Apple is an existential threat to the PC

Here is some comparisons for 3DMark Wild Life Extreme, Baldur's Gate 3 and Cyberpunk 2077 (through Game Porting Toolkit and therefore translated from x86 to Aarch64). You can watch the video here.

3DMark Wild Life Extreme.jpg
BG3-1080p.jpg
BG3-1440p.jpg
BG3-2160p.jpg
 

Attachments

  • CP2077-4k.jpg
    CP2077-4k.jpg
    101.8 KB · Views: 4
Here is some comparisons for 3DMark Wild Life Extreme, Baldur's Gate 3 and Cyberpunk 2077 (through Game Porting Toolkit and therefore translated from x86 to Aarch64).
The 4060 and 4080 in the video are laptop GPUs though, with unkown TDP. The DF comparison was made a gainst desktop GPUs with the full TDP.
 
Yes, the M3 Max is a laptop SoC 🤷🏼‍♂️
No, there is no difference in performance between a laptop Mx Max and a desktop one, unlike AMD or NVIDIA GPUs.

The purpose of the comparison is to gauge the true performance of M1/M3 Max GPUs in comparison to PC GPUs and consoles APUs, which only makes sense if you compare against dekstop GPUs not laptop GPUs with their cut down nature, different TDP configurations even in the same SKU and their variable performance.
 
No, there is no difference in performance between a laptop Mx Max and a desktop one, unlike AMD or NVIDIA GPUs.

The purpose of the comparison is to gauge the true performance of M1/M3 Max GPUs in comparison to PC GPUs and consoles APUs, which only makes sense if you compare against dekstop GPUs not laptop GPUs with their cut down nature, different TDP configurations even in the same SKU and their variable performance.
Sure you can certainly compare them like that. I just want to point out that the form factor will throttle the Mx Max SoCs based on whether it's in the 14-inch or 16-inch chassis on sustained loads (as gaming represents).

There really is no arguing what the intended purpose of the Mx Max SoCs are meant for though. Otherwise they wouldn't wait 6 months before releasing it in the incredible small form factor (19.7 x 19.7 x 9.5 cm) Mac Studio.

It's first and foremost designed as a laptop SoC while, usually, becoming available in a low selling desktop variant.

The only data points we have are quite clear about this as well (Steam Hardware Survey).

LaptopDesktop
M1 Max82%18%
M2 Max85%15%
M3 Max100%0%

The last entry is, obviously, tongue-in-cheek.

It's all academic because Apple has not really made a truly "balls to the walls" desktop SoC with a power budget to match an Intel or AMD and Nvidia combo. Even the entire Mx Ultra SoC uses less power under load than an high-end Intel CPU.
 
Last edited:
I just want to point out that the form factor will throttle the Mx Max SoCs based on whether it's in the 14-inch or 16-inch chassis on sustained loads (as gaming represents).
That's true. That's why the DF comparison was made using the 16 inch form factor, which as I showed above gives the same fps as the M1 Max desktop.

Here are other links showing little performance differences between M1 Max in the 16 inch laptop form factor and M1 Max in the desktop form factor.



 
Yeah, we can also use that to infer the performance of the M3 Max GPU, which is around 40% faster than M1 Max in rasterization, this should put the M3 Max around the level of the desktop RTX 2080, which means M3 Max is about equal to PS5/Series X as well.
You can't infer anything though.

The M3 Max GPU is totally different from the M1 Max GPU architecture.

Just looking at the 3DMark Wild Life Extreme test should tell you as much.

3DMark Wild Life ExtremeScore
M1 Max17984
M2 Max25070
M3 Max31386
 
Just looking at the 3DMark Wild Life Extreme test should tell you as much.
Where are these numbers from?

The best I could find is the M3 Max being 55% faster than M1 Max in that benchmark, which is not that far off the RTX 2080 estimation considering the gains in the Wild Life test is an exception to the rule rather than the typical gains over the M1 Max.

 
Keep in mind, these Mx Max SOCs are in MacBook Pros which cost over $4000 or even $5000.

So if they match PS5/XBSX SOCs, that maybe be notable but they have to be relatively small volume.

Average Selling Price of MacBook Pros is probably around $3000, maybe less.

So these laptops with high-performance SOCs are not an existential threat to PCs, where the ASP of laptops is $600-800? A fraction of these Mx Max MacBook Pro SKUs?
 
Where are these numbers from?

The best I could find is the M3 Max being 55% faster than M1 Max in that benchmark, which is not that far off the RTX 2080 estimation considering the gains in the Wild Life test is an exception to the rule rather than the typical gains over the M1 Max.

From the review I linked to earlier. It's right there in the pictures I posted in this thread except the M1 Max, which I found from the first result on Google.
 
Microsoft going all in on AI when Windows is such a total clusterfuck is pathetic.

They don't have any usable GUI API which isn't in maintenance mode except webview (WinRT/WinUI is a joke). They don't have a convenient way to sandbox win32 applications. They don't have single logon account transfer/sync. MacOS did live tiles right (ie. Live Tiles Anywhere) just as Microsoft made the most castrated widget system ever seen. System design of 99% of windows laptops is terrible. Plus of course their fundamental inability to compete because they lack a full ecosystem.

Their cloud division will want to earn sweet subscription fees for copilot on MacOS too ... no matter how good, AI isn't going to add dick to Windows competitiveness.
 
Last edited:
Supposedly the Qualcomm SOC that they will use in the next Surface devices are faster than M3.

If true, then Apple Silicon advantage would be over.
 
Supposedly the Qualcomm SOC that they will use in the next Surface devices are faster than M3.

If true, then Apple Silicon advantage would be over.
12 performance cores at 80 Watt TDP to beat the base M3 with 4 high-efficiency cores and 4 performance cores at 23 Watt TDP. The problem with Windows laptop standby battery life is all the un-core stuff gobbling up power while doing nothing. If Qualcomm can solve that it will surely help a lot but then the other problem arises as nothing noteworthy on Windows is Aarch64 yet.

I will wait for final hardware to pass judgement though, although with a die size as big as faster Apple Silicon SoCs, up to 80 Watt TDP and a slower GPU (that doesn't support DX12) things doesn't look all rosy.
 
The 23W configuration and single core performance was good too.

It seems they can fit more and faster cores and get a competetive GPU in there at a node behind. Architecturally that's great, shame there is no platform for it.
 
The 23W configuration and single core performance was good too.

It seems they can fit more and faster cores and get a competetive GPU in there at a node behind. Architecturally that's great, shame there is no platform for it.
The problem is that with 0% market penetration and five devices no one is going to port from x86 to Aarch64 unless Microsoft brings heavy incentives to bridge the issue (or get serious about vertical integration with hardware, software and perhaps moving their consoles to Aarch64 as well).

It's still an issue for macOS four years later in the transition despite currently sitting at around a 70% (Aarch64) and 30% (x86) split. My only data source is the monthly Steam Hardware Survey though, so big caveat there 😅

Many games still rely on Rosetta2 as developers struggle to port to the Metal graphics API and Aarch64, although most professional and serious applications have already made the transition with big performance improvements and responsiveness to boot.

SteamSurveyMarch2024.png
 
So MS is going all ARM with Surface or it's just a couple of SKUs where they will use this Qualcomm SOC and the rest will be Intel?

And if it really is a 80W part, over 3X the power envelope of iPads and MacBook Airs, maybe they are specifically targeting power or gaming devices?
 
I'm not sure about this, but two power profiles were mentioned (80W and 23W), which sounds more like total SoC power than total device power.
The performance numbers released last year seems to suggest that it's similar to a full M3 Pro, which has the same core counts (12 cores). Even 80W SoC power does not sounds too bad as it's the peak power and we don't know how much power it actually consumes when running something like Geekbench. M3 Pro is ~30W when running some CPU benchmarks.
Of course, the main problem Microsoft will face is software. Right now most software on MacOS are already ported to ARM, but the majority of software on Windows aren't. This will take time.
 
Back
Top