Apple A14 Benchmarks and Apple Silicon Macs | Hacker News Reader

Apple A14 Benchmarks and Apple Silicon Macs

twitter.com

17 pointstosh7 comments

Raw performance is great, and it’s easy to salivate over that. (Or be concerned that you're getting a 5% bump at the cost of no x86 virtualization.)

But I think Apple is doing this for a different performance metric: battery life in the MacBook Air.

Apple doesn't really break out what model line sells what, but it's safe to assume the Air is the big seller, followed by MacBooks Pro, then the iMac, then the mini, and then a rounding error of Mac Pro. For consumer machines, it’s not that compelling to say “it compiles 20% faster”. It is really compelling to say “Apple introduces MacBook Air, powered by Apple Silicon, with 12 hours of battery life, and it never gets hot.”

Pro users need floating point, and this is roughly half the score of an older iMac Pro.

It's not a bad result, but it's still some way short of the kind of performance you need for professional 3D rendering or audio workstation applications.

As someone who lives in a hot climate the only metric I care about is my Macbook not being a hand heater. I miss the old Macbook Air and how nice the cool aluminium surface was under my palms. Every Macbook i've had since then has been uncomfortably sweaty and if this Apple Silicone can reduce the palm rest a few degrees then i'm sold. Of course, if they had put proper cooling in the new Macbook Air it would have done it as well. But gotta push people to those more expensive models every way you can right?

As a developer, the compelling thing for me is not power but battery life. So I agree with your sentiment.

That is, as long as the price is right.

> I personally expect even the thinnest & lightest devices to beat out almost all of the 2020

What's this? And why is this on the front page?

> The big wildcard is graphics; Apple has to get from a 9,000-odd Metal compute score from the A12X to the 38,789 of a Radeon Pro 5500 XT in the top-end iMac today — that'll mean they likely need 4–5X the GPU cores, which seems very achievable in a Mac-sized thermal profile

Is there something that would prevent them from associating a Radeon gfx with an ARM CPU as an interim solution while they amp up their own graphics?

Is there even a PCI Express on ARM architecture?

Of course PCIe is a standard on Arm. (as are UEFI and ACPI)

On Linux, Nvidia officially supports their drivers on Arm for PCIe GPUs since... June 26, 2013 for 32-bit arm. (see https://www.nvidia.com/en-us/drivers/unix/linux-arm-display-...). For Arm 64-bit and PCIe GPUs, it's far more recent ( https://www.nvidia.com/en-us/drivers/unix/linux-aarch64-arch... )

nouveau and amdgpu run fine too.

Reply on news.ycombinator.com