Apple M1 Max: is it really at the level of a mobile Nvidia GeForce RTX 3080? It can be discussed – Frandroid

Is the Apple M1 Max at the level of a mobile Nvidia GeForce RTX 3080 as Apple boasts? Here are plenty of answers. Get ready, the comparison is not that easy.

While waiting for the first tests where we will have independent benchmarks that will give us a precise idea of ​​what the Apple M1 Pro and Max chips have to sell, let’s take a look at Apple’s claims concerning the graphics power of its chips.

As a reminder, Apple has unveiled its two new MacBook Pro, 14.2 and 16.2 inch, which are entitled to two home ARM chips intended for professionals: the M1 Pro and M1 Max chips.

As we specify, you have a dozen different configurations, which is not necessarily to be understood for neophytes. In any case, we can summarize the characteristics of the two chips like this:

• Apple M1 Pro : an 8 or 10 core CPU (6 to 8 high performance cores and 2 high energy efficiency cores), 14 or 16 core GPU, 16 core Neural Engine, 16 or 32 GB of unified RAM, up to 200 GB / s of memory bandwidth
• Apple M1 Max : a 10-core CPU (8 high-performance cores and 2 high-energy-efficient cores), 24 or 32-core GPU, 16-core Neural Engine, up to 64 GB of unified RAM, up to 400 GB / s of memory bandwidth

Regarding the integrated graphics chip (iGPU), it has no name because we are generally talking about an SoC (System On Chip) which integrates all the components. We just know that Apple has worked on a custom GPU architecture that goes from 14 to 32 cores depending on the versions. These cores benefit from a much larger memory bus. Apple has especially insisted on the efficiency of this iGPU, that is to say its power-consumption ratio.

In this case, Apple pitted the M1 Pro chip against a 4GB Nvidia GeForce RTX 3050 Ti, with the Apple chip achieving similar performance with 70% less power. From what we understand, this GPU block would consume around 30 Watts, compared to 75-80 Watts for the RTX 3050 Ti dGPU. In the graph provided by Apple below, we learn that the device used by this benchmark was a Lenovo Legion 5 laptop (82JW0012US) equipped with an AMD Ryzen 7 5800H CPU with 16 GB of DDR4 RAM memory.

In the graph below, Apple set about comparing its Apple M1 Max chip to two laptops. Apple does not specify which version of the M1 Max that was used, we imagine that it is the one with 32 cores. The comparison was made with two gaming laptops, the MSI GE76 Raider (IPS Full HD @ 360 Hz, Intel Core i7-10870H, Nvidia GeForce RTX 3070 and 32 GB of RAM) and the Razer Blade 15 Advanced (OLED 4K @ 60 Hz, Intel Core i9-11900H, Nvidia GeForce RTX 3080 in 16 GB version and 32 GB of RAM). These are two big configurations that we had the opportunity to test.

Most of the matrix size of the M1 Max chip is taken up by the 32-core GPU, with power reported at 10.4 TFLOP (teraflops). The graphics part takes up so much space that we could speak of a GPU surrounded by an SoC, and not of an SoC that hosts a GPU. You follow ? When it comes to performance, Apple struggles with the best on the market, thus comparing the performance of the M1 Max chip to a mobile GeForce RTX 3080. The M1 Max would therefore achieve similar performance with 100 Watts less power (60 Watts vs 160 Watts).

What seems obvious is that Apple is just rubbing shoulders with the most powerful portable machines on the market. These machines are bought by creatives, developers and gamers. Apple, for its part, obviously targets creatives and developers. As Cassim points out, in his editorial, this approach is very frustrating because it puts players aside. MacBooks don’t offer the versatility offered by AMD-Intel-Nvidia machines running Windows 10/11.

As you can see, the graphics provided by Apple are very imprecise and do not really make it possible to understand what was measured. What does this look like in reality?

The delicate comparison

Upon arrival of PS5 and Xbox Series X, we compared the two consoles lengthwise and broadly. They share the same architecture, provided by AMD, so it’s pretty straightforward. The teraFLOPS unit of measurement (TFLOPS) is often used, strongly supported by the departments marketing brands. Microsoft, for example, announced with great fanfare that the Xbox Series X will be entitled to a graphics power of 12 TFLOPS, or 12 teraFLOPS.

As you can see, the theoretical GPU power of the Apple M1 Max chip is comparable to that of the GPU of a PS5, but far below the chips of Nvidia. For comparison, the very last Nintendo Switch OLED is … 0.4 teraflops, while a PS4 is 1.84 teraflops. You have to be careful with this measurement. It is a way of measuring the raw mathematical computing power of a processor. The problem is, raw power isn’t enough to sum up the complex architecture of a processor or graphics chip. We will take other criteria such as architecture, bandwidth, use of RAM, difference in video memory capacity and so on.

All these elements make it risky to compare the teraFLOPS of processors using different architectures to conclude on a difference in performance, moreover when they come from different brands.

We must therefore instead turn to synthetic benchmarks and comparisons of real tests on software. While waiting for the possibility of getting their hands on machines, synthetic benchmarks let the scores obtained by the first users pass.

The first synthetic benchmarks

We thus obtained the scores of GFXBench, a tool which has a certain reputation in the industry. Apple’s M1 Max chip achieved 703.2 fps in the normal “Aztec Ruins” level (screen OFF, 1080p) and 275.9 fps in the high level (screen OFF, 1440p).

We have synthesized these first results in a graph, in green (screen OFF, 1080p) and in blue (screen OFF, 1440p), the tests of the mobile RTX 3080 obtained were carried out on a low consumption variant of the chip (its power is dependent on various variables such as the frame of the device for example). Note that screen OFF results are more accurate, as this removes screen definition as a potential factor.

What to conclude from this first test? First, we have no idea what the test conditions for the Apple M1 Max are, is it a 24-core or a 32-core version? Then, the Apple M1 Max chip is an unprecedented chip with unprecedented performance. We are talking about an ARM SoC, it should be remembered. On a laptop PC, the dedicated graphics chip often has its own video memory and its own active cooling system. This is not the case at Apple, the Apple M1 Max is an all-in-one chip with unified memory.

Finally, Apple has not designed its chip for games, so it will be much more relevant to use other synthetic benchmarks than those of GFXBench and especially to compare the performance of professional software under specific conditions. It will also be necessary to compare the consumption, when the laptop is on mains and on battery, this is where the Apple M1 Max shines and that bodes very good things for the future. One can imagine how Apple will be able to continue to increase the number of graphics cores in a more energy-intensive configuration, like that of a Mac Pro.

Completely consistent pricing positioning

You have to pay 3,209 euros for the cheapest configuration with an Apple M1 Max chip (24 cores and 16 GB of unified memory). At this price, there are very solid configurations of laptops, with GeForce RTX 3080 mobile, a powerful AMD or Intel CPU and 32 GB of RAM. Apple’s proposal appears to be, at first glance, fairly consistent in pricing. This statement is not necessarily valid on less expensive portable machines equipped with an RTX 3060 or 3070 dGPU, which could be compared with MacBook Pro equipped with an Apple M1 Pro chip, there are between 1000 and 1700. euros.

In addition, let us also mention the advance of Nvidia in certain areas like AI. GeForce RTX graphics cards, for example, incorporate RT cores dedicated to computing the ray tracing. Thus, new resources are allocated to developers and creatives, as well as gamers. What is more, many software aspects have been developed, including the DirectX Raytracing library on Windows 10 and 11 that allows developers to take advantage of technology.

Once again, Apple’s machines are certainly more efficient at working, especially since the performance obtained seems to be relatively stable even when the laptop is on battery (which is not at all the case on laptops) . On the other hand, they exclude, de facto, the players who wish to have a versatile machine to work but also to play.