blastdoor

It will be very interesting to see how this plays out for the Mac Pro. 

Next year, AMD is expected to have a 96-core processor with 12 channels of DDR5 memory (https://wccftech.com/amd-epyc-genoa-cpu-platform-detailed-up-to-96-zen-4-cores-12-channel-ddr5-5200-sp5-lga-6096-socket/). That thing will be a beast. 

I suspect Apple will still have the better core design and the better performance/watt. So Apple could compete with that thing. But will they choose to do so? I'm confident that some Mac Pro users would appreciate a mega-beast like that.

So, Apple will have to make a choice between:

1. design an ASi SOC competitive that really does target that market
2. lose the small number of (vocal and big spending) customers who want that super high end Mac Pro
3. stick with x86 in that super high end Mac Pro

I would love Apple to choose option #1. I don't think it would be entirely nuts to do it. There is such a thing as a 'halo' product. Those are worth doing sometimes. Also, if Apple were to replace x86 in their own servers/Cloud, then the 'market' for this super high end SOC might be bigger than just Mac Pro buyers. 

But the other two options are very realistic, too. I could really see it going any of these three ways. 

Hokey smokes, the Mac sales! 

It's quite the coincidence that dropping Intel from the Mac happens and Apple achieves Intel-like gross margins. I mean, I know that can't be the whole story, because so much of Apple's revenue comes from other sources. But their margins this quarter are CRAZY. 

tht said:

This type of news could be a consequence of the pandemic. If the pandemic was going to cause a problem, it’s going to be in 2021 and 2022, assuming enough people get vaccinated by the end of this year, which seems more of a hope than a plan now. 

In 2020, all the M1 hardware development was done in 2019. What they were working on was getting the supply chain for mass production ready. Hardware that was planned for 2021 release? Yeah, I could see a slowdown there. 

Maybe so. Although they had said it would take two years to complete the transition. If we get to the end of this year and the only models left needing an update are the big iMac and Mac Pro, then they might actually be right on track for making their two year timeline. 

robaba said:

Hope this doesn’t go through.  We need less consolidation within chip manufacturing, not more.

You're so, so wrong. To get more competition, we need more consolidation (up to a point). 

Due to the increasing fixed costs of developing new manufacturing nodes, companies need massive economies of scale. GloFo already lost the economies of scale necessary to compete at the cutting edge, and so they stopped competing. Intel has either already lost, or is on the cusp of losing, the economies of scale needed to compete with TSMC and Samsung. 

With this consolidation, maybe (*maybe*), the combined Intel+GloFo will have the scale needed to compete with TSMC and Sumsung. That will give the world three firms competing on cutting edge nodes. 

Without this consolidation, then we will most likely have two firms competing at the cutting edge. And I wonder how long Sammy can hang in there.

libertyforall said:

I am willing to bet this was part of a negotiated deal w/Apple to supply it chips in the future, in exchange for sale of their cellular modem business...  

I think you would almost certainly lose that bet. 

GloFo decided to abandon pursuit of more advanced processes a couple of years ago https://www.extremetech.com/computing/276185-globalfoundries-radically-restructures-kills-7nm-spins-off-asic-design-team. They are now focused on clients that use older nodes (12nm and above). Other than maybe some specialty chips here or there (and I'm not sure what those would be), Apple has little use for GloFo. 

If there's any kind of 'behind the scenes' negotiation here at all, I would guess it involves the US government and a desire for a more secure supply chain for defense contractors. With Intel owing GloFo assets (including old IBM assets in NY), in addition to their own, plus TSMC in Arizona, US defense contractors can be assured of a wide range of semiconductor nodes to meet all their needs.