Mac Studio & M1 Ultra

[ParkerJames]

For giggles. I opened a 300 mb file and the AM immediately showed a 3.4 mb usage and its stsyed the same when in the process of rendering Redshift of whole model….reference the above system.

 

[line-weight]

5 hours ago, Mark Aceto said:

@line-weightthat's interesting. So even though you were working with a file that's roughly have the size, it sounds like there was no RAM savings. Do you notice relative performance improvements or slow-downs with either option enabled or disabled?

 

The only other thing I was wondering about is that the timing of VW upgrades coincides with macOS upgrades. Did you upgrade to Monterey around the time you upgraded to v2022? Were you ever using v2021 on Monterey? Would be great to rule out macOS in this process of elimination game... 

 

If I remember right, I used VW2021 on Big Sur for a short time. I can't remember if I ever ran 2022 on Big Sur, or if I upgraded to Monterey first.

 

In any case, VW2021 on Monterey is what I'm using day-to-day at the moment. I have 2022 installed but have only tested stuff on it so far; I have not moved any "real" projects to 2022 yet.

 

NB the test I described above, was on VW2021 not 2022. I might try it on 2022 and see if similar, better or worse happens.

[Anders Blomberg]

I don't know if it would help but I'm running Clean my Mac and it has a little button for freeing up RAM. Maybe worth a try?

 

[line-weight]

9 hours ago, line-weight said:

Just did a quick test of what you suggested - I had "save viewport cache" and "save VGM graphics cache" both ticked, but unticked them, saved doc, closed VW and restarted.

 

This is in VW2021:

- File reduced from >2GB to 1.2GB

- Opening the file in VW immediately occupies 14.9GB of memory according to Activity Monitor

- By cycling through a few saved views, without editing anything, I could get it up to 21GB quite quickly

- closing the file left VW occupying 6GB of memory (no other files open except for an empty "Untitled 1" file).

 

Ok, a repeat of the above test but this time in VW2022. I opened the exact same file in 2022, which meant that on first opening it had to be converted. Once it had converted and opened, AM said VW was using 20GB of memory. I then saved it as a 2022 document, closed it and quit and re-started vectorworks. Then I opened that file fresh.

 

- File size still about 1.2GB

- Opening the file in VW immediately occupies 11.2GB of memory according to Activity Monitor

- By cycling through a few saved views, without editing anything, I could again get it up to 21GB quite quickly

- closing the file left VW occupying 1.7GB of memory (no other files open except for an empty "Untitled 1" file).

 

So - in fact on some measures it performed a little better than VW2021, which runs contrary to my subjective impression that VW2022 suffers more badly from this memory issue. But, this was just a quick test, so not the same as having the file open for several hours and working on it.

[line-weight]

Final test for this evening - same file, again with "save viewport cache" and "save VGM graphics cache" unticked.

 

This time, instead of cycling through (3d shaded perspective) saved views to see what happens with the memory, I looked at a sheet layer with several viewports each containing a fair bit of geometry (not section viewports, 3d perspective renderworks viewports, but displaying in wireframe because I haven't updated them since opening the file).

 

In VW2021, this prompts the memory to go up to about 33GB. Zooming and panning is sluggish but not completely unusable.

 

But in VW2022, the memory goes off up to 40GB and beyond, everything reacts incredibly slowly with a lot of beacballs and it's essentially unusable.

 

So this seems to be where 2022 is different, and worse. If the root cause is a "memory leak" problem in Monterey, then VW2022 seems to be more badly affected than VW2021, at least working in sheet layers.

[Mark Aceto]

On 3/11/2022 at 1:00 PM, Christiaan said:

Why should I get a Mac Studio for this price when I could get a base model Mac Pro for a similar price and upgrade it over time with cheaper 3rd party components? But what sort of upgradability is a future Apple Silicon Mac Pro going to have?

 

I decided to order a Studio because I expect the Mac Pro will cost 2k more for the same specs because "Pro". I'm not sure what would be user-servicable in a Mac Pro ARM since both the GPU and RAM are SOC. Hard drive? The SOC itself? Maybe they'll offer it with Threadripper and RTX 🤣 🤣. 🤣. It's an unpredictable wildcard at this point.

[Diamond]

Hello Christiaan,

 

On 3/12/2022 at 5:00 AM, Christiaan said:

The Mac Studio has made me curious about the Mac Pro now. Why should I get a Mac Studio for this price when I could get a base model Mac Pro for a similar price and upgrade it over time with cheaper 3rd party components? But what sort of upgradability is a future Apple Silicon Mac Pro going to have?

 

For the best guess of the Mac Pro, I would listen to the latest https://atp.fm/473 podcast.

 

As the M1 Ultra is the last chipset in the M1 family, the 'Jade 4C-Die' version that will likely be one of the offers available in the Mac Pro will be based on the M2 family of chips.

 

No one is quite sure how the Mac Pro system is going to deal with extra RAM (for those that want terabytes), or other I/O like PCI cards, and so on. But it makes sense that Apple will maintain their existing Intel enclosure, as they spent years developing it and would have had the Apple Silicon transition in mind. Obviously it would have a different motherboard and chip interconnects to reflect the speeds these chips can handle (i.e. Ultra-Fusion).

 

And regarding cheaper third party components, the Mac Pro is never going to be cheap to upgrade.

 

Also note that the Intel Mac mini is still available (while the Intel iMac was discontinued). I would not be surprised to see an Intel based version of the Mac mini in addition to the Mac Pro to maintain compatibility for specialist users.

 

Combined with the Studio Display, it is the replacement for the iMac, Mac Mini and in many cases, the Mac Pro. In short, unless you need external PCI cards, this is your machine. And if you want portable, then go for the MacBook Pro instead. For those that have workflows that need the full grunt of the M1 Ultra, then go for both. With the headless Mac Studio enclosure, update it when it is end of life.

 

Moving to Apple Silicon, what you will notice is the instantaneity and responsiveness of macOS, similar to iPhone or iPad. That is where you will find the most improvement. Hopefully Apple will continue to push their speed increases aggressively to break the Intel slowdown we have seen over the last decade. But I suspect the largest benefit is just transitioning to Apple Silicon.

 

To be fair, I am not sure Vectorworks is using all of that power of the M1 Ultra at the moment. IMO the more important figure is the RAM/VRAM. Since moving from a Intel MacBook Pro with 16GB RAM to an iMac with 40GB RAM, most of the crashes I was experiencing have gone away.

If you work on very large Vectorworks projects, 32GB would be minimum, with 64GB ideal, and 128GB just bragging rights. I am sure a 64GB M1 Max would be an amazing machine. If I was to spec one of these at the moment, this is the configuration I would choose as a solid mid-high tier Vectorworks machine…

 

Edited March 14, 2022 by Diamond

[Diamond]

@Christiaan Regarding your Mac Pro FOMO read this…

www.theverge.com/22974998/apple-xmac-myth-midrange-mid-tower-mac-studio

[Christiaan]

Brilliant, thanks @Diamond

[Mark Aceto]

13 hours ago, Diamond said:

For the best guess of the Mac Pro, I would listen to the latest https://atp.fm/473 podcast.

 

I miss Siracusa's annual macOS reviews (especially for Big Sur and Monterey).

[Mark Aceto]

On 3/10/2022 at 6:34 PM, jnr said:

Hate to do it but may follow your advice and do a full uninstall

 

I do it often for testing, and it typically takes about 10 minutes. I wouldn't recommend doing it often; only as a troubleshooting step or bi-annually (after SP3).

 

On 3/10/2022 at 6:34 PM, jnr said:

Apple will give me $1300 for the iMac Pro on trade in.

 

Apple trade in is great for phones and watches but you can get a lot more return selling it. It's never taken me longer than 2 days on Craigslist (ironically eBay fireballed me with scammers when I tried with my iMac Pro, so I've never gone back) in NY and LA. I also just sold the MBP M1 Max that I never took out of the box for the same price I paid. Mac's hold their value, especially during a chip shortage with long ship times.

 

On 3/10/2022 at 6:34 PM, jnr said:

And since I'm already swimming in the Big Sur pond, in the interim perhaps forced to upgrade to Monterey. Fingers crossed that 2022 issues with it were solved with latest 12 version.

 

The Studio I ordered delivers in May, so I'm dragging my feet until Monterey 12.3.1 or later. Presumably SP3.1 will be out around that time too. Good time to make the move on the homestretch. Blue skies until the next vicious annual release cycle starts all over again... 

[Christiaan]

15 hours ago, Diamond said:

I am sure a 64GB M1 Max would be an amazing machine. If I was to spec one of these at the moment, this is the configuration I would choose as a solid mid-high tier Vectorworks machine…

 

 

I have 128 GB of RAM in my iMac, which I routinely near max out. I know the M1 architecture means I probably won't need as much RAM as I have now in my iMac but I would still be reluctant to order anything less than the 20 core so that I can max out the RAM at 128 GB, given that there's no ability to add it later and I imagine my use of VW over the next few years will demand more use of RAM, not less.

 

Edited March 14, 2022 by Christiaan

[E|FA]

Looks like I'll be waiting for the next Mac Mini update, hopefully with more RAM available than the current M1.  

[Diamond]

1 hour ago, Christiaan said:

I have 128 GB of RAM in my iMac, which I routinely near max out. I know the M1 architecture means I probably won't need as much RAM as I have now in my iMac but I would still be reluctant to order anything less than the 20 core so that I can max out the RAM at 128 GB, given that there's no ability to add it later and I imagine my use of VW over the next few years will demand more use of RAM, not less.

Wow 😮! What are you doing to max out all of that RAM. The only thing that tends to push me that far over the edge are memory leaks relating to rendered viewports (section viewports mostly).

 

Yes, unfortunately that places you in the expensive M1 Ultra space. And remember, the reason they chips are so expensive is because Apple use this fab process of joining chips together and they become massive chips using a massive proportion of the chip wafer. On the plus side, mentioned by @Mark Aceto, they will hold their value, so I am sure you will able to sell it when the Apple Silicon Mac Pro comes out and you will have had a year’s usage.

 

And according to @Siracusa, in Apple Silicon, RAM is still RAM. And perhaps more so because it forms the dual role of VRAM as well.

[Mark Aceto]

Renderworks will use 64gb or more of RAM if it’s not constrained by the CPU. TM will use 64gb or more of RAM if it’s not constrained by the GPU. Google Chrome will use 128gb or RAM if one tab is open 😉 

 

I’m not sure how much VRAM/RAM Redshift is capable of using but the min requirement is 8gb VRAM.

Also keep in mind multiple external monitors hitting the GPU (unified memory), referencing multiple PDF’s in Preview, Mail and Safari memory leaks etc etc while we’re using VW at the same time.

[zoomer]

3 hours ago, E|FA said:

Looks like I'll be waiting for the next Mac Mini update, hopefully with more RAM available than the current M1.  

 

 

I hope that too but it may be likely that the M2 stays with 16 GB of memory.

 

I am aware that the M1 Max or Ultra are much faster in any task than my M1.

But if my low spec M1 would only allow 32GB of memory or more,

it may be slower than possible - but I could open and work on all of my

projects. But with only 16 GB of shared memory - I can't !

[M5d]

11 hours ago, E|FA said:

Looks like I'll be waiting for the next Mac Mini update, hopefully with more RAM available than the current M1.  

 

8 hours ago, zoomer said:

I hope that too but it may be likely that the M2 stays with 16 GB of memory.

 

I haven't seen anything mentioned on RAM for the M2, but it has already leaked / been suggested that when the M2 Mac Mini arrives, there will be an M2 Pro option. That should mean a BTO with at least 32GB available beneath the Mac Studio models. 

[Diamond]

2 hours ago, M5d said:

I haven't seen anything mentioned on RAM for the M2, but it has already leaked / been suggested that when the M2 Mac Mini arrives, there will be an M2 Pro option. That should mean a BTO with at least 32GB available beneath the Mac Studio models. 

 

Remember the lowest level Apple Silicon chips are meant for everything entry level. It is also clear in this chip architecture reset Apple is taking cues from the original Johnny Ive / Steve Jobs iMacs and iBooks. They were definitely entry level. And I can’t see an iPad with the possibility of having 32GB Memory, at least for the next couple of years.

 

Hopefully, Apple do bring a 32GB chip to the Mac Mini earlier than that. We can hope, but I wouldn’t expect it. 😏

[Christiaan]

15 hours ago, Diamond said:

Wow 😮! What are you doing to max out all of that RAM. The only thing that tends to push me that far over the edge are memory leaks relating to rendered viewports (section viewports mostly).

 

Medium size project. A 30 unit residential block. But it's a complicated footprint following a curved site and stepping down a sloping site, so there are many elevations and sections.

 

It really comes down to needing to keep quite a few Sheet Layers in the model file for certain things to work properly. All our elevations, for instance, so that the Data Tag for window numbering works (they don't work in workgroup reference files because they lose their connection to the window when you refresh the WGR).

 

We also worked with our section drawings in the model so that we can use Edit Section In-Place. Although we now have a nice workaround for this, which is the ability to copy and paste viewports across files. So now we have a separate file for our sections and copy paste them into the model file if we want to use Edit In-Place, then delete them and update the references.

 

I also recently worked on a 160 unit residential scheme that we put through planning. That was a memory hog. We had a lot of Renderworks Section Viewports, but split onto multiple files. That scheme also needed a lot of elevations and sections, so If we'd taken it into construction I think we would have struggled even with 128 GB RAM using our preferred workflow.

 

15 hours ago, Diamond said:

On the plus side, mentioned by @Mark Aceto, they will hold their value, so I am sure you will able to sell it when the Apple Silicon Mac Pro comes out and you will have had a year’s usage.

That's a good point 👍🏼

Edited March 15, 2022 by Christiaan

[M5d]

1 hour ago, Diamond said:

 

Remember the lowest level Apple Silicon chips are meant for everything entry level. It is also clear in this chip architecture reset Apple is taking cues from the original Johnny Ive / Steve Jobs iMacs and iBooks. They were definitely entry level. And I can’t see an iPad with the possibility of having 32GB Memory, at least for the next couple of years.

 

Hopefully, Apple do bring a 32GB chip to the Mac Mini earlier than that. We can hope, but I wouldn’t expect it. 😏

 

Yes, I’ve been waiting, patiently, for Apple to give us another iMac Pro level device. The Studio/Ultra fits the bill perfectly. Hopefully we’ll hear from NV on RAM expectations and the current problem in line-weight’s and Zeno’s posts soon. Based on what I have worked with, 64GB should be plenty, but if NV says there’s an issue or tangible benefit . . . ?

 

The Mac Mini rumours referred to above sound pretty solid, based on the A/Si tiers we already know about:

Mini = Mn & Mn Pro  |  Mac Studio = Mn Max & Mn Ultra  |  Mac Pro = Mn Ultra and Mn ??? 

 

Credit to Apple, it’s a very simple, understandable, lineup if this is how it turns out.

[line-weight]

More fun times in VW2021 today - and a new record of >50GB being used by VW on my 16GB M1 mini.

 

 

I don't really understand how it can need that much memory. In this case, I was looking at a single sheet layer (file opened looking at it, so it's not cached anything on other layers from this session), on which were 9 perspective view renderworks viewports. All un-rendered - so showing the model as wireframe.

 

A few JPG images on the sheet too - but their size is measured in 10s or 100s of MB only.

 

If VW needs 50GB of memory to show me 9 viewports, that suggests it needs around 5GB for each viewport. That's for a model the entire file size of which is <2GB.

 

One thing I noticed, once I updated the class visibilities for each of those VPs (so they each were viewing a smaller number of objects) then on next file opening, memory was sitting around 20GB and I was no longer getting constant beachballs.

 

My understanding of how memory works in these scenarios is no doubt naive but I just don't get how it can need that much.

[zoomer]

Hmmh,

that is VW 2021, non ARM native App.

I think very likely a memory leak or misusage.

But who is in charge, VW or Apples Rosetta or both.

 

I am not sure if VW 2022 is free of such memory issues.

But whatsoever,

I fear a possible solution would come for VW 2021 only.

[zeno]

3 hours ago, line-weight said:

More fun times in VW2021 today - and a new record of >50GB being used by VW on my 16GB M1 mini.

 

 

I don't really understand how it can need that much memory. In this case, I was looking at a single sheet layer (file opened looking at it, so it's not cached anything on other layers from this session), on which were 9 perspective view renderworks viewports. All un-rendered - so showing the model as wireframe.

 

A few JPG images on the sheet too - but their size is measured in 10s or 100s of MB only.

 

If VW needs 50GB of memory to show me 9 viewports, that suggests it needs around 5GB for each viewport. That's for a model the entire file size of which is <2GB.

 

One thing I noticed, once I updated the class visibilities for each of those VPs (so they each were viewing a smaller number of objects) then on next file opening, memory was sitting around 20GB and I was no longer getting constant beachballs.

 

My understanding of how memory works in these scenarios is no doubt naive but I just don't get how it can need that much.

 

Take this! 118 GB!!!

 

 

 

Edited March 21, 2022 by zeno

[line-weight]

2 hours ago, zoomer said:

Hmmh,

that is VW 2021, non ARM native App.

I think very likely a memory leak or misusage.

But who is in charge, VW or Apples Rosetta or both.

 

I am not sure if VW 2022 is free of such memory issues.

But whatsoever,

I fear a possible solution would come for VW 2021 only.

 

Same sheet layer, same drawing, in VW2022 got it up beyond 55GB and eventually the whole computer crashed and restarted. So if anything, worse in VW2022.

 

[gijsmans]

Our workflow is VW and Enscape. Forced to use PC but also Mac user since 1987 and Mac programmer. Mac Studio could kick the PC out if it runs enscape for Mac in the near future.