Mac Studio & M1 Ultra

[Christiaan]

M1 Ultra, do I need it? Discuss.

Edited March 9, 2022 by Christiaan

[line-weight]

The Mac Studio seems like a significant development. Are apple looking to get some of their old customer base back, having somewhat abandoned the desktop "power user" for a while?

[Anders Blomberg]

I was watching the event and I must say that it was really cool to see Vectorworks appear in the presentation. I hope this means Apple and VW have a good relationship which would benefit us users. Apple's website also mentions performance increases for Vectorworks specifically and this is where I get kinda worried. The performance increases for other software are reported to be 3x or 4x but for VW it peaks at 1,6. I'm guessing this is because performance increases for processors these days seem to come mainly from more cores, and Vectorworks only use a single core in many tasks, as I understand it. So basically the base model M1 would not differ much from the highly specced Ultra in a lot of tasks. This is just my understanding, I'm sure others here know much more about this.

 

[line-weight]

Here are the specifics of what they tested

 

Quote

Testing conducted by Apple in February 2022 using pre‑production Mac Studio systems with Apple M1 Ultra, 20‑core CPU and 128GB of RAM, and pre‑production Mac Studio systems with Apple M1 Max, 10‑core CPU and 64GB of RAM, each configured with 8TB SSD, as well as production 3.2GHz 16‑core Intel Xeon W–based Mac Pro systems with 192GB of RAM and 4TB SSD, and production 3.6GHz 10-core Intel Core i9–based 27‑inch iMac systems with 128GB of RAM and 8TB SSD. Tested with Vectorworks 2022, using the Publish command to export sheet layers with Renderworks-rendered 3D viewports to a 7‑page PDF document. Performance tests are conducted using specific computer systems and reflect the approximate performance of Mac Studio, Mac Pro and iMac.

 

I have to say, most of the limitations I feel I face in VW at the moment are caused by VW rather than hardware. The biggest hardware bottleneck I find at the moment (using an M1 mac mini) is memory, as the 16GB gets filled up when I do Renderworks renders. But as far as I can work out, this doesn't need to happen and seems to be caused by some kind of memory leak in VW.

 

For sure it would be nice for those renders to complete in half the time, but what would be nicer would be for VW not to crash due to aforementioned memory problems when I try and do several renders in a batch.

Edited March 9, 2022 by line-weight

[fabrica]

5 hours ago, Christiaan said:

M1 Ultra, do I need it? Discuss.

do you need it : no

do you want it : yes!

[Guest SteveJ]

Hello all interested in Native Apple silicon!

 

The publish command is certainly a lower bound for performance improvements. The multi-core and GPU/Metal/DirectX performance wins for Vectorworks currently occur with the everyday, every minute, every second workflows of changing views and navigating your complex design. Switching and redrawing in Top/Plan, switching to and navigating 3D views, working with large files with large number of classes and more. I think/hope those using Apple silicon already can remind us of the these wins.

 

Concerning 16GB and the frustrations updating multiple viewports in batch, that is one of the biggest offerings of ultra. Updating multiple viewports with complicated section cuts does push beyond 16GB. With 32, 64, and now 128GB, that limit is lifted. 

 

So four points to add at this point:

• The memory limitation of 16GB has a new upper bound!
• The real multi-core and GPU benefits of native Apple silicon are not well celebrated with the publish command - those of you already using Apple silicon might remind us of the experience you see. (16GB batch update limit not withstanding)
• This discussion thread will collect more benchmarks as we and others get hold of a shipping Mac Studio. We will be able to report on workflows other than the publish command.
• Apple silicon is great technology. And, don’t forget, much of what we do is cross platform. Ultra and UltraFusion are quite compelling. Here at VW, we are constantly working to improve Vectorworks to take advantage of all this new technology. We hear and understand your concerns and are laser focused to continue to improve your workflows and experiences.

[J. Miller]

Will the ultra speed things up at all since its 2 chips acting as one (mac voodoo magic)

My trash can is getting buggy and need to be replaced soon.

is the studio monitor worth it or is there a suggetstion for (2) 4k monitors. I would like to have good speakers and a web cam.

Maybe (1) studio 27" and a 2nd basic model that will line up with the studio (4k) as and screen for resources, oip, navigation.

I could really use some help here. VW entertainment user with lots of renders and a pile of 3D modeling.

Thanks,

Jeff

 

[J. Miller]

So Steve, Is that a thumbs up to invest in a studio M1 ultra (32Gb ram or 64 gb ram)

Jeff

[Guest SteveJ]

7 minutes ago, J. Miller said:

Will the ultra speed things up at all since its 2 chips acting as one (mac voodoo magic)

Recall the single M1Max is quite fast - some might chime in here. We will get our hands on ultra asap - and share what we find! I would not expect it to be slower than M1Max. and you have the 128GB option. If you currently use 128GB for your projects, this is very nice news!

[Guest SteveJ]

2 minutes ago, J. Miller said:

So Steve, Is that a thumbs up to invest in a studio M1 ultra (32Gb ram or 64 gb ram)

Jeff

Has anyone here been disappointed with Apple Silicon? (16GB limit not withstanding)

[markdd]

I have the M1 Max having migrated from a MacPro Trashcan. The difference in performance is night vs Day. Absolutely delighted so far.

Edited March 9, 2022 by markdd

[Mark Aceto]

On 3/9/2022 at 2:19 AM, Anders Blomberg said:

So basically the base model M1 would not differ much from the highly specced Ultra in a lot of tasks.

 

Yes, and it sounds like the claim that M1 Ultra is "60% faster than the 28-core Mac Pro" was in regards to single core (multi-core is closer to 20% faster). That's a dirty marketing trick considering that more cores on an Intel chip lowers the single core clock speed. All of the M1 family have pretty much the same single core performance.

 

That said, I'm overjoyed that the Ultra is benchmarking just shy of the Threadripper 3990x (for under $6k !), so I'll be ordering one soon to make use of Renderworks and Redshift (with 128gb of unified memory available for VRAM).

 

Edited March 10, 2022 by Mark Aceto

[Mark Aceto]

44 minutes ago, J. Miller said:

is the studio monitor worth it

 

Absolutely not. There are so many excellent monitors that are 32" and above from BenQ, Dell, LG... and the funny thing is that they all use the same panels.

However, the nice thing is that now users have a great headless option, so we can get whatever monitor we want.

[Mark Aceto]

46 minutes ago, J. Miller said:

(32Gb ram or 64 gb ram)

 

Keep in mind the other apps you might be using (possibly at the same time). Twinmotion will use 64gb or more. 32gb is plenty for VW by itself but you may have 3 Chrome tabs open at the same time 😉

[Mark Aceto]

4 minutes ago, Mark Aceto said:

 

Absolutely not. There are so many excellent monitors that are 32" and above from BenQ, Dell, LG... and the funny thing is that they all use the same panels.

However, the nice thing is that now users have a great headless option, so we can get whatever monitor we want.

 

The polishing cloth is sold separately for $19 because... courage... environment... privacy 🤣🤣🤣

[line-weight]

54 minutes ago, SteveJ said:

Concerning 16GB and the frustrations updating multiple viewports in batch, that is one of the biggest offerings of ultra. Updating multiple viewports with complicated section cuts does push beyond 16GB.

The thing is, the same operation (render multiple renderworks viewports on the same sheet layer) that I could do in VW2018 on my old mac pro, with 16GB memory, now causes a crash in VW2021 on my newer M1 with 16GB memory.

 

There is something else that's going on that looks a lot like some kind of memory leak. This has been discussed elsewhere.

 

We shouldn't have to spend on more expensive hardware to compensate for something that is wrong inside VW.

[Mark Aceto]

Keep in mind that the M1's have unified memory.

 

We're used to buying a computer with 32gb of RAM + 8gb of VRAM but the M1 just lists a single number to be shared by both. That's a blessing if you have 128gb to be shared but it's a curse if you have 32gb or less.

[line-weight]

4 minutes ago, Mark Aceto said:

Keep in mind that the M1's have unified memory.

 

We're used to buying a computer with 32gb of RAM + 8gb of VRAM but the M1 just lists a single number to be shared by both. That's a blessing if you have 128gb to be shared but it's a curse if you have 32gb or less.

 

Sure. But my old mac had 16GB RAM + 3GB VRAM.

My M1 has 16GB unified memory, so it's a little less but I'm not sure it explains the problems I describe.

There's discussion of it here

https://forum.vectorworks.net/index.php?/topic/72941-mac-silicon-os-and-vw/page/16/#comment-411469

 

[Mark Aceto]

1 minute ago, line-weight said:

 

Sure. But my old mac had 16GB RAM + 3GB VRAM.

My M1 has 16GB unified memory, so it's a little less but I'm not sure it explains the problems I describe.

There's discussion of it here

https://forum.vectorworks.net/index.php?/topic/72941-mac-silicon-os-and-vw/page/16/#comment-411469

 

 

Not sure what's to discuss. 16gb is not enough. 32 is minimal. 48 is good. 64 is great. 128 is excellent.

 

Also Redshift requires 8gb VRAM in VW, so there's that.

[line-weight]

1 hour ago, SteveJ said:

The multi-core and GPU/Metal/DirectX performance wins for Vectorworks currently occur with the everyday, every minute, every second workflows of changing views and navigating your complex design. Switching and redrawing in Top/Plan, switching to and navigating 3D views, working with large files with large number of classes and more. I think/hope those using Apple silicon already can remind us of the these wins.

 

Sorry for another negative post, but -

 

I'm using Apple silicon already and spend much of my time navigating and working in "shaded" 3d views. Moving to M1 did not result in a greatly improved experience in this regard. For the most part it continues to be "fine" just as it was on my older non M1 mac, but with the exception of a highly distracting problem that is significantly worsened in VW2022, which is described in this thread:

https://forum.vectorworks.net/index.php?/topic/87645-vw2022-shaded-view-flickering-hatches-patterns/

I'm hopeful it might be fixed in SP3. That thread has gone dead in terms of response from VW. But it's another example of the user-experience bottleneck being to do with VW problems rather than the hardware. There's not much point having hardware that can render fluid, detailed motion around a complex model if the view is then trashed by flickering zigzag lines everywhere.

 

 

[Mark Aceto]

Another thing I like about the Studio is that I'm hardly ever onsite anymore. 95% of my time is spent working at a desk with a 4k monitor. Spec for spec, the Studio is a better value (and less thermally constrained) than a laptop or all-in-one for the right person.

[Mark Aceto]

1 hour ago, Mark Aceto said:

 

Absolutely not. There are so many excellent monitors that are 32" and above from BenQ, Dell, LG... and the funny thing is that they all use the same panels.

However, the nice thing is that now users have a great headless option, so we can get whatever monitor we want.

 

And the same display is not compatible with the other options: VESA, height-adjustable, non-height-adjustable. And for some reason, also not compatible with Macs made before 2016? Is it April Fool's already?

[J. Miller]

Mark,

Can you tell me what monitors you are using. I also sit at the desk full time now-a-days.

speakers - would be nice , video cam - yes, 4k - yes, 27" - min width.

I am planning on getting (2)

Jeff

 

 

[Sky]

The Apple Studio Display is 5K in a 27" package.

Does that mean the VW UI elements would get smaller?

Or are those dynamically resized depending on the monitor?

[Mark Aceto]

Ugh, early Cinebench scores not looking good... 

 

https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r23_multi_core-16

 

 

Much further down the list, are the M1 Max chipsets, scaling predictably at 1/2 the Ultra scores:

 

 

 

So, it stands to reason, that the imminent Mac Pro with 4x Max (double Ultra) would theoretically scored 48,000 (not even close to Threadripper).

 

Edited March 9, 2022 by Mark Aceto