Apple introduces M6 and M5 Ultra • Hacker News Board

Re: Apple introduces M6 and M5 Ultra

#21

Post by bel8 » Tue, Aug 25, 2026, 1:18 PM UTC

LeBit wrote: post #7

rvz wrote: post #2

> M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB, and delivers a staggering 1.2TB/s of unified memory bandwidth that is 50 percent higher than M3 Ultra. Apple never needed to participate in the AI race to zero. Because they were already at the finish line years ago building their own chips that can run large >100B parameter AI models locally.

1.2TB/s is 2/3 the speed of an nVidia 5090. But you get a generic computer and much more RAM. And you lose a couple of organs.

The real downside for me is not having Linux support.

It would take Apple one or two engineers to make Linux life much easier on macs. But Linux is outside their walled garden so it's ignored.

Re: Apple introduces M6 and M5 Ultra

#22

Post by jjice » Tue, Aug 25, 2026, 1:18 PM UTC

rvz wrote: post #2

> M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB, and delivers a staggering 1.2TB/s of unified memory bandwidth that is 50 percent higher than M3 Ultra. Apple never needed to participate in the AI race to zero. Because they were already at the finish line years ago building their own chips that can run large >100B parameter AI models locally.

I was so blown away at all the discourse surrounding "Apple fumbling on models". They should never have been in the model game to begin with. Apple crushes hardware over the last decade and that's a huge advantage today. In the end, massive models have proven to be very strong, but small models have proven to be good enough (especially with the recent Qwen 2.8 27B drop) and that's where I imagine the future will lie for consumers.

Re: Apple introduces M6 and M5 Ultra

#23

Post by Lunar5227 » Tue, Aug 25, 2026, 1:18 PM UTC

teekert wrote: post #4

rvz wrote: post #2

> M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB, and delivers a staggering 1.2TB/s of unified memory bandwidth that is 50 percent higher than M3 Ultra. Apple never needed to participate in the AI race to zero. Because they were already at the finish line years ago building their own chips that can run large >100B parameter AI models locally.

Is there anything comparable that runs Linux, doesn't necessarily look as good, but is perhaps (a lot) cheaper/fixable? Or is this really pretty optimal? I mean this is not nvidia based right? It's all custom? So we can use it under Asahi perhaps? I want to get something for my company to run local models, wondering what would be a good option.

Strix platform maybe?

Re: Apple introduces M6 and M5 Ultra

#24

Post by tester756 » Tue, Aug 25, 2026, 1:19 PM UTC

When will Apple's Mx CPUs use Intel's 18A/18A-P/14A node(s)?

Re: Apple introduces M6 and M5 Ultra

#25

Post by SilverSlash » Tue, Aug 25, 2026, 1:19 PM UTC

The 512GB Ultra is amazing, sure. But who is it for exactly? VC funded big spender founders? In that case why would they need local AI? The ultra rich enthusiast? But there can't be too many of those. So who actually buys these?

Re: Apple introduces M6 and M5 Ultra

#26

Post by Matl » Tue, Aug 25, 2026, 1:20 PM UTC

delduca wrote: post #5

I know they are a phone company, but I think they should focus on local models software, not only hardware.

Software wise there's plenty to choose from already. Ollama/llama.cpp, LM Studio, Lemonade, vllm etc. Anything Apple would bring to the table?

Re: Apple introduces M6 and M5 Ultra

#27

Post by kksweet » Tue, Aug 25, 2026, 1:21 PM UTC

delduca wrote: post #5

I know they are a phone company, but I think they should focus on local models software, not only hardware.

They aren't a phone company, and haven't been ever. They're a hardware company first

Re: Apple introduces M6 and M5 Ultra

#28

Post by mleo » Tue, Aug 25, 2026, 1:21 PM UTC

delduca wrote: post #5

I know they are a phone company, but I think they should focus on local models software, not only hardware.

Apple has the mlx framework. Most/all major software for running models locally support it. Apple also has RDMA for interconnecting multiple machines across Thunderbolt connections.

Re: Apple introduces M6 and M5 Ultra

#29

Post by givinguflac » Tue, Aug 25, 2026, 1:21 PM UTC

delduca wrote: post #5

I know they are a phone company, but I think they should focus on local models software, not only hardware.

Uhhh, they are? They’re a hardware co for sure, but to say they aren’t focusing on on-device models is absurd on its face. They’ve spent over 2 years on Siri AI which is (mostly) local.

Re: Apple introduces M6 and M5 Ultra

#30

Post by terminalcommand » Tue, Aug 25, 2026, 1:22 PM UTC

teekert wrote: post #4

rvz wrote: post #2

> M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB, and delivers a staggering 1.2TB/s of unified memory bandwidth that is 50 percent higher than M3 Ultra. Apple never needed to participate in the AI race to zero. Because they were already at the finish line years ago building their own chips that can run large >100B parameter AI models locally.

Is there anything comparable that runs Linux, doesn't necessarily look as good, but is perhaps (a lot) cheaper/fixable? Or is this really pretty optimal? I mean this is not nvidia based right? It's all custom? So we can use it under Asahi perhaps? I want to get something for my company to run local models, wondering what would be a good option.

AFAIK, apple does not release drivers open source, asahi is a reverse-engineering endeavour and does not support GPU. For nvidia, there are both proprietary and open-source linux drivers. CUDA and inference works on linux with nvidia. I would recommend checking out this video of Alex Ziskind to shop for a computer to run local LLMs: https://www.youtube.com/watch?v=mevUEQcumzU&t=224s. TL;DR besides Apple he recommends, DGX Spark, Tenstorrent Wormhole N300, AMD Radeon 7900 and NVIDIA RTX 5090.