The fastest Macbooks and Macs appear with the processor newcomers M2 Max and M2 Pro. However, the presentation of the two systems-on-chips (SoCs) makes one thing all too clear: Apple also only cooks with water. There are no revolutionary leaps in terms of performance and range of functions, but there are important innovations such as HDMI 2.1.
Apple is once again having its processors produced by the chip order manufacturer TSMC with 5-nanometer structures, although TSMC has now started 3-nm series production. Apple improves the performance with additional arithmetic units and larger caches, which makes the chips even larger: If the render image comparisons are accurate, the M2 Max has an impressive 510 mm². For comparison: Nvidia’s AD102 GPU for the GeForce RTX 4090 measures 608 mm².
Due to the consistent 5 nm manufacturing technology and without extensive changes to the CPU and GPU architecture, the M2 Max and M2 Pro should only calculate significantly more efficiently than the M1 Max and M1 Pro if the arithmetic units are not fully utilized. Then it says: More cores with a reduced clock require less energy than fewer cores with a higher clock. In return, the M2 versions get two additional CPU efficiency cores (8 performance plus 4 efficiency cores) and larger GPUs.
Moderate speed gains
With both the M2 Max and the M2 Pro, Apple promises a 20 percent faster CPU and a 30 percent faster GPU compared to their respective predecessors. Both models share the same CPU, there are only differences in the GPU, the media engine and the memory expansion. The neural engine of both models gets a 40 percent speed boost – its use is being expanded more and more, for example in image processing with Photoshop or sound mixing in djay.
In the The Geekbench database already contains the first results of the new Macbook Pros, in which the CPU cores consistently clocked at a good 3.6 GHz. This is probably a boost beyond the standard specification, with which the single-core value rose over 2000 points.
(Image: Apple)
GPUs still without ray tracing
The GPU structure with crooked 19 or 38 shader clusters is unusual, mainly because there are 10 such clusters in the normal M2. Actually, it is reasonable to assume that Apple could have designed the Pro and Max versions with 20 and 40 clusters, respectively, but deactivated a part in order to increase the yield of usable chips. The rendered die shots reflect Apple’s specifications, but do not have to be accurate in that respect.
Apple, meanwhile, is the last company to design powerful GPUs without ray tracing capabilities. Corresponding hardware acceleration would be helpful in render programs, which the company likes to promote. In addition, the media engines can still do nothing with the AV1 codec; instead there is ProRes support as usual.
The HDMI 2.1 support reveals the message indirectly: New Macbooks and Macs can drive 8K displays via HDMI at 60 Hertz or 4K displays at 240 Hertz. In future, it will also be possible to connect 4K 120 Hertz televisions to the devices without an active, expensive USB-C adapter. At least one additional display can be connected via Thunderbolt 4 including DisplayPort mode: The M2 Max can control a total of five screens and the M2 Pro three. Appropriately equipped Macbooks are no longer limited to an external monitor – a welcome innovation compared to the cheaper M2 Macbooks.
Comparison Apple Silicon M2 vs. M1 | ||||||
SoC |
M2 Max |
M1 Max |
M2 Pro |
M1 Pro |
M2 |
M1 |
production |
TSMC 5nm |
TSMC 5nm |
TSMC 5nm |
TSMC 5nm |
TSMC 5nm |
TSMC 5nm |
size |
~510mm² |
~432mm² |
~288mm² |
~246mm² |
~145mm² |
~120mm² |
transistors |
67 billion |
57 billion |
40 billion |
33.7 billion |
20 billion |
16 billion |
CPU |
||||||
performance cores |
8th |
8th |
8th |
8th |
4 |
4 |
P cores L2 cache |
32 MB |
24 MB |
32 MB |
24 MB |
16 MB |
12 MB |
P cores clock |
maybe 3.49GHz |
3.2GHz |
maybe 3.49GHz |
3.2GHz |
3.49GHz |
3.2GHz |
efficiency cores |
4 |
2 |
4 |
2 |
4 |
4 |
E cores L2 cache |
4 MB |
4 MB |
4 MB |
4 MB |
4 MB |
4 MB |
E-core clock |
maybe 2.4GHz |
2.06GHz |
maybe 2.4GHz |
2.06GHz |
2.4GHz |
2.06GHz |
GPU |
||||||
cores |
38 |
32 |
19 |
16 |
10 |
8th |
shaders |
4864 |
4096 |
2432 |
2048 |
1280 |
1024 |
FP32 processing power |
13.6 TFlops |
10.4 TFlops |
6.8 TFlops |
5.2 TFlops |
3.6 TFlops |
2.6 TFlops |
Storage |
||||||
RAM type |
LPDDR5-6400 |
LPDDR5-6400 |
LPDDR5-6400 |
LPDDR5-6400 |
LPDDR5-6400 |
LPDDR4X-4266 |
Max amount |
96 GB |
64 GB |
32 GB |
32 GB |
24 GB |
16 GB |
interface |
512 bits |
512 bits |
256 bits |
256 bits |
128 bits |
128 bits |
transfer rate |
409.6 GB/s |
409.6 GB/s |
204.8 GB/s |
204.8 GB/s |
102.4 GB/s |
68 GB/s |
System level cache |
? |
48 MB |
? |
24 MB |
8 MB |
16 MB |
Again with 5 nanometer structures
Meanwhile, there is a question mark over Apple’s reluctance to switch to TSMC’s 3 nm process generation. Normally, the company is always early – i.e. first – to new production processes and also pays a lot of money for it.
However, the advantages of new manufacturing processes are diminishing, especially with the large SoCs for MacBooks and Macs, because Apple is going the same way as AMD, Intel and Nvidia with growing caches, but SRAM cells can hardly be shrunk any further. This is why the question also arises at Apple: How can the scaling problem be avoided? Will SoCs consisting of several chiplets also be an option in the future apart from the Ultra series?
Meanwhile, memory manufacturers are making an innovation possible with 24 GB LPDDR5 memory chips, which allow crooked amounts of RAM on straight memory interfaces. The maximum configuration of the M2 Max thus increases from 64 to 96 GB, while the transfer speed remains the same at almost 410 GB/s. The M2 Pro is still limited to 32 GB. If you want more, you have to go for the expensive Max Macbooks – those interested in the Mac Mini will get nothing due to the lack of a Max model.
(mma)