[g4.6pre-65-g6701cd284] Freezes on Apple Silicon (M1 SoC)

[Rachael]

Well - apparently the M1 Macs still have Boot Camp.

Boot Camp requires a Mac with an Intel processor.

I stand corrected. Apparently Linus used a virtual machine - my bad.

It is still possible, however, to run it natively - that I am absolutely sure of - but I have no idea how difficult it would be or how much effort it would entail. It likely would not take advantage of Apple's ARM extensions, either, and likely also would not have proper drivers to run the hardware in said machine.

[Graf Zahl]

Apple would be smart to allow it - it'd give them an edge on that market that may be impossible to overcome later. But I somehow doubt they'd go that route.

[Warrex]

All I say about that is: This is the first chip of this family. It cannot be expected to be the ultimate killer.

Soon:

https://www.macworld.co.uk/news/12-core ... o-3800060/

https://www.notebookcheck.net/Apple-s-M ... 468.0.html

[Rachael]

Adding more cores is nice and all but core speed is still ARM's weakest point, in my opinion. But the M1 surprised many of us - including me - so we'll see how that pans out.

[Cacodemon345]

Any chance of the competition coming up with a ARM chip able to compete against Apple this year? I know of the Samsung Exynos chip with AMD RDNA 2 GPU but I don't believe it will make its way to the laptop market.

[Graf Zahl]

I don't expect anything big to come out this year, maybe some more performant laptop with a better CPU than Microsoft's offering.
That part doesn't really matter that much. Performance is not everything and even despite its power the M1 also comes with some really annoying limitations - it's not all rosy in Apple land.

This was a severe kick in the ass the CPU industry needed, but not a revolution.

[Warrex]

Adding more cores is nice and all but core speed is still ARM's weakest point, in my opinion. But the M1 surprised many of us - including me - so we'll see how that pans out.

As I wrote clock for clock it already has the highest single threading performance. Period. And otherwise do not worry:

"One aspect we’ve never really had the opportunity to test is exactly how good Apple’s cores are in terms of memory bandwidth. Inside of the M1, the results are ground-breaking: A single Firestorm achieves memory reads up to around 58GB/s, with memory writes coming in at 33-36GB/s. Most importantly, memory copies land in at 60 to 62GB/s depending if you’re using scalar or vector instructions. The fact that a single Firestorm core can almost saturate the memory controllers is astounding and something we’ve never seen in a design before."

(Source: https://www.anandtech.com/show/16252/ma ... -m1-tested)

Any chance of the competition coming up with a ARM chip able to compete against Apple this year? I know of the Samsung Exynos chip with AMD RDNA 2 GPU but I don't believe it will make its way to the laptop market.

You can buy a Server: https://www.anandtech.com/show/16315/th ... tra-review

[Graf Zahl]

As I wrote clock for clock it already has the highest single threading performance. Period. And otherwise do not worry:

Well, it still does not soundly beat my 8 year old i7 on GZDoom, so this is certainly not everything. Of course with the numbers at hand I cannot tell how much of it is waiting for the GPU.

[Warrex]

As I wrote clock for clock it already has the highest single threading performance. Period. And otherwise do not worry:

Well, it still does not soundly beat my 8 year old i7 on GZDoom, so this is certainly not everything. Of course with the numbers at hand I cannot tell how much of it is waiting for the GPU.

I am refering to Cinebench (see above) and Geekbench (https://browser.geekbench.com/v5/cpu/co ... ne=6182840). As I said it is only rivaled by AMDs ZEN3 CPUs. In contrast to you I cannot say anything about GZDoom's CPU requirements and how well it is optimized for one architecture or the other. I would assume that GZDoom is optimized for x86 and not for ARM64. Maybe you want to share some info on that.

[dpJudas]

The anandtech article has this note about its benchmarks: "As we’ve had very little time with the Mac mini, and the fact that this not only is a macOS system, but a new Arm64-based macOS system, our usual benchmark choices that we tend to use aren’t really available to us". When that's what they had available their conclusions needs to be taken with a certain grain of salt.

[Graf Zahl]

Benchmarks aren't everything. I'd rather judge by measuring real life applications. It wouldn't be the first time some hardware or compiler was designed around cheating a benchmark.

GZDoom wasn't optimized for anything. It's plain vanilla C++ code doing nothing special. IMO if an architecture needs specially designed code to be fast, it's defective by default.
So, to sunmarize: From what I have seen so far, the M1 is surely faster than any other laptop chipset I have seen, but there is no magic - depending on the task at hand a modern desktop CPU still has the upper hand.

[Guest]

Thank you for the info!

CineBench is a real life application btw. as it is based on Cinema 4D. And - as you know - in real life applications get heavily optimized for the architecture or platform they run on (see x264 / x265 with their ASM code and finely tuned use of the x86 extensions). So I would not know which benchmark would convince you anyway.

The good thing is I do not need to. I am just happy that GZDoom now runs fast on my Mac and that your views on Apple and ARM do not seem to be as pessimistic as they were in 2018 (ZDoom on Mac OSX Thread).

Thanks again to Alexey for the quick fix!

[Graf Zahl]

Thank you for the info!

CineBench is a real life application btw. as it is based on Cinema 4D. And - as you know - in real life applications get heavily optimized for the architecture or platform they run on (see x264 / x265 with their ASM code and finely tuned use of the x86 extensions). So I would not know which benchmark would convince you anyway.

Some general purpose benchmark. So it's exceptionally good at (de)compressing videos, which is not surprising, considering its coprocessors.
And no, most applications do not get platform specific optimizations. Only heavy duty stuff like video processing gets that because here it translates into significant time savings.

Most code out there is actually written in a mostly platform independent manner to increase its reusability.

[Warrex]

Some general purpose benchmark.

Name one. In the Anandtech Article I sent you they included the SPEC benchmarks. Is does not get more industry standard and general purpose as this I might argue:

"In the overall new SPEC2017 int and fp charts, the Apple Silicon M1 falls behind AMD’s Zen3 in the integer performance, however takes an undisputable lead in the floating-point suite. Compared to the Intel contemporary designs, the Apple M1 is able to showcase a performance leap ahead of the best the company has to offer, with again a considerable strength in the FP score. While AMD’s Zen3 still holds the leads in several workloads, we need to remind ourselves that this comes at a great cost in power consumption in the +49W range while the Apple M1 here is using 7-8W total device active power." (https://www.anandtech.com/show/16252/ma ... 1-tested/4).

So it's exceptionally good at (de)compressing videos, which is not surprising, considering its coprocessors.

Yes, it is but this does not relate to what I wrote. Cinebench / Cinema 4D - were it shows clock for clock class leading ST performance - is a 3D software rendering suite and x265 where it does not show class leading performance but good performance is a pure software HEVC encoder. So its fixed funtion HW encoders and decoders don't help the M1 there at all and the exessive amount of x86 optimization works again it.

[Graf Zahl]

were it shows clock for clock class leading ST performance -

Come on!
Parroting those marketing bullshit phrases does not convince anyone.
All I see is that with GZDoom it performs roughly the same as GZDoom when using Vulkan, and I'd give it a few percentage points for the MoltenVK overhead, but if this was "world class leading performance" the difference would be larger - my CPU is 8 years old, after all.