Thought experiment: What if Apple stops making computers one day?

[richmlow]

Thats your problem, you are projecting an Intel MacBook Pro which are going to be obsolete pretty soon. Apple has moved on to a much more futuristic and better Architecture.

Hello TechnoMonk,


A current (2023) state-of-the-art Apple Silicon computer (PowerBook, Mac Studio, Mac Mini) would NOT last for 10+ for me. This is because of the current restrictions (non-upgradable RAM, no expansion, etc.) on the Apple Silicon chips.

That is my point.


richmlow

 

[TechnoMonk]

Hello TechnoMonk,


A current (2023) state-of-the-art Apple Silicon computer (PowerBook, Mac Studio, Mac Mini) would NOT last for 10+ for me. This is because of the current restrictions (non-upgradable RAM, no expansion, etc.) on the Apple Silicon chips.

That is my point.


richmlow

You just made my point about the legacy mindset. Unified memory and CPU/GPU computing are most likely the future. Apple, NVIDIA, and AMD will squish Intel. NVIDIA didn't offer to buy ARM for fun; they all want the piece of unified CPU/GPU with integrated high bandwidth RAM.
The latest 4090 or 3090 on my workstation with 128GB RAM, AMD thread ripper processor runs circles around my M1 Max. The big problem for me is GPU runs out of memory, and the extra RAM needs to be more helpful. Sure, I can reserve lower ram on GPU, but that slows down, and the GPU load goes up with throttling.
Apple is in the right place with unified High BW RAM for both CPU/GPU. Anyone needing legacy computing can use Windows or Linux Workstations.

 

[kschendel]

A current (2023) state-of-the-art Apple Silicon computer (PowerBook, Mac Studio, Mac Mini) would NOT last for 10+ for me. This is because of the current restrictions (non-upgradable RAM, no expansion, etc.) on the Apple Silicon chips.

That is my point.

You assume that the resources you need for your workloads will continue to expand as they have in the past. I admit that that's a reasonable assumption, but I believe it's becoming less true with time - or, said another way, the workloads that require ever-expanding resources are likely to become more and more specialized. General office tasks are already easily handled by 10-year-old computers, perhaps with a bit more memory than was expected back then. (By which I mean 4GB memory and the like. My 10-year-old "office" computer was configured with 16GB memory and is fine.)

I can imagine that certain kinds of database serving tasks, and maybe certain kinds of content creation, will continue to expand. Even there, though, I wonder; we've hit 4K and it's unclear to me that there is any visual improvement by going higher, except for very large screens which may well be unsuited for ordinary consumer use.

For my own work, the only thing I expect to need (ok, WANT) significantly more of than I can easily configure today, is CPU power, and that's only because I want my compiles to take 1 or 2 minutes instead of 15. Cue the tiny violins.

 

[russell_314]

You just my point about the legacy mindset. Unified memory and CPU/GPU computing are most likely the future. Apple, NVIDIA, and AMD will squish Intel. NVIDIA didn't offer to buy ARM for fun; they all want the piece of unified CPU/GPU with integrated high bandwidth RAM.

Exactly this. You can't replace the tubes in your radio anymore. Technology moves on. Computers are going to be one piece non serviceable items. It's harder for some people who have been in one mindset for decades to grasp something totally new. I love nostalgia but it's in the past.

 

[sam_dean]

You assume that the resources you need for your workloads will continue to expand as they have in the past. I admit that that's a reasonable assumption, but I believe it's becoming less true with time - or, said another way, the workloads that require ever-expanding resources are likely to become more and more specialized. General office tasks are already easily handled by 10-year-old computers, perhaps with a bit more memory than was expected back then. (By which I mean 4GB memory and the like. My 10-year-old "office" computer was configured with 16GB memory and is fine.)

I can imagine that certain kinds of database serving tasks, and maybe certain kinds of content creation, will continue to expand. Even there, though, I wonder; we've hit 4K and it's unclear to me that there is any visual improvement by going higher, except for very large screens which may well be unsuited for ordinary consumer use.

For my own work, the only thing I expect to need (ok, WANT) significantly more of than I can easily configure today, is CPU power, and that's only because I want my compiles to take 1 or 2 minutes instead of 15. Cue the tiny violins.

You'd only need to replace for reasons unrelated to preventive maintenance if your use case changes.

Many here extended the useful life of their Macs with OpenCore Legacy Patcher (OLCP) because their use case has not changed and their situation does not make preventative maintenance a priority.

 

[sam_dean]

Hello TechnoMonk,


A current (2023) state-of-the-art Apple Silicon computer (PowerBook, Mac Studio, Mac Mini) would NOT last for 10+ for me. This is because of the current restrictions (non-upgradable RAM, no expansion, etc.) on the Apple Silicon chips.

That is my point.


richmlow

RAM is fixed at purchase but SSD can be added externally.

What triggers a replacement for me would be after the final Security Update released.

 

[richmlow]

OK.....let me be very specific.


1. I need more RAM and expandability than what the current Apple Silicon computers offer.

2. I hope that future iterations of Apple Silicon CPUs will give me "enough" RAM.

3. If not, then I will have to go to the PC world for my math research computational needs.

4. At that point, I'll have a Mac which will be used to surf the web, play music, watch movies and perform other lightweight (and pedestrian) tasks.


richmlow

 

[nvmls]

On our end it's easy, no Mac no more Apple devices. The Mac is basically the pillar for the rest.

 

[AJK13K]

I will know it is the end times!

 

[kschendel]

2. I hope that future iterations of Apple Silicon CPUs will give me "enough" RAM.

3. If not, then I will have to go to the PC world for my math research computational needs.

That's a valid use case. I have a feeling though that you will quickly bump your head on PC limitations as well. All desktops I know of top out at 128GB. Threadripper hits 256GB, I think, but very expensive. If you want more than 128, with any serious likelihood of needing yet more, you're into the server world, either Xeon or EPYC. Both are very expensive both for CPU's and for the associated components such as motherboards. I don't see Apple Silicon / ARM going that way any time soon. It's not the same world as when consumer Xeon was a realistic thing.

 

[turbineseaplane]

If the Mac goes away, I go away as an Apple customer

 

[TechnoMonk]

OK.....let me be very specific.


1. I need more RAM and expandability than what the current Apple Silicon computers offer.

2. I hope that future iterations of Apple Silicon CPUs will give me "enough" RAM.

3. If not, then I will have to go to the PC world for my math research computational needs.

4. At that point, I'll have a Mac which will be used to surf the web, play music, watch movies and perform other lightweight (and pedestrian) tasks.


richmlow

Looks like what you need is research computational software/libraries to start using GPU. It’s 40X faster computing mathematical simulations using GPU, compared to CPU. when I started using Numba instead of regular Numpy in Python. Some of those 20-30 G data sets used to take an hour, take couple of minutes with GPU.

 

[pacalis]

Hello sam_dean,


You are right. It might be the case that a current iPhone is already more "powerful" than my 2013 Mac Pro. However, an iPhone cannot run Mathematica!

The other day, I was running a custom-designed Mathematica program / calculation on my 2013 Mac Pro. It took 4 hours and ended up using 64GB RAM (towards the end of the calculation)!


All the best,
richmlow

I think your general use case is going away as mathmatica moves to the cloud.

 

[richmlow]

Looks like what you need is research computational software/libraries to start using GPU. It’s 40X faster computing mathematical simulations using GPU, compared to CPU. when I started using Numba instead of regular Numpy in Python. Some of those 20-30 G data sets used to take an hour, take couple of minutes with GPU.

Hello TechnoMonk,


Indeed.

Do you know any resources that you can recommend for GPU computing with Mathematica? I would definitely like to learn more about this possibility.

Any info. is greatly appreciated.


richmlow

 

[RokinAmerica]

The other 90% of the computing world will survive.

 

[sam_dean]

Hello sam_dean,


Your question "Are you planning to replace your 2013 Mac Pro Xeon with a 2023 Mac Pro M2 Ultra" is a great question!

The answer is very dependent on the tech specs for the anticipated 2023 Mac Pro. This next computer purchase will be used specifically for mathematics research. As such, I want to have the following:

192GB RAM
16-cores

I will not have extensive need for GPU capabilities.

An additional challenge is that Apple (currently) designs and sells "throw-away" computers. Almost all of the hardware is locked-down with no expandability/upgrade/repair options. To further compound the problem, Apple releases poor quality OS versions every year. The current Apple computer platform (laptop, all-in-one, desktop) does not have the longevity of the Apple computers of times past.


All the best,
richmlow

I did a cost analysis of your tech specs vs a 2022 Mac Studio & 2019 Mac Pro

Base model 2019 Mac Pro starts at $6k with these specs

- Xeon 8-core
- 32GB RAM
- 8GB GPU
- 512GB SSD
- magic mouse, lighting cable & keyoard

2022 Mac Studio as close to your specs as possible sells for $4.8k

- M1 Ultra 20-core
- 128GB RAM
- 1TB SSD
- no magic mouse, lighting cable & keyoard

My assumption for the 2023 Mac Studio based on the M2 Max of the 2023 MBP 16"

- $4.8k
- M2 Ultra 24-core
- 128GB RAM (with option to 192GB)
- 1TB SSD
- no magic mouse, lighting cable & keyoard

Cost-wise a Mac Studio with GPUs you do not need is cheaper by $1.2k.

You could keep it for 5-6 years until the next process node shrink.

By comparison a Mac Pro with user replacebale RAM, SSD and PCIe expansion slot would start at least $6k. That's extra $2k over the $4k Mac Studio M2 Ultra

The Mac Studio siphoned most of the potential Mac Pro buyers who did not wan user replacebale RAM, SSD and PCIe expansion slot. It is also to cover the extra parts & extra shipping size of a Mac Pro.

If macOS is that bad why not move to Windows/Linux?

 

[Ploki]

I use Logic Pro.
I could switch to Studio One or Cubase, but I really don't want to.