Apple Mac goes ARM with BIG SUR

The patches are here:
https://developer.blender.org/T78710

2 Likes

Cool. I’ll be curious to see what this entails.

Me not be a programer and just glancing at the commit notes, it looks like a port to ARM and nothing more. Or are there future implications for Metal?

That’s right.

1 Like

but what does this mean? there were working ARM builds of blender before that. did they require some manual changes before building and now there is an official patch which makes this work out of the box?

i guess if the apple transition turns out to be a success maybe players like qualcomm will give it a try too with mainboards + cpus for pcs?

They had to compile the libraries for ARM. Besides that, there are plenty of adjustments in Cycles to make sure it uses CPU accelerations.

1 Like

There were builds for ARM Linux. This is a patch for ARM macOS.

1 Like

Hi, working on Blender for ARM starts:

https://developer.blender.org/rB9b9cbabd96f5f3c8911c3670cd8ddd6789619c1a

Even I am not an Apple fan it would be nice if Blender for ARM is ready before Mac ARM systems are shipped.

Cheers, mib

I am not an Apple fan but prefer macOS hahaha.

So this is actually some good news.

I am still curious about the future of Blender openGL UI and if the UI would actually be ported to Metal too.

And then there is the big GPU question …

Hi, there he (Stefan Werner) go again:

https://developer.blender.org/rB52d469531b1ec1ea237b1b40a5e8dcbe06eb5c4c

Cheers, mib

2 Likes

I am not sure about those Youtube people

He makes a claim that metal API and optimisation helped speed up significantly

I agree with almost everything he said in that video except about the Mac Pro.

I expect the Mac Pro will move over to Apple Silicon too and with integrated Apple GPUs but they’ll use multiple SOCs in the Mac Pro to get the high core counts for the CPU and multiple GPUs. Apple’s desktop SOCs are said to be very different to their mobile SOCs so expect the new Apple Silicon Macs to be a step forward from where current Intel Macs are even those with dGPUs.

If the Mac Pro had two or four SOCs it’ll have two or four GPUs which is far more likely than Apple creating a single powerful PCIe based GPU for the ridiculously small Mac Pro market. It would be incredibly expensive to produce on such a small production run even if their Arm silicon is cheap to make.

Also GPUs on PCIe buses is everything they’re trying to avoid with their new strategy.

I don’t think Apple will update the Mac Pro one final time with Intel or AMD hardware, if there is another Mac Pro it’ll be Apple Silicon based and will be very different to the cheesegrater of 2019. There’s a chance the Mac Pro may be retired and the iMac Pro is the highest end Mac Apple will sell, only time will tell.

Good points actually

I think many assume the new macs will be iPad chips in a Mac which will not be the case

Only the instruction set will be the same.

The 2019 Macpro is still very good also when some say you can get it for cheaper as a self made pc. Commercial alternatives are equally priced.

Those machines will work for quite some time before they get replaced.

The 2019 Macpro has a very smart cooling system which with Apple silicon you don’t need so if there will be a Macpro it will be quite different.

they can’t cheat the laws of physics. that arm cpus will consume much less power when performance is comparable to desktop x86 chips (at the same node size) will have to be seen. i would be very surprised if there was much difference. a little maybe. :slight_smile:

Actually I disagree

RISC vs CISC has major advantages

Look at the amd thread ripper
It is not directly a radical new design but just a very smart x68 stacking technique

It is kinda funny / interesting that Apple goes back to RISC now

There are major architecture differences between apples SOC and intel/amd

Intel cpu gpu communication vs apples cpu gpu communication is another good example to show the difference

I agree with the term Apple does not go arm because they design their own chips unlike android who uses premade cpus

modern x86 cpus mostly use a risc-like instruction set too internally. only on the outside it still appears to be cisc.

it will be very interesting to compare performance per watt on the same node size. apple could be the first to get 5nm chips though.

As you say the current Mac Pro has some life left in it and when it comes to its natural EOL there may be no need for one.

Apple’s next Mac Pro might be rented from iCloud. A lot can change in 4-5 years particularly with the Cloud and I would be really surprised if Apple doesn’t have an iCloud strategy to rent hardware, think of HandOff but instead of your iPad handing off to your Mac it’s smartly uploading data to a cloud based Mac. You connect a keyboard, display and mouse to your iPad and it’s now a Mac Pro.

When you think about the suggested speeds of 6g you know the future is the cloud.

but decoding of instructions in x86 i suppose always will be more energy costly than pure RISC
Also current x86 does not have big.LITTLE, (but Intel’s Alder Lake will implement that).

I really hope that SoftBank wont sell ARM to Apple/Nvidia/Qualcomm/Intel so they could vendor lock that architecture.

Amd Turku killed intel recently so it is hard to predict the future of x86 but I buy into the argument of optimization.

Not everybody needs a threat ripper or Xeon system and for gaming the SoC in Xbox and PlayStation are pretty powerful

Obviously 3D where the more the better is excluded here

But writing text etc is fine even in hardware like an iPad

So it will be interesting to see if through apples decision a bigger movement to arm will happen

All those atom notebooks were nice web browsers and Chrome books well are still web browsers

Apples SoC or MS surfaceproX have vs x86 just fantastic battery life for mobile use

Desktop can very different

But then even there not having fans of your GPUs go on would be cool too for once

What I found most shocking was a report that the Mac ARM chip was going to be made with 2 nm manufacturing. It is easy to theorize chip companies have been holding off on their tech to they can come out with 20% faster every year, but for them to have 2nm tech is crazy impressive. To put that in perspective a water molecule is about 0.282 nm. Each transistor is the width of 7 water molecules!

1 Like