Hacker News new | ask | show | jobs
by m_mueller 2776 days ago
I think one thing you’re missing is how good Apple’s GPU have become (and will be in 1-2 years).

I’m thinking you’re going to have a hybrid architecture for MBP. The T2 will expand to support all of Apple’s own software plus all upgraded one sold through a revamped Mac Appstore (plus hopefully your own compiled stuff when security settings are off). This can power down (all but ~2 cores) and instead power up an x86 coprocessor that supports everything else. If Intel doesn’t deliver that, AMD will. For Macbook I think you’re on the right track, just that the Air will be replaced with a 13 inch Macbook model with the same architecture.
2 comments
kllrnohj 2776 days ago
> I think one thing you’re missing is how good Apple’s GPU have become (and will be in 1-2 years).

Not compared to other discreet GPUs they haven't. They are still 5 years behind consoles. The only time mobile ever "catches up" and achieves "console-class" is near the end of a consoles ~5-8 year lifetime.

The performance is superb for mobile, and certainly good enough for integrated (it'd be perfect in something like a macbook air), but it's still getting destroyed by the relatively crappy Radeon Pro 560X in the 2018 MBP.

link
m_mueller 2776 days ago
Right, but I figure this could be scaled up with larger dies. Sure, Apple is not yet on eye level with Nvidia & AMD, but given their massive R&D budgets I don’t think that’s for long. Also, AR & VR drives demand for GPU FLOPS/Watt in a tight envelope, which has proven to scale quite well to larger chips in CPUs (see mobile ARM -> Cavium & co.).
link
kllrnohj 2775 days ago
Scaling up is one of those easier said than done things. Until Apple actually does it there's no reason to assume they could do it.

Cavium is sort of your proof that scaling is hard. The 32-core ThunderX2 @ 2.2ghz with 56 PCI-E lanes has the same TDP as the 32-core AMD Epyc with 128 PCI-E lanes. And it's slower than the competition from AMD & Intel at comparable power budgets.

link
m_mueller 2775 days ago
Sure, Cavium is not yet competitive for compute heavy workloads that make use of vectorization. For everything else, e.g. memory bandwidth bounded algos from what I gathered in benchmarks they are quite competitive. And those kind of workloads are actually quite common in HPC from my POV.
link
theluketaylor 2776 days ago
I'm a little scared to consider the hybrid approach since I think it's highly likely Apple will only let mac store code execute on the ARM side of the house. They can tout the security benefit (which is real), but they get to collect their 30%. If apple ever gets to a point where they ditch x86 macOS is no longer a general purpose OS and that's pretty much its whole appeal.

I'm not personally that interested in a hybrid mac as getting rid of the power draw of an intel cpu and all the weight and technical baggage it comes is the real draw of an arm mac for me.

What I really want is to hand apple a little over $1000 and walk away with a laptop form factor with nothing but one of their fantastic A series SoCs inside and running full macOS where I make all the decisions. The gpu being so much better than the anemic intel integrated stuff in my current air is a big part of the attraction of the A series chip running macOS.

I don't think the air branding will go away. Apple spoke so much during this last event about macbook air being people's favourite mac I can't see them giving up on such positive branding. It's far and away the best selling mac and has been for close to a decade.

link