AFAIK, Tensorflow does not officially support AMD GPUs on Intel Macs. Nvidia and Google TPUs are listed as supported, but no AMD. :-< https://www.tensorflow.org/install.
Might be a ROCm port somewhere that can be added with a TF source build.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
8GB RAM is a disappointment?
- Thread starter MacBH928
- Start date
- Sort by reaction score
Might be a ROCm port somewhere that can be added with a TF source build.
i'm not going to read all 36 pages of this thread, but i have read some pages at a random. i'm a writer by trade. i mainly use microsoft apps, web apps, stock apps like itunes, notes, and fantastical. the most taxing thing i can imagine myself doing is fooling around in movie and iphoto. would 16gb be overkill? would like to hang onto this machine for 3-4 years...at which point my youngest child will be in high school and then i could pass it down to her.
currently have the entry-level 2018 MBA with 128GB. looking forward to having more storage and getting away from the butterfly keyboard. plus i locked in a resale price at SELLMYMAC for more than 50% of the original purchase price.
currently have the entry-level 2018 MBA with 128GB. looking forward to having more storage and getting away from the butterfly keyboard. plus i locked in a resale price at SELLMYMAC for more than 50% of the original purchase price.
I have completely different results than you are reporting...
Attachments
I tried it again, and it shot up to more than 3GB of RAM. I then exited cleared history and tried again. Now I am a 300-500MB RAM while playing, like I said something in Safari is leaking memory.
It may not be Safari itself, but a Safari extension that is causing the behavior you are describing. Right now, Safari is using less than 75MB RAM on my MBP, and that's while downloading the latest XCode RC from the developer site.
You should be great with 8GB RAM. The only reason you should even consider 16 is if you anticipate your child doing memory-intensive tasks if and when you hand it down.
You need to use Apple-provided backend:
Apple's ML Compute Framework Accelerates TensorFlow Training
As part of the recent macOS Big Sur release, Apple has included the ML Compute framework. ML Compute provides optimized mathematical libraries to improve training on CPU and GPU on both Intel and M1-based Macs, with up to a 7x improvement in training times using the TensorFlow deep-learning library.
GitHub - apple/tensorflow_macos: TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework.
TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework. - GitHub - apple/tensorflow_macos: TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework.
github.com
They have not yet been merged in the official TensorFlow
Nice. Timing is strange through. I wonder if it was tied to M1 Macs release.You need to use Apple-provided backend:
Apple's ML Compute Framework Accelerates TensorFlow Training
As part of the recent macOS Big Sur release, Apple has included the ML Compute framework. ML Compute provides optimized mathematical libraries to improve training on CPU and GPU on both Intel and M1-based Macs, with up to a 7x improvement in training times using the TensorFlow deep-learning library.www.infoq.com
GitHub - apple/tensorflow_macos: TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework.
TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework. - GitHub - apple/tensorflow_macos: TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework.
They have not yet been merged in the official TensorFlow
M1 performance looks great. My MacBook Air M1 will arrive today. I will install Apple's TF release and see how it runs on some models.
Well, it uses the new tensor compute framework that has been introduced in Big Sur, so I suppose it makes sense...
Please report with results! I am not using TF, so I can’t test it unfortunately
I have the 8gb ram and like I expected it serves my needs well.
The swapping is fast and unnoticeable. This is the first time in a while where I really did not have to police what apps are open or not, just based on the number of Apps I have on the dock compared to my previous intel MacBooks.
I saw this choked though on my workflow where I opened 10 Facebook Ads and 10 Google Ads tabs launching at the same time. Those are very heavy websites and one tab consumes 1.5gb in memory. It still worked out well but it showed where the limitation is.
There are indeed stress tests that will not choke this machine, but I have a specific use case where it actually did, and this is not a stress test since I was hoping to have more than 10 tabs working on these heavy web apps.
I would like to blame it on Facebook and Google as those web apps (for creating ads) are very heavy, and the amount of JS + network latency makes me think a higher RAM is needed. Not sure how UMA handles heavy web pages where the majority of data needs to be fetched online.
Everything else works really well.
Will observe Memory pressure.
The swapping is fast and unnoticeable. This is the first time in a while where I really did not have to police what apps are open or not, just based on the number of Apps I have on the dock compared to my previous intel MacBooks.
I saw this choked though on my workflow where I opened 10 Facebook Ads and 10 Google Ads tabs launching at the same time. Those are very heavy websites and one tab consumes 1.5gb in memory. It still worked out well but it showed where the limitation is.
There are indeed stress tests that will not choke this machine, but I have a specific use case where it actually did, and this is not a stress test since I was hoping to have more than 10 tabs working on these heavy web apps.
I would like to blame it on Facebook and Google as those web apps (for creating ads) are very heavy, and the amount of JS + network latency makes me think a higher RAM is needed. Not sure how UMA handles heavy web pages where the majority of data needs to be fetched online.
Everything else works really well.
Will observe Memory pressure.
Last edited:
Okay I got it to choke with 20 Google Ads tabs open at the same time.
This would/should be a usual case for me, though Google ads does suck a lot of memory.
This would/should be a usual case for me, though Google ads does suck a lot of memory.
@dingclancy23 Which browser is that? Not Safari I assume (I don't think Safari has a renderer process). That memory usage is horrendous. Half a gig of auxiliary RAM to render a website?
This is Brave. But let me try in Safari now.
Google Ads still got it to choke.
The difference though is that Safari seems to be able to handle it better, and once I quit the tabs, the lag and slowness disappeared instantly. And there were no beachballs.
In this case of course, 16gb ram helps but that just means you can run 10 more of these tabs, which is a waste. Google is just not a good software for RAM.
If your work is heavy on web apps, I am not sure what magic the UMA can do if memory leaks happen, especially since it is so easy to open a new tab and a new instance of a heavy web app. That is just the nature of the beast with working with web apps. Just have to police yourself.
Also Google is not a responsible citizen of RAM in their apps. Now that they have overhead with M1 chips, I expect them to somehow take full advantage of it.
I just purchased a MacBook Air (open-box) from Best Buy. since it was not a customized order I was worried (and still am) about the 8Gb Ram. I am using a 16Gb MacBook Pro 15inch from 2018. On that computer, I usually reach around 12gs of ram while I use Outlook, Teams, Onenote, onedrive, VSCODE, Macvim, Preview, vnc, Work and excel on my work computer.
So far, I have been using the same without the use of VNC. and so far, it's been pretty good. The swap memory is at 2.60gigs which annoys me slightly because I don't like my SSDs to be cycled if they don't need but I don't feel any sluggishness when doing tasks which is great.
So far, I have been using the same without the use of VNC. and so far, it's been pretty good. The swap memory is at 2.60gigs which annoys me slightly because I don't like my SSDs to be cycled if they don't need but I don't feel any sluggishness when doing tasks which is great.
Is that because IOS isn’t a full MacOS operating system running large full software suites?
I remember when Steve Jobs got on stage & said his iPhone was running the complete full MacOS with “real” full apps, not those baby apps that windows run. Lmao
Steve sure knew how to gas light ya then?
I bought the 1st iPhone & to my surprise I couldn’t install it run Final Cut Pro nor ProTools full real apps!
if i had a 16gb notebook, would people like my photos i post here better?
You may not like the comparison, but I am here to tell you that I have the M1 Air with only 8 GB of RAM and it rocks through what I need it to do easily--far better than my top of the line 2015 MacBook Pro 15" with an i7 and 16 GB of RAM. So...I don't know what to tell you. Perhaps the unified architecture is that good.
Actually I feel the opposite, the RAM compression on the M1 gets higher more quickly than on my 2014 Mac mini. But I don't notice because it swaps a lot. And I mean A LOT (on my Mac mini it's usually zero, only uses swap memory when the machine is very stressed, on the M1 it easily jumps into 3-4GB of Swap memory) So the SSD is fast enough for you to not notice it, but you're possibly wearing your SSD if you're using a Mac with less RAM that you need, even if you don't notice a performance drop.
I would usually recommend 16GB unless you're going to do office and web browsing with few tabs. Or if you're gonna sell that Mac in 2-3 years, because SSDs are not replaceable.
ram is ram nothing change about it . Yeah more burst speed ddr4x but 8 gb still 8gb whatever plarform .
Unless we can control "services" then we can reduce the ram usage . Keep simple no cluttered desktop, no startup and limit spotlight and manual sync backup if possible. More ram is okay if sell in the shop which are the most problem for people like me ( i dont like to purchase online )
So although I have a 16GB M1 Mac Mini I must say that I find the M1 much more efficient in using RAM compared to the 16GB i3 2018 Mac Mini that I also have. It must have to do with the unified architecture, sharing video memory, it's almost like using an ipad just running macos instead of ios.
I think 8 or 16GB are perfectly fine, much better thant 8GB/16GB on an intel platform. 32GB RAM would have been recommended for people using virtual machines in my opinion and since there isn't any windows support on the M1...
I think 8 or 16GB are perfectly fine, much better thant 8GB/16GB on an intel platform. 32GB RAM would have been recommended for people using virtual machines in my opinion and since there isn't any windows support on the M1...
I'm running Windows in a VM on a Air M1. I have tried other settings but settled on using 8GB of RAM and 4 cores. So far it runs great with the apps I have used. So, yes, if you plan to use VMs probably better to get a 16GB system.
At work we have the 2020 i7 quad core Intel Macbook Pro w/ 16GB RAM.
Same as my 2019 16" MBP - both get super hot, gets laggy and battery drains in 3 hours using Zoom for remote learning.
Now the M1 base with 8GB running the same apps runs: super cold, never slows down, and battery lasts all day long.
Same as my 2019 16" MBP - both get super hot, gets laggy and battery drains in 3 hours using Zoom for remote learning.
Now the M1 base with 8GB running the same apps runs: super cold, never slows down, and battery lasts all day long.
Apple will either need to offer more RAM like 32gb or offer a user replaceable SSD