64GB RAM is not a wise buy, and here's why

[TimothyB]

As a graphic designer, 64 is tempting for peace of mind in a fast paced work environment where software and projects easily spiral out of control, but not likely to happen on my personal machine.

There are times when the iMac I use at work with 40gb will run out of memory, requiring me to sacrifice one program/projects to unfreeze other software with unsaved work.

For personal projects at home, I’m likely not going to overwhelm the system. If one day there’s a project that needs the memory, I could fall back to my gaming PC where adding another 32gb is half the cost.

 

[pshufd]

Most of my development work is done on development cloud servers. I have specs, reports, etc. on my laptop but those take up trivial amounts of space. The thing that takes up space is personal videos. I do my personal stuff on my 2014 with 500 GB and keep the videos on an 512 GB SD card. Actually email takes up a lot of space too but I'd guess that most people don't need to keep 30 years of email online - it's definitely stuff that I could archive. So I don't need 1 TB. It's nice to have but I could probably get by with 256 GB of primary and 512 GB of secondary - shame that the new Macs don't have an SD slot.

What would be cool us a USB-C drive that stuck out maybe 3/8 of an inch that you could always leave plugged in.

 

[Glockworkorange]

My point was if you purchase 64GB now to cover your basis and "future" proof you are bottlenecking yourself because in 3-4 years when 64GB may be more feasible the rest of your MBP will have a 3-4 year old graphics card and processor.

I don't know if anyone would pay those prices for RAM just to "future proof" their machines (unless they have $$ to burn).

If someone buys 64GB, I'm going to assume they have a use for 64GB right now; if not, then I think it's probably a waste of money.

I guess their are a very small number of people who need that much RAM and this machine is for them, but I think the cost of that RAM is like $2,000.00.

 

[bsbeamer]

32GB to 64GB is a $400 upgrade. If you’re buying this machine for video production work, motion graphics, 3D, animation, even some audio production - is that savings really helping you long term?

 

[pshufd]

I don't know if anyone would pay those prices for RAM just to "future proof" their machines (unless they have $$ to burn).

If someone buys 64GB, I'm going to assume they have a use for 64GB right now; if not, then I think it's probably a waste of money.

I guess their are a very small number of people who need that much RAM and this machine is for them, but I think the cost of that RAM is like $2,000.00.

You'd be better off with something like the Dell Precision 7740 which has 4 SODIMM slots so you could use 16 GB SODIMMs which are a lot cheaper than 32 GB SODIMMs. The 7740 can hold up to 128 GB of RAM.

 

[Glockworkorange]

I’m considering 64GB mainly as working in IT, I want to use laptop for studying/lab environments with virtualization and multiple VMs running - currently I’ve got a 16GB macbook and it starts to struggle with the workload.

However more frustrating then struggling with resources is the fan noise like a jet engine once you have 3 / 4 VMs running at the same time - I would go for the top spec MacBook Pro 16 with i9 and 64GB ram if it could run multiple VMs comfortably and without the annoying fan noise.

I could probably go with 32GB memory wise - but for my use case , 64GB would make the machine more future proof.

Yeah, but is the insane increase in price worth it? Would you spend the $$? It's not a trivial cost.

 

[Falhófnir]

Think it's pretty obvious that it's currently applicable to only a very small number of people, and those people will know who they are? Still going to see people recommending 32GB and 64GB to people who are browsing Facebook to 'future proof' their purchase, though 🙄

 

[Glockworkorange]

32GB to 64GB is a $400 upgrade. If you’re buying this machine for video production work, motion graphics, 3D, animation, even some audio production - is that savings really helping you long term?

I thought it was thousands, not hundreds? Or maybe I am thinking of the costs from 16 to 64?

 

[php0614]

9980hk is kind of powerful, but anyway it is a mobile processor, so it cannot handle 64gb ram well in most cases including 3d stuffs. Also the 5500m is better than Vega 20 but when compared to desktop cards or top tier mobile cards like 2080 max-q or Quadro 5000, it is quite limited. In short, 16 inch Macbook pro this year cannot handle large 3d projects that need more than 32gb ram well. It is much wise to save the 400 USD for a desktop upgrade like purchasing AMD 3990wx processor or something, or for next Macbook buy in 3-4 years.

 

[bsbeamer]

I thought it was thousands, not hundreds? Or maybe I am thinking of the costs from 16 to 64?

Nope.
16GB > 32GB is $400 upgrade
16GB > 64GB is $800 upgrade
32GB > 64GB is $400 upgrade

Think about the cost/value for your individual needs. These prices are really not that bad, especially for Apple BTO.

 

[Stephen.R]

I think the cost of that RAM is like $2,000.00.

Upgrading the base 16" machine to 64GB is $800.

which has 4 SODIMM slots

Another thing it has, is Windows.

I thought it was thousands, not hundreds? Or maybe I am thinking of the costs from 16 to 64?

Nope, not that either. 512GB to 8TB SSD is $2.4K - maybe that's what you're thinking of?
[automerge]1573919970[/automerge]

Think it's pretty obvious that it's currently applicable to only a very small number of people, and those people will know who they are? Still going to see people recommending 32GB and 64GB to people who are browsing Facebook to 'future proof' their purchase, though 🙄

If someone is asking for recommendations and their use case is "browse facebook" my recommendation to them would be an iPad.

For someone who actually wants a computer to do something more than that, an upgrade from 16 to 32GB adds just 16% to the cost, but will undoubtedly extend the usable life of the machine. My 2011 MBP17 was used practically every day for software development and ops/sysadmin work until about a year ago, thanks to maxing the RAM at 16GB (yes aftermarket max, Apple max was only ever 8GB) shortly after purchase.

Most consumer apps are not CPU or GPU intensive, but plenty use a metric **** ton of memory because files get bigger, developers get 'lazy', etc. I generally recommend not to use Electron based apps, but if you want/need to, you'll appreciate as much memory as you can get - sure some are not complete trash, but the majority of them are red hot flaming dumpster tire fires when it comes to memory use.

So, if you can't afford the extra $400, then no, don't get 32GB. But if you can, you're extending the life of it - either for you personally, or, for someone else - the higher spec, the more likely someone else can make use of it when you're done - either someone you give it to (e.g. a relative or charity), or someone you sell it to.

 

[The Mercurian]

Sigh this again.

Ok this is me TRYING to keep parallel statistical models small:


My larger models are 4GB in size and I can only run 4 threads in parallel out of a possible 12. Running more at once directly = faster results and increased productivity. I would like to have an 8 core machine running 14 of these side by side (save 1 core to do other things). Thats 14 x 4GB in RAM = 56GB.

So if you find yourself saying "nobody needs XX GB thats ridiculous".... remember....

NOT EVERYBODY DOES WHAT YOU DO!!!!

 

[pshufd]

Sigh this again.

Ok this is me TRYING to keep parallel statistical models small:
View attachment 877528

My larger models are 4GB in size and I can only run 4 threads in parallel out of a possible 12. Running more at once directly = faster results and increased productivity. I would like to have an 8 core machine running 14 of these side by side (save 1 core to do other things). Thats 14 x 4GB in RAM = 56GB.

So if you find yourself saying "nobody needs XX GB thats ridiculous".... remember....

NOT EVERYBODY DOES WHAT YOU DO!!!!

We have a lot of systems in my office with 1TB+ of RAM. So yeah, there are applications which require a lot of RAM. But they are fairly rare. The advent of running large datasets on PCs is relatively recent.

I remember the 70s where we had 16-bit systems with addressable program space of 32KB and we didn't have virtual memory back then. We had to write code-section overlays which would swap in small parts of your program at a time to run the code, then swap in other pieces of code to run when they needed to. We would make programs smaller as well and chain programs together storing work product on disk to go from one phase to the next.

I still have a working computer in my basement with 250 words of executable memory.

 

[CarbonCycles]

^^^ Been there done that....even as recent as last week.

Threads like these really annoy me..it's somewhat insulting that someone thinks their workloads or use patterns are more relevant than others.

I think a lot of folks fail to realize that MBPs are used extensively in large companies like Amazon, Facebook, etc. in departments ranging from administration, engineering to research/development.

Not everyone wants or needs to fire up the Spark Cluster or MPI platform to debug their code or run a quick n' dirty simulation.

 

[The Mercurian]

We have a lot of systems in my office with 1GB+ of RAM. So yeah, there are applications which require a lot of RAM. But they are fairly rare. The advent of running large datasets on PCs is relatively recent.

I remember the 70s where we had 16-bit systems with addressable program space of 32KB and we didn't have virtual memory back then. We had to write code-section overlays which would swap in small parts of your program at a time to run the code, then swap in other pieces of code to run when they needed to. We would make programs smaller as well and chain programs together storing work product on disk to go from one phase to the next.

I still have a working computer in my basement with 250 words of executable memory.

So this the other thing. People see that and automatically think 'big data'. My datasets are about 150k. The statistical models themselves are big. And we mathsy types can always make them bigger. And this software R statistical software is not rare - its one of the main software options in a small little thing called data science - heard of that? Note that I said I was trying to keep the models small

 

[pshufd]

deleted

 

[CarbonCycles]

So this the other thing. People see that and automatically think 'big data'. My datasets are about 150k. The statistical models themselves are big. And we mathsy types can always make them bigger. And this software R statistical software is not rare - its one of the main software options in a small little thing called data science - heard of that? Note that I said I was trying to keep the models small

That's the part ppl fail to understand is that the intermediate calculations need to get cached in memory, which blows-up the size exponentially.

PPL in our field are definitely in the minority on this forum, but we're the ones who are always asking for more technical enhancements. I really couldn't care if the MBP was painted pink w/ polka dots and ran 8k...it would be lost on me if it didn't have the physical esc key and I would love a NVIDIA GPU.

 

[The Mercurian]

^^^ Been there done that....even as recent as last week.

Threads like these really annoy me..it's somewhat insulting that someone thinks their workloads or use patterns are more relevant than others.

I think a lot of folks fail to realize that MBPs are used extensively in large companies like Amazon, Facebook, etc. in departments ranging from administration, engineering to research/development.

Not everyone wants or needs to fire up the Spark Cluster or MPI platform to debug their code or run a quick n' dirty simulation.

People don't realise that using high performance computing if you have it available is hassle in its own right. First you gotta apply for access - depending on where you are justfiy your use case, maybe wait in a queue - then oh **** I made a mistake gotta do all the process all over again.... etc etc....

 

[pshufd]

So this the other thing. People see that and automatically think 'big data'. My datasets are about 150k. The statistical models themselves are big. And we mathsy types can always make them bigger. And this software R statistical software is not rare - its one of the main software options in a small little thing called data science - heard of that? Note that I said I was trying to keep the models small

Google and Oracle offer their employees Macs or Lenovos and Macs are very popular.

Macs are very popular in the bio world too. So there are use cases for a lot of RAM. But, I think that the percentages that need this are still very small. You have the trading world too - I have no idea why professional trading applications need so much memory - I suspect that it's bad programming - but I can't argue with how they develop software; I'm just a user.

 

[The Mercurian]

Google and Oracle offer their employees Macs or Lenovos and Macs are very popular.

Macs are very popular in the bio world too. So there are use cases for a lot of RAM. But, I think that the percentages that need this are still very small. You have the trading world too - I have no idea why professional trading applications need so much memory - I suspect that it's bad programming - but I can't argue with how they develop software; I'm just a user.

Yeah I dont' know anything about trading software either - but there is a big trend across fields towards Bayesian computing and that often means Markov Chain Monte Carlo methods and if your model is anything more complex than a simple regression these calculations get big and slow real fast.

The other point is - not everyone can do everything. My models above are my code calling someone elses code - I can't optimise the other guys code as I don't understand it's several thousand lines of code on a line by line level (which is what you need for this kind of tweaking). In this case I was able to optimise my end of it to run a bit leaner but that in itself is extra effort. I simply can't take on the other guys code too!

 

[PROFESS0R]

Like some others here, the more memory I have, the better.

It seems that some here prefer to speak in terms of overly broad sweeping statements without data to back up their words. As a professor, when I see the words “all”, “never” or “most”, I cast a critical eye towards the individual writing or speaking these words. Unless one has completed a probability sample of a population of whatever is being studied, and unless one speaks in terms of confidence intervals and / or margins of error, one cannot say “most” with any degree of accuracy.

Pedantic? Perhaps.... but I think it is not too much to expect a bit more critical thinking.

Joe

 

[maflynn]

why do you care?

Who says I care. They can do what ever they want, but when people ask about getting memory, I offered my opinion

 

[kmahmood]

Yeah, but is the insane increase in price worth it? Would you spend the $$? It's not a trivial cost.

Depends on how you look at it, some would say it’s insane to spend so much on a MacBook and not go for a cheaper laptop option like a HP or Dell.

but given if one is prepared to spend up to 3k on a premium laptop then increasing that cost by another 7-10% that might make the machine last a couple more years then it may be worth it.

but yes for me my priority would be a laptop that can do what I want (run multiple VMs) comfortably without getting stressed and loud annoying fan noise — I might be prepared to pay a premium price for this, but only if it does what I’m after.. otherwise maybe as someone else suggested I should give up on the laptop dream and get a desktop like the IMac Pro for my needs. (Not that that comes cheap either)

 

[CM_1986]

Like some others here, the more memory I have, the better.

It seems that some here prefer to speak in terms of overly broad sweeping statements without data to back up their words. As a professor, when I see the words “all”, “never” or “most”, I cast a critical eye towards the individual writing or speaking these words. Unless one has completed a probability sample of a population of whatever is being studied, and unless one speaks in terms of confidence intervals and / or margins of error, one cannot say “most” with any degree of accuracy.

Pedantic? Perhaps.... but I think it is not too much to expect a bit more critical thinking.

Joe

What kind of work are you doing "professor" that requires a 5000+ MBP you have in route?

 

[pshufd]

People don't realise that using high performance computing if you have it available is hassle in its own right. First you gotta apply for access - depending on where you are justfiy your use case, maybe wait in a queue - then oh **** I made a mistake gotta do all the process all over again.... etc etc....

We have queues like that at work. There's a web page listing different architectures and disk/memory/core sizing. You pick your system, then operating system and then wait until a machine is available. Then you move your environment over and run your tests and do your debugging.