MacBook Air M1 beats 16″ MacBook Pro i9 in Performance

Topic

New Reply

MacBook Air with M1 chip beats 16-inch MacBook Pro performance in benchmark test

…Earlier today, an independent analysis from AnandTech argued that the M1 chip in fact has the potential to be the fastest laptop CPU on the market, and the new benchmark results seem to prove this. The new MacBook Air with M1 chip scored 1687 in single-core and 7433 in multi-core tests.

For comparison, the higher-end 16-inch MacBook Pro model with an Intel Core i9 processor scores 1096 single-core and 6870 multi-core. The fact that the M1 chip in a MacBook Air was able to perform better than an Intel Core i9 processor in both single-core and multi-core seems extremely promising…

1 user thanked author for this post.

Nathan Parker

Reply | Quote

Replies

The Apple chip has some interesting design features that Intel and AMD would be wise to notice (and I am sure they will). It’s already known that as clock rates increase, power consumption increases more rapidly than performance. Intel bet on clock rate with the Pentium 4 series, though not all of the clock rate they had designed the chip for actually materialized, and they ended up with a CPU that ran hot, used a lot of power, had the highest clock rates in of any consumer CPU, and still performed worse than the lower-clocked AMD offerings that got more done with each clock cycle.

Intel is kind of doing the same now, it seems, with clock rates around 5 GHz for desktop turbo CPUs (higher than Pentium 4 ever achieved), relying on process node improvements to enable higher clock rates while keeping the (high) power consumption about the same. That requires certain architectural design parameters, and those hinder performance at lower clock rates (and at higher ones, but it’s supposed to be offset by the higher clock rates it enables).

Apple’s new chip is designed around getting more done at lower clock rates (instructions per cycle). It benefits with the lower power consumption of a lower clocked CPU, but without the performance loss.

I don’t think there’s anything that Apple can patent about the basic design parameters of using wide decoder units and other such things.  It remains to be seen how the x86 instruction set can work within a paradigm of lower clock rates and greater data path widths, but there’s nothing that prevents them from doing so that I know of. When Intel abandoned the Pentium 4 design in favor of the lower-clocked but more performant Core series, upon which all of their current i-series are based, they made a similar kind of move that Apple’s making with the new design (ignoring the platform change from x86 to ARM), and they did it within the existing x86 universe. Of course, that Core series has evolved once again to chase higher clock rates over greater efficiency, as it’s kind of a cheap way to get more performance. Overclockers can do the same with a few tweaks, but designing a better CPU is a little harder.

I still am not so sure of what the impact that this would have if the new platform emerges as the clear performance leader in the laptop or even desktop segment. One of the benefits of the x86 Macs is that they allow access to the large number of Windows programs out there, and the ARM Macs won’t have that. If Apple uses an emulation layer as Microsoft has with its ARM devices to run x86 software, the performance benefit will evaporate, as emulation comes with significant overhead. Apple would presumably see that as a temporary thing, as it was when they moved from the POWER architecture to x86, but that only works for the software that gets ported over. Apple, as a single source of both the OS and the hardware, can certainly dictate that Macs will now use any architecture they wish, but they can’t force third parties to release software for it.

If when these Macs hit the market, the native Mac ARM software is scarce and most of it will require x86 emulation, resulting in a PC that isn’t faster than x86 units for the software they want to use, and that can’t run the large Windows software library through BootCamp or the like, I would think a lot of people would hold off buying the new ones, and that could cause a chicken-and-egg thing. If Apple is able to get enough developers on board, something Microsoft tried and failed to do with its own Windows 8 (and then 10) phone, it could result in enough sales to convince more of them to get on board rather than simply let the Mac platform go and just develop for Windows.

Intel is bound to x86 in people’s minds, and it’s not that they refuse to try to escape from it. They tried with IA64, but AMD’s x86-64 extensions were backwards compatible with existing x86 software, and that ruled the day.

 

Dell XPS 13/9310, i5-1135G7/16GB, KDE Neon 6.2
XPG Xenia 15, i7-9750H/32GB & GTX1660ti, Kubuntu 24.04
Acer Swift Go 14, i5-1335U/16GB, Kubuntu 24.04 (and Win 11)

2 users thanked author for this post.

Nathan Parker, OscarCP

Reply | Quote

The units of the scores are not published, so I find impossible to make anything of these. If they are in some way measures of speed, then I find it intriguing that the single-core results are some 40% better (faster?) for the ARM CPU chip than for the Intel one, but their multi core results are a lot closer, the M1 being just some 10% faster than the Intel.

I am prepared to believe that the M1 chips, as far as performance goes, are probably better than the Intel ones, as Ascaris explains, but perhaps by a rather modest margin. The problem, right now, based on what I have seen, heard and read, is that the information being put out is not terribly informative, being mostly Apple’s PR and Mac fans exuberant lionizing. (Say I, who am very pleased with my Intel Mac and, am sure, some day I’ll be proud of my nth generation Mn Mac!)

For now, I think it’s better to wait for the brass band and chorus to get to the end of their current triumphalistic performance, pack up their instruments and go home to have a quiet dinner, then go to sleep. Later, the actual reporting of how the new machines and their software work, or don’t, coming from bona fides serious users, will start to draw the picture of what is really better and what the problems are with this new Macs and their software. And what are the unintended consequences of such a profound change in architecture, and which their remedies.

Surely good things are coming, eventually; but the advice is already out there when it comes to buying these new machines and their novel software, and can be summed up in one word: Wait!

Ex-Windows user (Win. 98, XP, 7); since mid-2017 using also macOS. Presently on Monterey 12.15 & sometimes running also Linux (Mint).

MacBook Pro circa mid-2015, 15" display, with 16GB 1600 GHz DDR3 RAM, 1 TB SSD, a Haswell architecture Intel CPU with 4 Cores and 8 Threads model i7-4870HQ @ 2.50GHz.
Intel Iris Pro GPU with Built-in Bus, VRAM 1.5 GB, Display 2880 x 1800 Retina, 24-Bit color.
macOS Monterey; browsers: Waterfox "Current", Vivaldi and (now and then) Chrome; security apps. Intego AV

1 user thanked author for this post.

Nathan Parker

Reply | Quote

Synthetic benchmarks often have no conventional units, per se, as they’re relative performance indices, not necessarily an actual measure of how many of [x] were done in a unit of time or how long it took to perform [y]. As such, “Geekbench score” is the unit. The result is not comparable to any other benchmark, but it can be compared to other runs of the same benchmark on other platforms, or at least that is the idea (and I don’t have any reason to doubt it).

Dell XPS 13/9310, i5-1135G7/16GB, KDE Neon 6.2
XPG Xenia 15, i7-9750H/32GB & GTX1660ti, Kubuntu 24.04
Acer Swift Go 14, i5-1335U/16GB, Kubuntu 24.04 (and Win 11)

1 user thanked author for this post.

Nathan Parker

Reply | Quote

Ascaris, thanks for clarifying those benchmarks, always assuming that those undefined units are what you believe them to be, which is not immediately obvious to me.

But, assuming that they are what you think they are, that still leaves the mystery unrevealed of why, with more cores (and even more virtual CPUs), the benchmarks become so much closer than with a single core. To me that is quite counterintuitive, to put it as nicely as possible.

Ex-Windows user (Win. 98, XP, 7); since mid-2017 using also macOS. Presently on Monterey 12.15 & sometimes running also Linux (Mint).

MacBook Pro circa mid-2015, 15" display, with 16GB 1600 GHz DDR3 RAM, 1 TB SSD, a Haswell architecture Intel CPU with 4 Cores and 8 Threads model i7-4870HQ @ 2.50GHz.
Intel Iris Pro GPU with Built-in Bus, VRAM 1.5 GB, Display 2880 x 1800 Retina, 24-Bit color.
macOS Monterey; browsers: Waterfox "Current", Vivaldi and (now and then) Chrome; security apps. Intego AV

1 user thanked author for this post.

Nathan Parker

Reply | Quote

OscarCP wrote:

The units of the scores are not published

Which units of scores ?

1 user thanked author for this post.

Nathan Parker

Reply | Quote

Alex, thanks for clarifying those ratings.

A well-written professional article must contain the definition of all relevant terms used. If this were one of the papers in my specialty that I am sent regularly by the editors of professional journals to review, I would have recommended that such a poorly written article be rejected outright.

Ex-Windows user (Win. 98, XP, 7); since mid-2017 using also macOS. Presently on Monterey 12.15 & sometimes running also Linux (Mint).

MacBook Pro circa mid-2015, 15" display, with 16GB 1600 GHz DDR3 RAM, 1 TB SSD, a Haswell architecture Intel CPU with 4 Cores and 8 Threads model i7-4870HQ @ 2.50GHz.
Intel Iris Pro GPU with Built-in Bus, VRAM 1.5 GB, Display 2880 x 1800 Retina, 24-Bit color.
macOS Monterey; browsers: Waterfox "Current", Vivaldi and (now and then) Chrome; security apps. Intego AV

1 user thanked author for this post.

Nathan Parker

Reply | Quote