What's with the focus on insane amounts of RAM for these new apple silicon chips? Is there some performance benefit specific to macs that would require 32GB+ of ram? Who on earth would need that in a desktop/laptop?
For tech work computers, I'd say 32GB is the minimum one should have these days. My partner got a 16gb RAM laptop from her new company as a data analyst and she runs into trouble with it at least a couple times a week when she's got a lot of processes running.
And for multimedia work it matters often more, directly correlating with responsiveness of UI when editing multilayered tens-of-megapixel images or, worse, hi-res videos with tons of effects. I’m not familiar with 3D, but it’s probably even worse there, with less-well-packed data structures and constant GPU utilization.
I assume that’s for professional workstations? There were complaints about the „low“ maximum RAM with previous Apple Silicon generations if I remember correctly.
Extra RAM is always cool. I'm at a job now that has a shared dev environment powered by Docker. Our Monolith runs in about a dozen different containers. I have a 16GB M1 Pro, and performance is _okay_ giving it half of my resources. Had I known how this performs, I would have pushed for a 32GB machine. As is, if we continue to grow, things could get slow in another year or two.
Could we optimize the local stack more to use less resources? Probably, but it probably isn't worth our time. Until then, more RAM is good. There are plenty of users out there that have more needs than me. Giving them a solution to their problems is a step in the right direction.
Compiling and linking large projects can easily eat 4gb per core, so if there are 8+cores you need more than 32gb of ram.
I don’t think you need 64gb, my experience (on a 32gb machine) makes me think you need maybe 5gb per core to be safe from swapping, but presumably doubling is easier/faster than increasing in non 2^^x amounts.
I'm doing a GIS project[1] for fun to find out where I want to move in the next year. I have to constantly monitor RAM usage and save before running some operations because there's a chance OOM killer makes a fatal decision to kill the program. This is on a machine with 32GB and I could easily find myself using more.
1. Using QGIS. It's the first real GIS project I've done. I'm using to do this sort of work in code, so it's been a fun learning experience too.
For sure! And yes parts of it involve raster data. It doesn't help that I have very little control over the raster internals.
My project is limited to NY,OR,WA, but this clearly has a potential sparse raster problem. If I was doing it in code, I'd reach for internally tiled cloud-optimized geotiffs with low compression and I'd be pretty sure that all the empty tiles would get compressed down to nothing.
With QGIS, I can cross my fingers and hope it's doing something relatively sane with GDAL in the background.
Unified memory using high bandwidth buses to support both GPU and CPU and the rest.
I believe Apple uses very specific LPDDR5 packages then the pro and max have 2x and 4x buses so you need that many packages. This is why the M1 comes in 8GB ad 16GB, M1 Pro 16GB and 32GB and the Max is 32GB and 64GB.
The current M2 is 8, 16 and 24GB, so naturally a top spec M2 Max would have 4x 24GB so you get 96GB assuming they stick with a 512 bit bus.
15 comments
[ 23.1 ms ] story [ 1407 ms ] threadBeyond that though, 64gb+ of RAM makes for a great dev machine. :)
Could we optimize the local stack more to use less resources? Probably, but it probably isn't worth our time. Until then, more RAM is good. There are plenty of users out there that have more needs than me. Giving them a solution to their problems is a step in the right direction.
I don’t think you need 64gb, my experience (on a 32gb machine) makes me think you need maybe 5gb per core to be safe from swapping, but presumably doubling is easier/faster than increasing in non 2^^x amounts.
But I have to say, 12 CPU cores and 96GB of memory is a fascinatingly odd pairing. Is the base spec going to have a 128GB hard drive?
1. Using QGIS. It's the first real GIS project I've done. I'm using to do this sort of work in code, so it's been a fun learning experience too.
My project is limited to NY,OR,WA, but this clearly has a potential sparse raster problem. If I was doing it in code, I'd reach for internally tiled cloud-optimized geotiffs with low compression and I'd be pretty sure that all the empty tiles would get compressed down to nothing.
With QGIS, I can cross my fingers and hope it's doing something relatively sane with GDAL in the background.
I believe Apple uses very specific LPDDR5 packages then the pro and max have 2x and 4x buses so you need that many packages. This is why the M1 comes in 8GB ad 16GB, M1 Pro 16GB and 32GB and the Max is 32GB and 64GB.
The current M2 is 8, 16 and 24GB, so naturally a top spec M2 Max would have 4x 24GB so you get 96GB assuming they stick with a 512 bit bus.