I am really torn between 8GB vs. 16GB RAM

[pshufd]

So you have a main laptop and two backups? Hah, nice. I don't think of mine as back-ups but I have two older machines, one MBA and a Surface.

I have a 2015 which is clustered with a Late 2009 iMac and a 2008 Dell XPS Studio so it's used as more of a desktop.

I also have a custom built i7-10700 which I use as a trading workstation and it also runs my virtual machines. It's also the home NAS. All of this hardware is in the basement.

I also have a 2014 MacBook Pro 15 which I use casually in the living room and generally in living spaces.

A MacBook Pro/M1 could be used as a living room machine as I wouldn't need to charge it very often. I could also cluster it with the i7-10700 for office and M1 development. It really wouldn't fit in with the other cluster. I would likely sell the 2014 or give it away if I got the M1 for myself.

I've also been thinking of getting a Mini as well and clustering it with the 10700 for office and development and leaving an MBP/M1 in the living room for casual use.

 

[pshufd]

But then many, many people on this forum feel very strongly about not spending $200 for 8gb extra ram..... , never mind a whole computer just to fill a gap for 6 months.

We need to get an 8 GB Apple Stimulus upgrade program.

My guess is that those with bitcoin aren't too worried about $200:

 

[Quackers]

You might have multiple use cases and want to kick the tires. So buy an Air just to use around the house to try it out and then buy a real system when they are available.

Then buy a real system?
That is extremely dismissive and a vast over-simplification.
And I have a M1 MBP

 

[pshufd]

Then buy a real system?
That is extremely dismissive and a vast over-simplification.
And I have a M1 MBP

My trading station has 64 GB of RAM, discrete GPU and three 4k monitors and 5 TB of SSD. I'm waiting for Apple to come out with real systems.

 

[pshufd]

The performance is so close that either will satisfy 90% of the use cases, pro or casual; just pick the shape, keyboard, and level of noise you prefer. I'm glad Apple didn't limit the Air compared to the Pro.

Apple may have wanted to keep production as simple as possible and that would mean just making the MBP/M1 the same as the MBP/Intel.

 

[Quackers]

My trading station has 64 GB of RAM, discrete GPU and three 4k monitors and 5 TB of SSD. I'm waiting for Apple to come out with real systems.

Congratulations. So not a laptop then.

 

[krishmk]

My trading station has 64 GB of RAM, discrete GPU and three 4k monitors and 5 TB of SSD. I'm waiting for Apple to come out with real systems.

It might take a year or so for Apple to come up with those options.

On a side note which trading platforms do you use?

 

[Argon_]

Honestly, I couldn't disagree more with your post, respectfully. I mean, I am sorry but I don't share your opinion. Limiting artificially the MacBook Air, especially this year that both MBA and Pro are so similar in performance, would be ugly, and would force anyone who wants an improved machine (RAM/storage) to get a machine with a Touch Bar. And many of us don't like the Touch Bar.

Do you wonder why people with heavy workloads are going for a MacBook Air? Because they like the design more than a MacBook Pro, and they have seen the reviews, and the performance only drops a little with sustained workloads. It is almost as capable as a Pro, without the annoying TouchBar, at a slightly lower price point.

However, I do agree the Pro should have been offered in 16 and 32GB of RAM

I agree that the pro should be offered with more RAM, but that seems to be an M1 limitation. We'll see the performance version this year, which will hopefully be optionable up to 64. Not that I need 64 gigs of RAM, but some people do. I might consider 32 though, for heavy multitiasking.

 

[pshufd]

Congratulations. So not a laptop then.

Not a laptop or a Mini. I was disappointed that the Mini only supports two displays.

 

[pldelisle]

My trading station has 64 GB of RAM, discrete GPU and three 4k monitors and 5 TB of SSD. I'm waiting for Apple to come out with real systems.

Not quite sure a trading station actually needs all that ...........

 

[pshufd]

It might take a year or so for Apple to come up with those options.

On a side note which trading platforms do you use?

Active Trader Pro and Think or Swim. Active Trader Pro is a pig in terms of CPU. It also crashes or hangs on a regular basis. But it runs my algorithmic trading screens reasonably well.

 

[rui no onna]

The spare computer. A first world solution to a first world problem.

All joking aside, I do the same. It drastically reduces downtime if some calamity happens.

Lol, I've ran out of spare computers so I replenished last year.

Well, not quite ran out. I got tired of cleaning dust bunnies from the bigger, heavier desktops so when Windows 7 EOL came around, I bought a few NUCs to replace them (32GB RAM $120 + 1TB NVMe $120). For office type stuff, they're more than fast enough.

The NUCs are more than okay for the Plex Media Server and the light virtualization I do as well. I mean, even the Raspberry Pi 4 can handle HTTP+PHP server duties just fine.

 

[pshufd]

Not quite sure a trading station actually needs all that ...........

The 5 TB storage is for the NAS. I run a couple of virtual machines on it and will do Firefox and Thunderbird development on it from time to time. It typically only uses about 29 GB of RAM with about 20 GB cached. Firefox would likely use a lot of RAM for parallel compiles.

 

[pldelisle]

The 5 TB storage is for the NAS. I run a couple of virtual machines on it and will do Firefox and Thunderbird development on it from time to time. It typically only uses about 29 GB of RAM with about 20 GB cached. Firefox would likely use a lot of RAM for parallel compiles.

Okay, so not only using it for trading.

Yes compiling Firefox is insane. Tensorflow is impressive too

 

[pshufd]

The spare computer. A first world solution to a first world problem.

All joking aside, I do the same. It drastically reduces downtime if some calamity happens.

In WFH/SFH, your productivity depends on your hardware. When we went WFH, we had a few people on Slack that had hardware problems and the response was to take it into the Apple Store or make a service call. That can mean that you're down for three to five days. That's really awful when you have work deadlines.

In a trading sense, downtime is money.

A few seconds or certainly a few minutes can be very expensive.

 

[redfirebird08]

My parents are very light users, but I went ahead and updated them to 16 GB of RAM. Their M1 Mac Mini arrives around January 11th. Stimulus payment from the government ($1,200 for them as a couple) is covering the full cost of the computer, haha. I just figure with 16 GB of RAM, it helps future proof the machine for operating system updates over the next 5+ years.

 

[pldelisle]

My parents are very light users, but I went ahead and updated them to 16 GB of RAM. Their M1 Mac Mini arrives around January 11th. Stimulus payment from the government ($1,200 for them as a couple) is covering the full cost of the computer, haha. I just figure with 16 GB of RAM, it helps future proof the machine for operating system updates over the next 5+ years.

You are courageous saying this on this thread.... ? ? ?

 

[ScreenSavers]

I actually went with 8GB RAM for my personal M1 MacBook Air. I never thought I’d step back from my 16” MBP with 32GB and my Mac Pro with 64, but both of those machines are now gone. If my iPad Pro is as fast as it is with 6GB, I think Apple knows what it’s doing with memory management on its CPUs. I haven’t maxed it out once yet, and it comfortably has 2GB or so free, even while working on large projects in FCP X.

That being said, I did opt for 16GB in my Mac Mini, arriving tomorrow.

 

[dogslobber]

I bought a Mini 16gb with 512gb SSD which should arrive Thu. I think for longevity you want the 16gb. I buy these machines to last a long time. This one I got for free though : )

 

[Argon_]

In WFH/SFH, your productivity depends on your hardware. When we went WFH, we had a few people on Slack that had hardware problems and the response was to take it into the Apple Store or make a service call. That can mean that you're down for three to five days. That's really awful when you have work deadlines.

In a trading sense, downtime is money.

A few seconds or certainly a few minutes can be very expensive.

You're absolutely right. Downtime is money.

I've been saved in the past by the presence of a spare computer. Just like a fire extinguisher, it can prevent harmless issue from becoming a harmful one.

 

[Argon_]

Lol, I've ran out of spare computers so I replenished last year.

Well, not quite ran out. I got tired of cleaning dust bunnies from the bigger, heavier desktops so when Windows 7 EOL came around, I bought a few NUCs to replace them (32GB RAM $120 + 1TB NVMe $120). For office type stuff, they're more than fast enough.

The NUCs are more than okay for the Plex Media Server and the light virtualization I do as well. I mean, even the Raspberry Pi 4 can handle HTTP+PHP server duties just fine.

I prefer laptops for spare computers. Keep them in a neat stack, no worry about peripherals. Now with the USB C standard, current macs will make excellent future spares. No fumbling for chargers.

Your use of NUCs is a great idea. I'm contemplating one with no OS, plus Linux for a media PC. The low spec ones are so inexpensive.

 

[Sydde]

Really? So the code will NOT be faster retaining/releasing an object? It is a memory transaction which does mean the program itself will be faster.

And let me add to this, when the reference count is zero, the memory is cleared. So faster to release to where the reference count is zero -> memory gets cleared. NSObject is faster to release on Apple Silicon, therefore, it is faster clearing out memory.
o

Yes, I am sure about that. If an object gives up another object in its instance context, it will often deliver the other object after an autorelease; whatever receives that object will decide whether it wants to hang onto that object, in which case it will retain that (already allocated) object to prevent the autorelease pool from purging it. The retain and release operations are faster, but there is no change in memory, as the retain count goes from 1->2 after the retain and then 2->1 in the autorelease operation.

[[<NSObject> alloc] init] returns an object that already has a retain count of 1, which is part of the alloc method. Additional retains and releases only alter the value of the retain count, until it goes to 0, after which the memory gets freed up (which itself may only amount to changing two or three numbers in a table to mark the block as free). So, a fair number of release operations do not even directly free memory – sometimes not at all. But retain counts (a single number) can change quite a lot over the course of a process cycle, so having them run faster does help performance.

 

[Ethosik]

Yes, I am sure about that. If an object gives up another object in its instance context, it will often deliver the other object after an autorelease; whatever receives that object will decide whether it wants to hang onto that object, in which case it will retain that (already allocated) object to prevent the autorelease pool from purging it. The retain and release operations are faster, but there is no change in memory, as the retain count goes from 1->2 after the retain and then 2->1 in the autorelease operation.

[[<NSObject> alloc] init] returns an object that already has a retain count of 1, which is part of the alloc method. Additional retains and releases only alter the value of the retain count, until it goes to 0, after which the memory gets freed up (which itself may only amount to changing two or three numbers in a table to mark the block as free). So, a fair number of release operations do not even directly free memory – sometimes not at all. But retain counts (a single number) can change quite a lot over the course of a process cycle, so having them run faster does help performance.

You still did not say anything about when the release action goes from 1 -> 0 resulting in memory changes. The faster it goes to 0 references (as in the faster it releases to reach the 0 count), the faster the memory is cleared. It is all linked.

Also, unless the memory releases are asynchronous by default, the program will halt for those nanoseconds while it releases the memory counts. Therefore, those nanoseconds add up so the program can do more things.

 

[alien3dx]

You're absolutely right. Downtime is money.

I've been saved in the past by the presence of a spare computer. Just like a fire extinguisher, it can prevent harmless issue from becoming a harmful one.

I learn my lesson on iMac. Will assume every 3 year also need to beware.Laptop 1 to 2 year max usage

 

[pldelisle]

Learnt it the hard way too. I fried my 2013 MBP nvidia gpu a couple of months before end of apple care (I’ll never train an ML model again on any laptop, mac or PC). Been 4 days without a laptop. I had my iPad Air which saved a bit my life for school work, but 4 days without a computer can be very hard.