Apple M3 SoC analyzed: Increased performance and improved efficiency

So that we can make a meaningful comparison between the different laptops, we not only look at pure performance in the synthetic benchmarks, but also the power consumption, from which we then determine the efficiency. The consumption measurements are always carried out on an external display so that we can eliminate the different internal displays as influencing factors. Nevertheless, we measure the overall consumption of the system here and do not just rely on the displayed values for the CPU and GPU.

Apple has been able to increase the single-core performance of the new M3 processor by around 20% over the old M2 and around 18% compared to the M2 Pro. Compared to the old M1 processor, the M3 is 30% more powerful. The power consumption of the CPU itself (according to Power Metrics) was around 6.5 watts at the beginning of benchmark testing before increasing to 5.5 watts later on, which is why the power consumption has increased slightly compared to the old M2. Despite the significantly higher performance, Apple has still been able to increase efficiency by around 18% compared to the M2 in the MacBook Pro 13, which is an excellent result. However, the lead over the passively cooled MacBook Air models is smaller.

Multicore performance has also increased by around 20% compared to the old M2 in the MacBook Pro 13, although the power consumption of the CPU cores has increased slightly too. The M2 consumes just under 20 watts, while the M3 needs 20-21 watts at full load. This puts the new M3 just ahead of the old entry-level 8-core M1 Pro, but all other MacBook Pro CPUs are still faster. On the other hand, efficiency has also increased here, but the lead of 8% over the M2 in the MBP 13 is not quite as high as in single-core load.

Now we compare the Apple M3 with the current mobile processors from AMD, Intel and Qualcomm. In selecting the comparison devices, we focused primarily on the U series chips from AMD and the U/P series from Intel. We've also included individual H45 and HS chips with comparatively low TDP values (max. 50 watts) in for comparison, even if these do not represent direct competition for the M3. We have also included the Snapdragon 8cx Gen3 in the Microsoft Surface Pro 9 in the tables.

Starting with single-core performance, we can clearly see what progress Apple has made with the M3. Despite lower power consumption, the M3's performance core can easily keep up with the performance cores of the current Raptor Lake CPUs from Intel, which require significantly more power. Even the most efficient Intel chip in our comparison, the Core i7-1355U, falls behind by a factor of 3. The current AMD Zen4 chips "only" fall behind by a factor of two in terms of efficiency, but do not match the performance of the M3 either. The Snapdragon 8cx Gen3 can only keep up to a certain extent in the Geekbench test (-28%), but falls far behind in the significantly longer R23 single test (-68%) and only just comes in ahead of the Intel CPUs in terms of efficiency.

The tide turns a little in the multicore tests, as some competitors from AMD and Intel offer better performance, especially in the Cinebench R23 test. Since the M3 does not support hyperthreading, the result is nevertheless still very good. However, the two U-series processors from Intel (only two performance cores) fall behind and the Qualcomm chip once again has no chance at all (-68%). In the short Geekbench test, the Apple chip is significantly better and is only narrowly surpassed by the AMD Ryzen 7 7840S.

In terms of efficiency, the differences are significantly smaller than in the single-core load and the Ryzen 7 7840U is practically on par. The other CPUs from AMD and Intel fall behind and the Qualcomm chip once again stands no chance here.

In all graphics tests, the new 10-core M3 GPU has a 10-22% lead over the old 10-core M2 GPU, but maximum power consumption has also increased from ~13.5 watts for the M2 to ~15 watts for the M3. However, all past Pro GPUs from Apple (whether M1 or M2) are even faster, even the base 14-core M14 Pro. 

Compared to the Windows competition, the M3 GPU clearly outperforms the AMD Radeon 780M and the Iris Xe Graphics G7 in particular in both 3DMark Wildlife Extreme Unlimited and GFXBench, but the Apple GPU has to concede defeat to the Radeon 780M in the Geekbench OpenCL test. Dedicated mid-range Nvidia GPUs (from RTX 3050 upwards) are significantly more powerful and naturally also consume much more power (from 35 watts upwards). We have also included these results for better classification.

When examining efficiency, we more or less have to use Witcher 3, as we have the most comparative values for this stored in our database and the CPU load is quite low. Of course, this is not an optimal starting point for the MacBooks and the values are not completely comparable with the Windows laptops, since only the emulated version runs on the Macs via CrossOver, which also costs some performance. We carried out this test on MacBooks for the first time last year, namely on the M2 Pro models, and they took top place in the comparison field and were later only overtaken by the M2 MacBook Air 15. The new MacBook Pro 14 M3 is 8% more efficient.

Shortly before the announcement of Apple's new M3 processors, Qualcomm presented the Snapdragon X Elite and showed benchmarks for two reference designs with 23 and 80 watts. Of course, we also compare these results with the new Apple M3, for which we employed the Cinebench 2024, Geekbench 6.2 and GFXBench benchmarks (Aztec Ruins Normal Tier Offscreen) used by Qualcomm.

Qualcomm has a noticeable advantage in multicore performance with both reference systems (depending on the test, between 17-33% with 23 watts or 27-71% with 80 watts). However, this is a comparison with the slowest of the new M3 chips and we can already reveal that the M3 Pro (12 cores) with 28 watts is faster than the Qualcomm reference design with 80 watts, at least in Geekbench Multi.

In single-core tests, the Apple M3 easily outperforms the strong-performing Snapdragon X Elite. It will be interesting to see how much power the Qualcomm chip actually needs under single-core load.

In the GPU benchmark from Qualcomm, the result is quite clear, as the 10-core M3 GPU (with a maximum consumption of 15 watts) is able to outperform Qualcomm's small reference system by 12% and is only 7% behind the 80-watt system. However, the 18-core GPU of the new M3 Pro is clearly superior at 580 FPS.