Mac Silicon OS and VW

[line-weight]

30 minutes ago, Tom W. said:

 

 contrary to what I said in the other thread, VW2023 actually performs better although the underlying problem is the same.

 

My impression is that some memory related things have been improved between 2022 and 2023, but not necessarily such that they are any better than they were in 2021. And some have got slightly worse between 2022 and 2023.

 

[mjm]

3 hours ago, line-weight said:

I reported on some of my memory-hogging experiments here.

 

My impression is that having more RAM might not really help (it simply would take a bit longer for it to fill up and everything collapse). I think that was based on some comments from other users, for example this one here by @zeno describing memory on a (I think) 64GB machine getting eaten up, and going off up to 140 or 190GB.

CONCUR

[mjm]

1 hour ago, Mark Aceto said:

 

Doubling the low end from 16 to 32 would make a big difference.

 

Adding 16 to the high end would not be noticeable for most users, myself included.

 

I feel like a broken record but:

• Look at the Intel MBP in my signature: 32 + 8 (40 total; not 32)
  • So 32 is really more like 24 + 8 if we're comparing ARM's to Intels
• I've never run out of memory with 128gb, and I'm usually sitting somewhere between 25 and 50% (32 to 64gb)

If 32gb shared memory is in one's budget, it's a no-brainer for 2023 and beyond.

 

If a power user is concerned about running out of memory, and has the budget, I can vouch for the Studio not disappointing. I complain a lot. I've never complained about running out of memory.

I got the M1 Max 32GB MBP. So sad I did NOT get the 64GB.

 

[Tom W.]

1 hour ago, E|FA said:

@Tom W. Is your internal drive an SSD?  I'm assuming an SSD would be faster at virtual memory swapping than HD or Fusion drives. 

 

Yes it is SSD. What is virtual memory swapping exactly + when does it occur? Thanks!

 

 

 

 

 

 

 

 

 

 

[Mark Aceto]

1 hour ago, mjm said:

CONCUR

 

What was the VW task in that screenshot?

 

BTW there are som other column options in Activity Monitor that might be useful:

 

 

 

Edited January 27, 2023 by Mark Aceto

[Mark Aceto]

1 hour ago, mjm said:

I got the M1 Max 32GB MBP. So sad I did NOT get the 64GB.

 

Tim Apple (in his best Don Draper voice): That's what the M2 is for!

[E|FA]

1 hour ago, Tom W. said:

What is virtual memory

To the best of my understanding, when the machine runs out of RAM, it uses disk space to write down some of the RAM data, creating greater "virtual" RAM capacity.  However, this is much slower due to the read/write and swapping time.  An SSD should do this faster than a spinning disk HD.  https://support.apple.com/en-ca/guide/mac-help/mh11852/mac

[zoomer]

12 hours ago, E|FA said:

To the best of my understanding, when the machine runs out of RAM, it uses disk space to write down some of the RAM data, creating greater "virtual" RAM capacity.  However, this is much slower due to the read/write and swapping time.  An SSD should do this faster than a spinning disk HD.  https://support.apple.com/en-ca/guide/mac-help/mh11852/mac

 

I think in the past it was that way.

Once RAM was full, everything on top was swapped to HDD/SSD,

no matter what kind of.

Or your model will either fit into your VRAM or if not will not work at all.

 

I think Apple ARM :

- does this more intelligently

- has very fast SSDs (in RAID 0 config, if you do not go with the lowest SSD sizes)

- has overall very fast memory bandwith

- has tile memory (if an App is optimized to make use of it)

 

So this is not true knowledge but it "feels" like Apple ARM swaps more intelligently.

According to your memory size and current need, Apple will use your memory

pretty spacious from the beginning.

A 16 GB M1 will use more memory that an 8 GB for the same load.

But when memory demand rises, Apple will intelligently start swapping already

less urgent or static data to SSD.

Like for your Safari background tabs, Mail, .... and I even think parts of VW's

memory consumption is pretty static and can be swapped with little impact.

Maybe apple just evaluates which data is often in use and which is just laying

around in memory.

 

Therefore I can work with VW files even showing more demand than my 16 GB

memory available. Although I have other Apps open and memory is even shared

between CPU and GPU.

Not as snappy anymore, (yellowish memory pressure), but still OK.

And maybe that would still work on the 8GB M1 Mini too.

That is also true for Photo and Video work with large files.

 

I think it only starts to really lag if a large VW file needs more memory

than physically available, in a single chunk or alternating chunks of

memory. This makes working unrealistic.

(But it doesn't work noticeably better on the unlimited PC either !)

Also true for music production where you can have only a certain number of

tracks, which you need to access all at a time, until memory and CPU cores

reach their limit and exit.

 

 

So I think it is a total natural use case for Apple ARM devices to work under

demands that use more memory than available.

So you do not need the ridiculous priced max memory upgrades for your

M2 Mac in the large majority of use cases.

(Editing your Symphonie with 360+ Tracks, gigantic Revit Imports, ....)

Only if you are sure that your use case will need all that memory at a time,

it makes sense and you will really profit from the memory upgrade.

Edited January 28, 2023 by zoomer

[Pat Stanford]

I split the M2 Render  Speed test into a new thread. It seemed more appropriate that leaving it here buried 19 pages deep into this thread. Also, that information will be likely of more general interest.

 

Please post any update you have on the speeds of your machines there.

 

Thanks,

Moderator Pat

[neal-2002]

wrong post - moved!

Edited February 7, 2023 by neal-2002

[Don Seidel]

On 1/27/2023 at 3:27 PM, E|FA said:

What is virtual memory

as above, simply the machine using the hard drive as a scratch pad when the limit of installed ram is hit.

-Very slow w/ a traditional spin drive. Read/Write vs chip-to-ram communication is drastically slow

-Much improved w/ SSD drive read/write speeds

-Improved dramatically w/ Nvme drive (essentially a ram stick used as HD)

-improved dramatically again w/ M-chips having installed ram on the processor chip itself

[Mark Aceto]

14 hours ago, Don Seidel said:

-improved dramatically again w/ M-chips having installed ram on the processor chip itself

 

Nitpicking here but RAM isn't installed on the processor chip itself. With the SoC (system on a chip) layout, the M2 processor die is interconnected to the memory sticks on the interface block as seen below. Long story short, the CPU and GPU cores are on the M2 chip. The RAM modules are next to it... 

 

The nice thing about this architecture is that the ARM SoC's are "expandable" (at least at the time of purchase for an exorbitant penalty). This is also why I'm so pissed at Mapple for not offering more memory in the Mini when they clearly could but choose not to (forcing their customers to pay even more for a MBP or Studio).

 

https://www.apple.com/macbook-pro-14-and-16/

 

[Don Seidel]

On 2/8/2023 at 9:23 PM, Lunar Waneshaft said:

Nitpicking here but RAM isn't installed on the processor chip itself.

yes nitpicking. The direct connection to ram is a first-time design, if I'm not mistaken. Point being M-chips can do VM ram like never before.

[mar schrammeyer]

Now using a m1 macbook pr 16” with studio display VW23 only and an older i5 mac mini for mail and the rest a bit posh with three screens ( no body wanted to buy the mac mini)

I used to have a lot of crahses if i asked for a quick 4 elevations render . Had to do one and then the other three weird.