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.
You're Apple. You just built a 40-core SoC for Mac Pro. Now what?
- Thread starter senttoschool
- Start date
- Sort by reaction score
If you are primiarly running VMs , you can log into the hypervisor and do "under the instance" management about just as effectively as IMPI can under instances that are running 'raw iron' on some other system.
saying that apples hardware sucks for any real server use.
NO IPMI
NO Storage hot swap bays
Base Storage change needs an 2th system to reload it and are forced to wipe
low max ram
-------
apple does not have there own hypervisor os
Apple Reportedly Building M2 Ultra and M4-Powered AI Servers
Apple has started building its own AI servers that use the M2 Ultra chip, Haitong analyst Jeff Pu reports. Pu provided the analysis based on supply...
www.macrumors.com
Here we go. I think this is how Apple can make Ultra/Extreme chips worth it.
I'm guessing the M2 Ultra is just Apple testing. The M4 remark is likely to be the M4 Ultra or Extreme. I doubt the base M4 will be used.
My current prediction is that Apple will eventually unveil a premium Siri LLM service that costs $10-$20/month. They might want the free version to only run locally and the premium version will be more powerful and will require an Ultra/Extreme SoC in the cloud. Both versions will be optimized for Apple Silicon.
Apple can avoid Nvidia prices, optimize for their own chips, and be able to actually acquire enough AI chips.
Last edited:
On second thought, this "AI" server chip by Apple might not be a SoC. It could be a chip that only contains the Neural Engine but significantly larger than any NPU on Apple Silicon. IE, instead of a 16-core ANE, it's 256-core or something similar.
It'd be similar to Google's TPU, Microsoft's Maia 100, or Amazon's Inferentia. Basically all huge NPUs.
It's possible that Apple could do both: using full SoCs in their data centers and using a custom ANE chip.
It'd be similar to Google's TPU, Microsoft's Maia 100, or Amazon's Inferentia. Basically all huge NPUs.
It's possible that Apple could do both: using full SoCs in their data centers and using a custom ANE chip.
XCode Cloud was announced last year. And MacStadium, Amazon , etc were offering Mac instances in the cloud since about 6 months after the transition and had Intel ones before.
This server runs a thinned out version of iOS (according to "state of Platform" video). No local persistent storage ( for the service presumable). It boots primarily with just iOS iBoot protocols so presumably there is some boot SSD image. Just compute with little storage. ( similar in some respects to XCode Cloud which doesn't want your code repositories. Primarily focus on ephemeral builds/testing to moving a finished app off to 'app store' submission for any sense of persistence. )
The chatter about this is going into similar business as AWS , Azure, etc ... definitely not. This is an appliance (specialized to do one specific task) at least as much as a server. More closer to just having a very long distance eGPU that only does a few specific tasks.
Since it is a highly specialized appliance, all the generic modular features to compete in the general server workload market are not likely going to come into play. Historically, Apple hasn't paid much head to workstation/server GPUs with ECC. I wouldn't hold my breadth on that front either. (bit errors ... just a hallucination .. just start over later on a different server (if you can). )