I'm so glad Apple is giving people such cutting edge graphics hardware in phones now; games like Royal Match and Monopoly GO are going to look so good. I might even fire up Candy Crush again, and see how it performs on that super-fast 3nm chip.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
iPhone 16 Pro's A18 Pro Chip Rumored to Boost AI, Retain 6-Core GPU
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple’s processors are quite fast for LLMs.
I’ve been using LM Studio. It’s faster at running various LLMs on an M1 Pro than on a AMD 5900x with a Nvidia 3070.
It will be interesting to see how powerful these new processors are.
I’ve been using LM Studio. It’s faster at running various LLMs on an M1 Pro than on a AMD 5900x with a Nvidia 3070.
It will be interesting to see how powerful these new processors are.
With emulation there’s an actual use case for that GPU grunt on Android
Almost picked up a 13T Pro for that Mediatek chip. It’s crazy how Mediatek used to mean crummy and now is up there with the top chips
Not surprising considering they just stick with everything designed by ARM, and they just throw in all the best things available to them
It's the shared RAM that allows this. On the PC side, the GPU needs its own expensive dedicated ram alongside the system RAM, hence part of the reason why people are using Macs in the AI space. The Mac Studio with 128GB of shared memory costs less than its PC equivalent for machine learning tasks.
So it looks like the Apple Pencil will work in Apple Vision Pro, a VR headset with nothing to write on, before it works on the iPhone, the only thing I want to write on.
Dammit Apple. Why you gotta be like that?
Dammit Apple. Why you gotta be like that?
Eh … what about Photos and the Camera app? I’d like for Apple to finally catch up with Google and Samsung at this point for basic things like removing unwanted objects from pictures without 3rd party apps that do a more or LESS good job with this.
Definitely expecting to see some useful AI features this year with the new iPhones.
From the article you're commenting on:
"the larger die size would be related to the Neural Engine, which could power on-device generative AI features."
This might be the most uninteresting iOS update in a long time. I’m just not interested in AI.
Maybe Apple will impress me and prove me wrong
Maybe Apple will impress me and prove me wrong
It is partially a shared RAM effect. Some of it is how the models are programmed to run. My MBP has 32 GB of RAM. My Windows computer has 128 GB of RAM but the 3070 only has 8GB. Even with this GPU limitation, things like Stable Diffusion image generation (heavily CUDA-dependent) are much faster than on my Mac.
For my computers, macOS beat Windows for text generation but Windows/Linux beats macOS for image generation.
Apple has done a lot of work with LLMs (e.g., https://github.com/apple/ml-aim; https://machinelearning.apple.com/research/leveraging-large-language-models) so it will be interesting to see what can be run on device with the "AI" parts of Apple Silicon.
Last edited:
I use both Alexa and Siri. Siri is consistently much better than Alexa. They used to be fairly similar but Siri consistently outpaces Alexa’s performance in my household. We’ve mostly stopped using Alexa devices now because Alexa isn’t great.
Last edited:
I wonder when we are going to be able to use the iPhone as a computer at home. Like with usb c hook it up to a monitor and from there be able to control the OS with keyboard and mouse
Apple will be the only company who can truly monetize AI. Nvidia is just selling chips to msft meta on AI hype hope. Msft meta still need to figure out a way to substantially monetize AI or down goes Nvidia. How ironic “Apple is behind in AI!” that Apple with AI App Store and AI Megacycle will monetize AI the most.
I'm gonna go with never.
Apple has devices that already work that way, and they want to keep selling them.
Number of GPU cores is only part of the story. They could dramatically increase the performance of each core without the increasing the number of cores for example. For example, just last week Apple has published a patent that describes a redesigned GPU scheduler.