Apple Silicon has better performance macs on the way, but how fast are they going to be?
Apple is taking a two-year transition to its own Apple Silicon from Intel and AMD chips. The low-end items are already included in the M1 chip of the first generation, the MacBook Air, the MacBook Pro 13-inch, the Mac Mini, and the iMac 24" Both of these are well supported by the M1 chip that is essentially an A14 on steroids. Like the A12X, which has twice the CPU, GPU cores, and memory bus width, it is M1 compared with A14. The same applies to the A12X. So Apple put it in the latest iPad Pro.
This is a huge power to the lower-end Mac, but the bigger MacBook Pros and iMacs really will need a quicker move if the powerful Intel CPUs and AMD GPUs are to be replaced in current production. And, well, the Mac Pro is a different set of specifications and is likely to be the final transition product for Apple Silicon. So it's worth our hopes with reports that the M2 (or maybe the M1X is called?) enters production. What will be the next chip from Apple Silicon?
M1X or M2?
We have seen "leaked" specs and benchmarks for what is known as the M1X at CPU-Monkey.
These should be taken with a grain of salt since unreleased information on the web is always incorrect. They still have a very good collection of samples for Apple Silicon's next move.
It's almost academic if Apple calls an M1X or an M2 chip. M2 would be more meaningful if the chip were to be focused on the architectural enhancements of the processor A15 coming on iPhones and iPads this fall. To be honest, an M-series chip on A15 architecture would seem difficult to believe and would be already ready for development, but I suspect that any Apple silicone chip would be built on the A14 and M1 in 2021.
The A15 is expected to increase single-core efficiency across the A14 and M1, but Apple expects to strive for more nodes in its next Apple Silicon Chip over faster nodes.
The A15 is anticipated to be carried out on a somewhat improved 5-nanometer process, with a slightly higher CPU and GPU output along with the planned annual image/video and neural engine updates. This year too, I believe that Apple will eventually move to LPDDR5, giving it a lot more bandwidth. Such changes are precluded in an M1X, but in an M2 we would expect them.
Actually, even if the A14 and M1 are still dependent, Apple can call the M2 on the next chip. What is in the name? What is in the name? But the chip I expect to be an extension to the A14 and M1 lines in higher-end macs this year, whatever it's called.
A 12-Core CPU
The A14 has four CPU core high-efficiency (lower but low) and two HPCs (very fast but high-power). The M1 has maintained the four efficiency cores and doubled the high-performance cores to four, for 8 CPU cores in total. It makes sense to duplicate it again to eight high-performance cores for the next evolution, totaling 12 CPU cores.
These high-efficiency core cells definitely do the most when it comes to high-end performance and benchmarks. Intel processors have six or eight cores for the bigger 27-inch iMacs and 16-inch MacBook Pros (even 10 as an option in the 27-inch iMac). Matched from core to core, Apple's chips are much quicker than Intel's, so the upturn to eight high-performance centers will make Apple win a lot.
The M1 almost matches an 8-core Core i9 with just four high-performance cores and a single-center is much quicker. Imagine 8 high-quality cores!
A 16-core CPU
The M1 doubled the number of GPU cores of the 4-8 A14, and we believe the next step is to double the Apple Silicon. The size of a system-on-chip that is already massive would probably burst with a 16-core GPU (giving Apple 256 running units). But this is the most important of all updates for replacing what Apple already has on the market. The larger MacBook Pros and iMacs have discreet AMD Radeon graphics manufacturers with their own high-speed memory pool.
The 16 inch MacBook Pro provides a 5500M or 5600M Radeon with a special graphics memory of 4GB or 8GB and we are looking at around five teraflops of computing capacity. The iMac can be 27 inches, from a 4GB RAM Radeon Pro 5300 to a 5700XT 16GB RAM. We look at a spectrum of about 8 teraflops with twelve teraflops, perhaps more importantly a lot of high-speed GDDR6. We look into the processing capacity.
We were very impressed by the GPU performance in the M1 – it is a little more than its 2.6-teraflop weight class – but if the GPU's Radeon chips Apple's products are to be speeded up today, it still needs a major boost.
Intel still has a long way to go to match its Radeon Pro 5500M (let alone the faster iMac GPU) with its integrated graphics.
Double the RAM
The M1 can be either an 8GB or 16GB memory configuration. We find that the 8GB version is much better than a comparable Intel-based Mac, but it's nevertheless predicted that higher-end Macs can work hard with lots of memory.
The MacBook Pro 16-inch begins today with RAM 16GB and is up to 64GB configurable. The 27-inch iMac begins with 8GB but can reach 128GB. I doubt that Apple will try to match the highest RAM specs on these computers, but the next Silicon chip is essentially 16GB and 32GB.
More fascinating for me is how Apple can increase the bandwidth of its memory. The M1 increased the 64-bit A14 to the 128-bit memory bus. That means that LPDDR4x can deliver approximately 68 GB/sec. This is great for a super-thin-and-light laptop, also outstanding in the 27-inch iMac, but the Radeon graphics chips have a bandwidth of over 100 GB/sec dedicated graphics memory.
Apple's graphing processors are very powerful in terms of bandwidth, but I believe that a 16-core GPU will soon be bottlenecked by memory bandwidth, especially if it had to share the memory with all those CPUs. It's definitely possible to increase to a 256-bit memory interface and the chip will really fly.
What of the other topics?
If the next Mac processor is focused indeed on the A14 and M1, then we expect to remain unchanged from the A14 and M1 all the "other stuff" inside the chip - the audio and video processing, the encoders, and decoders, and the Neural Engine. If the next silicon chip from Apple really takes its heritage from the design of the next A15 through some engineering features, then we hope to enhance all these functions. In particular, Neural Engine will probably be upgraded significantly, because Apple is all-in to AI and machine learning.
Increasing the count of CPU and GPU would make the next Apple silicon chip an absolute screamer even if it didn't boost all that 'un-core,' especially if Apple doubles the memory bandwidth. It will also make the chip very enormous: The M1 has already an enormous 16 billion transistors and you would hope to have at least 24 billion versions of CPU and GPU cores doubled. That's not unheard of, but that will make the processor, far and away, the largest system-on-chip on the market, far greater than the chips in the Xbox series X or PlayStation 5. That's not unheard of. That's what Nvidia is for.
To help their work, Newsmusk allows writers to use primary sources. White papers, government data, initial reporting, and interviews with industry experts are only a few examples. Where relevant, we also cite original research from other respected publishers.