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.
16GB m1 pro is hot and fan is running for 75% of my normal workload,battery last only 10 hours
- Thread starter AlexOnBoard
- Start date
- Sort by reaction score
I've had more problems with chrome than anything else (and finally starting to move away from it) - and I don't mean just ram usage, but high processor use and the kill/wait dialogue coming up. Glancing occasionally at memory usage, firefox takes a fair amount of memory but doesnt' choke or run out of control.
I've been totally blown away by the speed this happening at. I am really, really hoping this excitement transfers over to Microsoft, and it convinces them to give the nod to changing their licensing model with Windows for ARM.
I know what you mean! Chrome is a weird app on Macs. It has a bazillion strange processes, sometimes it runs what appears to be totally random CPU processes, the way it uses RAM is odd, and it seems like it is the fastest browser when you have free RAM and by far the slowest when it has to start caching to the SSD. I very much have a love and hate relationship with Chrome.
That there's no good reason to think the limit is arbitrary, and some reason to think it isn't, makes the feeling that it is invalid. People may wish there was more RAM available, nothing invalid about that, but there's not any ground I can see for a sense of entitlement about it, which is implicit in the complaints about these things.
"Possible" is a slippery term. There are competing values to be weighed, mainly cost in this case. It's possible to get 64GB with the 16", but I don't most people should.
Oh, I don't think so. Just discovering that their past usage won't fit their new equipment in every respect. Looks like they've narrowed down the main issue to one particular program.
I don't see any reason that this would be a physical limit - more that they decided to standardise on basically two configurations because volume/scale and meets needs of a very large percentage of users that buy the entry-level machines.
There might be a chip/SoC design limitation that means redoing too much work to make it worthwhile at this stage - i.e. M1 derives too closely from the iphone/ipad chipsets that were designed to max out at 16gb. And that since they know they have the next chip/SoC coming out in x months, priority was put on simplifying manufacturing/distribution for now and no more work put into variations of this particular chipset.
But I wouldn't rule out the possibility they'll come out with an M1 package some months out that has more memory, either. If it isn't an inherent design limit but just an SoC mod, maybe demand will be high enough they decide it's worthwhile.
My personal guess is they won't, even if they physically could: they'll save this for the higher-end models where lots more profit. There's a fair bit of overlap between users who believe they need 32gb or more and also believe they need the ability to run four monitors at a time. Apple likes buyers having to choose between powerful low-end with a few identified limitations, and paying a thousand dollars more for their dream machine. Readers of threads like this are the perfect target for those expensive dream machines.
yes lets hope so...there are already rumours...but with microsoft you never know
2021 is shaping up to be a big year for Windows 10
Windows 10X, big UI and feature updates for desktop, and a new app streaming service are on the cards for 2021.
www.windowscentral.com
Physical limit in the sense that you can't fit than two modules on the package and making the package larger just to accommodate the few % of users who need more than 16GB doesn't make sense economically. But I definitely agree with you reasoning.
Right? My mailbox is almost exploding with GitHub notification of people submitting patches to homebrew and other projects. It's really cool to see this much enthusiasm
yes lets hope so...there are already rumours...but with microsoft you never know
2021 is shaping up to be a big year for Windows 10
Windows 10X, big UI and feature updates for desktop, and a new app streaming service are on the cards for 2021.
I'm keeping my fingers crossed. In the long run, I think this would likely be hugely beneficial to Microsoft because it would fully cement Windows on Arm and the large number of Mac users who also use Windows could strongly incentivize Windows developers to start writing native apps. The Surface Pro X never really generated waves of enthusiasm or media buzz. M1 Macs could, amusingly, be exactly what is needed to make Windows on Arm come of age.
So you think Apple probably could add RAM to the package just as easily as before for the Mini, but have changed their thinking about what the lower-end Minis should offer?
I think it's more likely it's not as practical to do that as before, most likely due to the physical properties of the chip. Seems likely to me that for more RAM they'll need a larger package, one that would require different architecture around it for power and cooling. The Mini got what was optimized for the laptops, where size, power, and heat are more important considerations.
The way I look at it, there are competing laptops with a similar form factor and very similar features that manage 32GB of LPDDR - most notably the Dell XPS 13 - but that have noticeably less CPU and GPU performance. So it is weird to see that missing with quite a bit more performance. Not pathetic IMO, but I don’t blame people for being frustrated that RAM capability didn’t increase with the other main system components.
My guess is it’s because of this: https://forums.macrumors.com/thread...tery-last-only-10-hours.2271835/post-29317636
Manufacturing standardization and/or forcing power users to move to a higher, more profitable tier.
Orrrrr... anything Google and Electron are wasteful regurgitations of an App that frankly does not belong to the M1.
I swear Google is the master torturer of computers.
Chrome is a hog.
Their development tools run on some JS framework that has no regard for compute cycles.
Also their web apps. Google Ads, Google Meet, and even Google Suites use a lot of resources.
They are just wasteful.
I swear Google is the master torturer of computers.
Chrome is a hog.
Their development tools run on some JS framework that has no regard for compute cycles.
Also their web apps. Google Ads, Google Meet, and even Google Suites use a lot of resources.
They are just wasteful.
As I mentioned above, increasing the RAM capacity would likely require changing other things as well, such as the power and heat provisions. The M1 chips are optimized for the least power and heat, and apparatus to deal with those in a small space and limited cost, that Apple engineers can achieve while still allowing for high performance.
People who don't care as much about power and heat and so on may find the XPS approach more to their liking, but my impression so far is that most of those who have purchased M1 laptops have different preferences, are glad they use so little power and generate so little heat, and so on. Most probably wouldn't trade much of that for the option to add more RAM.
Frustration (we're using different terms each time!) of the kind that underlies complaints is based in not taking such things into account. Frustration that's simply a matter of lack wish fulfillment is another thing. I'm frustrated there's no 18" version, myself, but it's not something I feel complaints are justified about.
(I explained above why I think the theory you link to is unlikely.)
An 18” MacBook Pro would be awesome! I’m a fan of 17” laptops, and MacBook Pros are so light I could see that being still very portable and usable, especially with Apple Silicon.
This is not very many tabs, folks ... That does seem like a lot of damn simultaneous testing, though. You are thorough ... If I'm writing code, I am just testing and using one browser. I also try to use the minimal number of necessary IDEs for what I'm doing. Anyway, I'll later go through and try the other browsers one by one. I mean, your workflow is your deal but keeping all these open is pretty intense. If you don't have a way to switch between them quickly, I don't see the point. Do you have it set up to update the state of the application for all these browsers? Why not just run a test suite? Obviously, it's your deal, but I'm wondering what you are doing where you really benefit from all this being open at once. As in, do I need to up my game? If there were a way to have four screens open that show your app as you have a script, click through your app, that would be cool. But you said you had just one external monitor.
So I keep pages and applications open for two reasons: 1) they re-open faster, and 2) I have a mental model for where they are, i.e., I can (try to) quickly reference them.
I think the apple silicon largely fixes 1). Applications open rather quickly, as do pages. Also, I'm not an OS expert, but I wonder if the usage pattern of keeping all these open may actually preclude it from achieving even faster re-opening times. 2) is more of a UI challenge that I don't believe is solved by macOS, Safari, Chrome, or Firefox.
We take it for granted that OS and browser UI is so bad. You can sort of use Alfred/Spotlight to quickly open applications and close them, but they should also return to their last state when re-opened. It's easier to close and not quit (macOS makes this strange default decision to only visually close applications). The Dock could be useful, but this default keeps everything running. Concerning tabs, sure, you can bookmark each, but that is a lot of steps - particularly if you are using Safari, which makes it time-consuming to create and manage bookmark folders (really, I have to click Bookmarks -> Create New Folder before I Add Bookmark then scroll and type to find this new folder ...). What you really need is, say, a 'session' - a single page that shows all of today's tabs (until marked for exclusion) that itself only is searchable in Alfred/Spotlight or via that page itself. Dynamically updated as you enter words for your query. Of course, the same system must allow other bookmarks to be optionally searchable. At the end of the day, this UI problem is a search (and tag) problem. An improved search could be quicker than hunting for open tabs and keeping some tabs open. Maybe the text search is complemented with visual previews. Anyways, when you stop looking at a page, it should close it. When you re-open, it retains state and AS ensures it opens as quickly as possible. RAM would certainly be used to ensure this, but perhaps it could be used more intelligently by the OS if it actually quits things not in use.
Thankfully I do have Swift Window Switcher in Alfred. I have a bookmark search there, but it's not intelligent, as I alluded to previously. I should be able to say "today" or a topic or "active" and also tag a page to a topic (well, actually change its content-based automatically selected topic). And when you close a page, you swipe left to end its active state, swipe right to keep it active but close, swipe up to share or add as a bookmark with an associated topic. Why are we using folders to manage context-less, tagless, and practically unsearchable bookmarks in 2020?
I didn't say "just as easily." I think it's not a physical limit per se - and that there were practical reasons, including market positioning, to not expand it much further. Just that it's not a 'physical' limit or that there's an insurmountable barrier.
Another poster wrote (much better than I phrased it) "Physical limit in the sense that you can't fit than two modules on the package and making the package larger just to accommodate the few % of users who need more than 16GB doesn't make sense economically." (Note Im not approving/confirming the physical/design limit sense, this is well beyond my knowledge level - I just doubt the barrier is insurmountable)
It may be it's a much higher barrier than I think, but practical/economical to address the users who need it is what I was aiming at describing. To put it simply, if it cost $100 million 'extra', for apple it may be a no-brainer. If it cost $10 billion, almost certainly not. Where is it in between these extremes, I have no idea. (I put this in terms of just $$, which is a bit crass, it may be more of a timeline limit or dedicated/critical personnel ie opportunity cost, staff they want to devote to the next gen chipset)
There used to be PCIe cards that held RAM sticks that presented to the system as a hard disk drive about 15 years ago and you used this as a RAMDisk. That is you accessed it like a disk but it was actually just doing disk operations to RAM. You lost all of your data when the machine went down.
It seems to me that you could use a similar concept for page and swap files on a RAM-constrained system. I took a quick look around and the only thing that looked close was Intel Optane NVMe cards with 32 GB of RAM and 480 GB SSD. I imagine that you could hook this up with a TB4 or TB3 adapter - the question is does this just work like a regular SSD and can you get it to only access the RAM? It strikes me that someone could make a rather small device to add 32 GB of RAM presented as SSD and then use it for page/swap and that it would be a lot faster than page/swap to the M1's SSDs.
It seems to me that you could use a similar concept for page and swap files on a RAM-constrained system. I took a quick look around and the only thing that looked close was Intel Optane NVMe cards with 32 GB of RAM and 480 GB SSD. I imagine that you could hook this up with a TB4 or TB3 adapter - the question is does this just work like a regular SSD and can you get it to only access the RAM? It strikes me that someone could make a rather small device to add 32 GB of RAM presented as SSD and then use it for page/swap and that it would be a lot faster than page/swap to the M1's SSDs.
Try using the new Webrender renderer with Firefox, it seems much faster and more efficient:
Firefox on Big Sur?
I am using Firefox on my M1 MBA via Rosetta 2. It is fine. However, I have a pro-tip for the slightly adventurous, go in an enable the new Webrender renderer on it and it will be night/day in terms of rendering speed and reduced CPU utilization, even under Rosetta. Here's a link that explains...
forums.macrumors.com
Look at WindowServer.. that's high. It's a 'Bug' Sur thing. You using a mouse? That makes it worse...
IntelliJ is maybe something that could benefit from Apple's neural CPU acceleration. Considering they are still running on Rosetta 2, they're probably not optimizing for that yet either.
This article touches on this discussion, with a similar point to what I was trying to say:
So yeah, there's a physical limit for the size of the package they rolled out for the first round.
They probably could have gone for more memory options/bigger package - I think without fundamental design changes - but one small package that they could use in three more or less entry level machines in identical configs simplified the prject tremendously. Great leverage of a single team to do one package and focus on getting it right for a large customer base.
And they have the data, they know their sales. And the vast majority of sales - 95% - doesn't go for more than 16gb.
There's probably a similar breakdown for the multiple screen users - just not that many users need it. And those who do, they're also the ones who want four ports and 32gb of memory. The multiple external screens issue may be a lot more complicated than more ram - more of a departure from the SoCs used in the ipad/iphones, but comes back to the same thing - a premium capability for a smaller chunk of the market.
Oh, and I'd love 32gb in a laptop, but I can more than get by with 16gb, and for $400 can live with it.
I'm not sure what we'll want ML accelerators for... considering they're more geared towards ML. Most of IDEA's performance issues seems to come from its excessive (and kinda inefficient?) indexing process. It very regularly tries to index every file that's marked as "source" within the project for any small change.
Perhaps just making it "native" to M1 instead of running it through Rosetta 2 will help tremendously in this case. We'll see. For now, we're stuck with having to run it through Rosetta 2. Performance is not an issue, but efficiency certainly is.
You don't understand how memory management works in a modern OS. The purpose is to use as much memory as possible at every time no matter how few application you use.
Memory not used is wasted memor. If you have a lot of memory macOS will use that memory. If you have less memory, macOS will use less memory.
Look at the memory pressure. It is green and therefore you have enough memory.