement, trying out several arrangements and choosing the most plausible one. This is an interpretation, not a measurement – and counting graphics cores in an image without a scale is quite prone to error.
Furthermore, there's a gap: the Neural Engine and image processor are not identified in the image. Both units carry the functions that Apple will likely highlight on September 9th, and the discovery reveals nothing about either of them.
The remaining values shown at least suggest a certain level of care: Two performance cores and four efficiency cores, as well as 16 MB of cache on the performance side, correspond exactly to what is documented for the A19 Pro. A completely fabricated chip would usually look different.
Why one more graphics core doesn't double the speed
An additional graphics core theoretically provides roughly one-sixth more graphics performance – noticeable in demanding games and video editing, but hardly noticeable in everyday use. The bottleneck in current iPhones is rarely the graphics unit.
The combination with 2-nanometer manufacturing becomes more interesting. We expect the real benefit of this generation to lie not in peak speed, but in how long high performance can be sustained – with the current Pro models, the 40 percent jump that Apple cites is explicitly for sustained performance, not peak performance. A more efficient manufacturing process plus one additional core contributes to the same effect.
If you're considering switching because of performance, this is the metric to look for on September 9th: not the number in the benchmark, but what's left after twenty minutes.
What the keynote will clarify
Apple traditionally provides core counts and benchmarks for its chips, but not cache sizes or bus widths. Therefore, only some of the information from the recording will likely be verifiable on September 9th – the number of graphics cores, for example, but not the memory interface. Until then, the overview of further expectations for the device remains available for the iPhone 18 Pro and iPhone Ultra; the keynote itself can be streamed.
Will processing power still matter to you in the next iPhone – or would you prefer longer battery life at the same speed? Tell us in the comments what you'll be looking at first on September 9th.
