Regretting my Mac mini and Mac studio purchase, 1 month later

Good for you!

No harm in trying :wink:

That comparison is a bit lacking. Yes you have gen 4 lanes, but to make use of that sweet bandwidth you need cards that operate at that very spec. If you have capture cards they are gen3 so you’re essentially wasting bandwidth and losing lanes.

EDIT: What I meant to say is that you can’t make a PCIe gen 3 card run at X16 with 8 PCIe gen 4 lanes, which theoretically have the same bandwidth.

Just what kind of work are you doing that you feel like a TR 3945WX is limiting your ability to get things done? And why did you think a Mac would be a good replacement, when you are using 2 compute GPUs and capture cards?

I’m assuming you didn’t have the time to try Asahi, but even then, you probably want some GPU acceleration for your workload. Lastly, while I think the M1 Macs are very performant, I don’t know I would want to use one, only because I have heard some bad things about the SSDs getting data corrupted, although that can be avoided by saving files to a ZFS NFS share, but still requires you to have a secondary machine.

Have to agree with others that it sounds like you just have a large amount of money burning a hole in your pocket. As @wendell would put it: first need to identify the problem you’re trying to solve with a new build before even looking at parts lists

3 Likes

i mean yeah sadly thats true, just turned 20 and have a job in Big Tech so i usually get given cool hardware or get good deals on things where im happy to just go out and buy new things, as money is something quite new to me, as grew up kinda poor and now on the google paycheck i can just spend a crapton a month and regret it later down the road…

1 Like

I kinda get it; I definitely spend kinda loosely on tech stuff now that I make good money (finally broke 6 figures by 29 wooo!) but I at least make sure it’s stuff I have a use for or definitely adds value to my life.

Like, I think my PC is in a good spot so i got into custom keyboards as my next project. I could upgrade my 3090 and 5950x to a 3090ti and 12900k or whatever comes out next, but it’s not like I’ve felt it holding me back

Edit:
I highly suggest just investing the money into stocks and forgetting it exists. Let it work for you to help with your retirement plan or something

* laughing with 128 PCIe lanes and 8 memory channels * :slight_smile:

4 Likes

Exactly!
Workstation needs extend beyond compute.

The AM5 stuff uses ddr 5 though.
If you actually need beyond AM5, build an EPYC workstation.

Good for you!

This is all part of living and learning the lessons of life. Please don’t regret you decision too much. I remember my early twenties and I can tell you I wasted money on much sillier things than some computer hardware.

1 Like

You might be better served to pick up a 5950X instead of the 5800X3D. They are coming down in price with the new chips being released and they are still absolute monsters. Even though it may seem silly to buy a socket and board design that is nearing end of life, the cost to performance ratio makes it hard to pass up.

@aj0413 Is that a sandwich or a toaster stroodle?


M1 support is improving but it is a relatively novel architecture and the conventional wisdom when something like that comes around is wait a generation (or two) before dipping your toes in the water.

I’m considering a Mac Studio (my work is distinctly within the macOS ecosystem and I’ve grown to love the mix between a UNIX-like OS and the ability to run proprietary software without any hitches, think best of Linux and Windows for me) but there are three things holding me back.

  • macOS Monterey is still half baked and I don’t do half baked. Memory leaks, random crashes and (for device restoration) a new way of doing things that is slowly unraveling itself, I’ll wait until it’s fully unraveled before risking my live work on it.

  • Apple Silicon is a energy-efficient performant platform but it’s a new platform and its benefits are realized when you use it natively (instead of over Rosetta) and it takes time, a lot of time for software libraries to have them and their optimizations ported to Aarch64 (just a reminder, companies like Adobe relied on the Intel Math Kernel Library for performance improvements on x86_64 systems). Not everything I need has Aarch64 native versions that have been battle tested. (Initial releases will always have hiccups, ask me how I know!)

  • Also, I don’t like how a good amount of the architecture is still relatively unknown. One bad thing about the ARM world is that isn’t as standardized as how things are in x86_64 land and everyone is free to implement things however they damn well please and then give that the highest level of opacity. So, binary blobs and undocumented routines galore! Developing on ARM64 is no trivial task.

    The fine folks at Asahi Linux are doing a great job reverse engineering whatever is necessary for Linux support and have even discovered erratum (nothing new, x86_64 chips also have erratum but they’re at least well documented, see here for one discovered by the Asahi team), so the way I see is, there is light at the end of the tunnel but we’re not quite there yet.

TL;DR: If you’re a professional who needs a reliable system, period. Apple Silicon isn’t quite there yet, it might have the compute but stability and flexibility (which often tie in with ecosystem maturity) reign supreme.

1 Like

In my scientific computing use case, the difference between Zen 2 and Zen 3 is more than neglible.

Just the clock speed improvement alone is significant besides the IPC improvement and doubled FP64 pipelines.

1 Like

Why did you buy this stuff when you had a really nice threadripper? Seems like you had everything you needed but were sucked in by marketing.

1 Like

being young and stupid mainly, just turned 20 and got a job at a top tech company, the threadripper is nice but I always want more and better if I can get it and wanted to give M1 Ultra a shot.

1 Like

Ah, well at least you have the self-awareness to recognize this. Somewhere an Apple executive will read this post and find themselves strangely aroused.

That’s alright. You live, you learn. What’s important is, that you gave it a shot. As Irate said, you at least have the self-awareness to recognize it.
But my advice would be to save your pennies for a while and use the company’s resources to play around as much as you can.
Oh, and buyer’s remorse is always there. Whether you can afford a thing or not. Sometimes you just justify it and keep whatever you’ve bought, sometimes you return it.

1 Like

Hi OP.
Did you ever run the OpenFOAM benchmark on your Studio ? I’d love to know how it compared to other hardware.

Computers are generally bandwidth constrained when running OpenFOAM. Not core/corespeed constrained. The Studios (M1 Max and M1 Ultra) are supposed to have massive memory bandwidth. They seem like a perfect fit for running OpenFOAM.

As far as using Mac OS, I have to agree with you. But Asahi now supposedly supports the Studio devices. Would Asahi Linux + Mac Studio make the ultimate (affordable) OpenFOAM processing station ?

1 Like

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.