Hacker News new | past | comments | ask | show | jobs | submit login

Why would you expect that Apple ARM is going to get close to i9 performance of 16" MBP? Nothing from benchmarks shows that.

All the Apple ARM benchmarks I've seen is that they're competitive with MacBook Air class offerings and not the high performance CPUs.




Best Benchmarks

MBPs with i9s:

Single Core: 1,119

8 Core: 6,900

A12Z (iPad 12.9 inch)

Single Core: 1,118

8 Core: 4,626

A13 Bionic (iPhone 11)

Single Core: 1,327

6 Core: 3,384

Neither of these ARM CPUs will be used in Apple Silicon. It will be an A14 class chip. It will be significantly faster due to

1) New 5 nm process vs. the older 7 nm process.

2) Greater thermal headroom ie they will be able to use more power in laptops and desktops than in tablets/phones.

3) More cores

It is expected that Apple Silicon will have at least 12 to 20 cores, and more for high end desktop versions.

4) GPU integration

The current A13 has an Apple integrated GPU that’s reputedly much faster than the Intel integrated GPUs. The 5 nm process will provide many more transistors for even faster GPU performance.

My expectation is Apple will have a family of Apple Silicon SOCs next year that start around 1,400 single core, and with multi-core starting at 7,000 and going over 20,000 for 30 or 40 core desktop versions.

They will probably include T2 functionality and more, further reducing power draw.

Also remember that i9s are $200-$300 each. A13s are around $100 each.

MacBooks will get the benefits of longer battery life, faster CPUs & GPUs while also being as much as $200 cheaper.


I thought the geekbench results for the A13 are already comparing favorably to a i9 and that is running on a phone. Imagine a version which has 10x the thermal headroom than the iPhone. Also, the ARM MB Pros are likely to run a variation of the A14 on the 5nm TSMC process.

Beyond the raw CPU power, it has been quite overlooked how much performance Apple can get by putting additional compute units onto the chip, like their neural net accellerators. As they own the whole hard- and softwarestack, they can much more easily make the best use out of those additions.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: