According to Vadim Yuryev from Max Tech, as reported by Pulsestacks on March 28th, Apple's M3 Ultra chip is likely to feature an entirely new design, functioning as a standalone chip, rather than being composed of two M3 Max chips like the previous M1 Ultra and M2 Ultra configurations.
PulseStacks noted that this speculation originates from another blogger, @techanalye1, who pointed out that the M3 Max chip seems to have removed the UltraFusion bridging interconnect technology previously used to connect two chips. As a result, Yuryev infers that the upcoming M3 Ultra chip will no longer be able to achieve packaging two M3 Max chips together. This also implies that the M3 Ultra is expected to be the first independently designed chip in the Apple Ultra series.
Independent design will enable Apple to customize the M3 Ultra for professional high-intensity workflows. For instance, Apple could completely remove the efficiency cores and opt for a full-performance core design while adding more GPU cores. With such a design alone, the performance boost of a single M3 Ultra chip is almost certain to surpass the improvement from M2 Ultra to M2 Max, as there won't be efficiency losses from UltraFusion interconnect technology.
Yuryev also boldly speculates that the M3 Ultra might feature a new UltraFusion interconnect technology, allowing for the packaging of two M3 Ultra chips to create the "M3 Extreme" chip with double the performance. Compared to packaging four M3 Max chips, this design would yield even greater performance gains, while also potentially supporting larger unified memory capacities.
Exact details about the M3 Ultra are scarce at the moment, but there are rumors suggesting it will utilize TSMC's N3E process technology, the same as the A18 chip expected to be featured in the iPhone 16 series set to launch in the latter half of the year. This also implies that it will be Apple's first chip to utilize the N3E process, with rumors indicating that the M3 Ultra will be released alongside the new Mac Studio in mid-2024.
