8GB RAM in M3 MacBook Pro Proves the Bottleneck in Real-World Tests

The unified memory is a different topic.

Focusing purely on performance... how much performance gain do you think it gives? In what software? Do you think there is more than a 0.5% gain in any usage scenario?
There are substantial energy gains from the physics of being extremely close to the SoC (reduced capacitance).
There are probably minor gains in having a cleaner signal path (easier, and less frequent, equalization).
What is missed in these discussions is a 3rd issue, namely that if the DRAM is fixed, it can be CHARACTERIZED.

Most DRAM (this is well known in academia) has the property that it is well above spec in most respects (loses charge much slower than spec so can be refreshed less often, requires less RAM to hold the charge, etc) but every RAM chip has a few pages that let down the team and only just meet spec.
IF your RAM is fixed, then you can
- characterize in the factory
- record the characteristics in a ROM associated with the memory controller
- exploit the characteristics. The most obvious of which is refresh each page at the rate it really requires, not at the rate spec says. It may also be possible, for example, to run some of the DRAM chips on lower voltages.

Does Apple do any of this? There are good reasons to believe they do. Much of the work in this field (characterizing real DRAMs, and suggesting how to exploit their features) has been done by Onur Mutlu at ETH, and Apple has worked with him (they have at least one joint patent, and at least one paper I've seen). There are also Apple patents that suggest their memory controller is consulting a look-up ROM to decide on how the exact parameters to use for handling parts of a DRAM.

I've no idea how to quantify any of this. However it is a fact that Apple seems to get DRAM throughputs that are like 98% of theoretical maximum, and Intel gets throughputs that are like 60% of theoretical maximum, so...
(I don't know where Intel is at right now. When I last looked at this around M1 time, my recollection is Intel is at ~60%, just something ridiculously bad compared to Apple. Maybe someone has current Intel numbers?)
 
There are substantial energy gains from the physics of being extremely close to the SoC (reduced capacitance).
There are probably minor gains in having a cleaner signal path (easier, and less frequent, equalization).
What is missed in these discussions is a 3rd issue, namely that if the DRAM is fixed, it can be CHARACTERIZED.

Most DRAM (this is well known in academia) has the property that it is well above spec in most respects (loses charge much slower than spec so can be refreshed less often, requires less RAM to hold the charge, etc) but every RAM chip has a few pages that let down the team and only just meet spec.
IF your RAM is fixed, then you can
- characterize in the factory
- record the characteristics in a ROM associated with the memory controller
- exploit the characteristics. The most obvious of which is refresh each page at the rate it really requires, not at the rate spec says. It may also be possible, for example, to run some of the DRAM chips on lower voltages.

Does Apple do any of this? There are good reasons to believe they do. Much of the work in this field (characterizing real DRAMs, and suggesting how to exploit their features) has been done by Onur Mutlu at ETH, and Apple has worked with him (they have at least one joint patent, and at least one paper I've seen). There are also Apple patents that suggest their memory controller is consulting a look-up ROM to decide on how the exact parameters to use for handling parts of a DRAM.

I've no idea how to quantify any of this. However it is a fact that Apple seems to get DRAM throughputs that are like 98% of theoretical maximum, and Intel gets throughputs that are like 60% of theoretical maximum, so...
(I don't know where Intel is at right now. When I last looked at this around M1 time, my recollection is Intel is at ~60%, just something ridiculously bad compared to Apple. Maybe someone has current Intel numbers?)
That’s very interesting, I never heard about that before. Thanks for sharing that, good points! 👍🏻
 
On this, I am no expert. I do know Anandtech tried to max out the M1 Max's memory bandwidth a few years ago and could not figure out a CPU workload to do it, although they apparently did manage to get pretty close.

From what I understand, it's primarily the GPU-intensive workloads that are much more likely to max out such a large amount of memory bandwidth, but this might be a better question for @name99
I remember the tests. No one could find even a slight gain from having all that bandwidth, try as they might. Why do we think Apple don't even bother proving all that unused bandwidth anymore on the M3 series. My 0.5% overall gain from the soldered RAM suggestion was exaggerated. The research suggests the benefit has been primarily been graphical, on extreme tasks, and theoretical in AAA games that don't exist on Mac.

The main benefit to Joe Bloggs has been battery improvement, but now the lower power chips are being produced in a form that could be socketed, if Apple wanted them.The could even run at higher frequencies than Apple use. The thinness saving is negligible in a MacBook Pro.

Apple save a buck in construction costs, and make their product non repairable.

Anyone want to try to defend the soldered SSD next? Saving 1/10th of a mm and $1 in construction costs aren't good arguments.
 
I remember the tests. No one could find even a slight gain from having all that bandwidth, try as they might. Why do we think Apple don't even bother proving all that unused bandwidth anymore on the M3 series. My 0.5% overall gain from the soldered RAM suggestion was exaggerated. The research suggests the benefit has been primarily been graphical, on extreme tasks, and theoretical in AAA games that don't exist on Mac.

The main benefit to Joe Bloggs has been battery improvement, but now the lower power chips are being produced in a form that could be socketed, if Apple wanted them.The could even run at higher frequencies than Apple use. The thinness saving is negligible in a MacBook Pro.

Apple save a buck in construction costs, and make their product non repairable.

Anyone want to try to defend the soldered SSD next? Saving 1/10th of a mm and $1 in construction costs aren't good arguments.
Nobody has LPDDR RAM in a socket-able form, it’s all soldered currently. There might be some coming out next year, but it currently doesn’t exist. Soldered components are more power efficient. And the SSD is fine as well, I will defend Apple. Seems odd you’re hanging out in an Apple Mac fan group, but all you seem to say is antagonistic towards Apple. Apple makes better systems, and they use soldered components. If you don’t like it, there’s plenty of other options out there…🤷🏼‍♂️
 
Nobody has LPDDR RAM in a socket-able form, it’s all soldered currently. There might be some coming out next year, but it currently doesn’t exist. Soldered components are more power efficient. And the SSD is fine as well, I will defend Apple. Seems odd you’re hanging out in an Apple Mac fan group, but all you seem to say is antagonistic towards Apple. Apple makes better systems, and they use soldered components. If you don’t like it, there’s plenty of other options out there…🤷🏼‍♂️
How many Apple complaints do I have? It's a short list. Do you have a similarly short (but different) list of improvements Apple could make, or would you argue that they are infallible?

Reports suggest LPDDR RAM is now available for order in a socketed form, so we could see such products next year.
 
Dude has no idea what's a bottleneck or limiting factor in a design space. There's only one set of configuration that has all components perfectly matched.

There is absolutely nothing wrong with the existence of a bottleneck.
There is no bottle without a bottleneck. It would have been called a jar otherwise. Jokes asides, every complex system has a narrow/weak point and I haven't seen enough compelling comparisons to pin everything onto the small RAM capacity of the base level MBPM3. Such a test should go beyond simply throwing a large file at a machine and watch the timer.
 
Wrong, at some point RAM won't get your computer to be faster, only from 8 to 16, from 16 to 32 the "faster" effect will be much minimal, and from 32 to 64 will be even less.

At some point getting more RAM won't get you any benefit.
You realize you claimed it wouldn’t and then admitted it does right?
 
How many Apple complaints do I have? It's a short list. Do you have a similarly short (but different) list of improvements Apple could make, or would you argue that they are infallible?

Reports suggest LPDDR RAM is now available for order in a socketed form, so we could see such products next year.
I’m not saying you can’t have any criticisms, I just find it odd that they don’t even seem tempered by anything you appreciate about the setup. Most of the time you can balance the pros and cons of a system, but it comes across like you’re just overemphasizing the cons with no consideration of any of the pros. I’m not trying to tell you how to think or what to say, sorry if I come across as too hard, I’m just genuinely curious if you have anything you consider pros of the new MacBook Pro, or if it’s all just negative? Honestly, I don’t really have many criticisms, none that come to mind at the moment. Obviously I don’t think Apple is infallible, but I also don’t feel the need to try to find something they’re doing to critique. 🤷🏼‍♂️. I think they generally offer great products that are better than the competition. Again, not saying your opinions aren’t valid (I disagree with the ones you’ve expressed so far), but you obviously have the right to think whatever you want about it.

About socketable LPDDR RAM…. A. It doesn’t exist today, so criticizing a product for not using it today isn’t valid. B. Who knows whether this “socketable” LPDDR RAM will be nearly as power efficient. Mounting the LPDDR RAM directly to the M chip package has benefits for energy efficiency, who knows how a compression socket would affect that. I doubt a compression socket would be mounted to the CPU package, as it would be difficult to access, so if the idea were upgradability, that wouldn’t be very easily upgradable. So they’d likely have to move it away from the CPU package if upgradability were the goal, and further distance equals less energy efficiency and increase in latency. And why should Apple switch back to user upgradable RAM when practically nobody upgrades their own RAM anymore. They’re going to incur extra production costs, potentially take energy efficiency and latency hits by moving the RAM away from the CPU package, all so that a couple hundred nerds can upgrade their RAM? I don’t think this would be a very reasonable expectation. It could be cool, but I highly doubt it would make any kind of business sense, and we must remember that Apple is a business, and they have to balance changes to production processes with the realistic improvements or gains of doing so. If there isn’t enough of a realistic gain from making a change, then it doesn’t make any business sense to incur the extra costs etc. anyways. Case in point, optical drives. Macs no longer have optical drives because they add to the bulk of the device, the cost of the device, and very few people would actually use them. It’s a similar situation with slotted user upgradable RAM, it would add some internal bulk, add some cost of production, and very few people would be likely to actually use it.

And it’s interesting to think of it alongside with the optical drive, because the last upgradable RAM model had an optical drive. Every MacBook for the last 11 years has had soldered RAM. So it would be an unrealistic regression. And it’s also a bit odd some people are complaining about it a decade later…
 
Last edited:
And it’s interesting to think of it alongside with the optical drive, because the last upgradable RAM model had an optical drive. Every MacBook for the last 11 years has had soldered RAM. So it would be an unrealistic regression. And it’s also a bit odd some people are complaining about it a decade later…
I think people are complaining about it a decade later because currently Apple isn't including enough at each price point. At least that's where I stand anyway. And the benefits of adding more RAM or storage down the road have not gone away.
 
It's a pretty safe bet that upgradable RAM isn't coming back on the MacBook market. They run too many memory channels to make it practical (even M3 Pro lineup runs a 192 bit bus, which would require three sticks of RAM if they used SO-DIMMS. The base M3 runs only two channels, but that's the chip that goes into the ultrathin ones that don't have internal space for it.)

Of course, Lenovo solves this problem in the their T-series Thinkpad market by just using two channels and having one of the sticks soldered to the board (asymmetric dual channel if you upgrade to a non-matching stick), but there is no way I could ever see Apple even considering that idea.

On the Mac Pro market (the one that starts at $7,000), I was a little disappointed (though not surprised) to see that they soldered it there too. They'd would have to (if I'm not mistaken) run 16 channels/16 sticks of RAM in order to get the kind of bandwidth they're running on the Ultra, but at the price these things are priced at (and with the internal space these things have), it COULD actually be done if they really wanted to. I'm not surprised Apple didn't go that route, but it'd make for a pretty inspiring machine if they did.
 
I think people are complaining about it a decade later because currently Apple isn't including enough at each price point. At least that's where I stand anyway. And the benefits of adding more RAM or storage down the road have not gone away.

And the small number of users who do legitimately need more ram know how to get it. They just don’t want to pay the extra premium for it, and here we are, more than 1400 comments later.
 
Well, more than a small number of Macbook Pro users need more than 8GB of RAM, and yeah, no one wants to buy something that they see as poor value.

Maybe it’s a “them” problem then. There are a ton of things around me that I feel cost too much (to me) and which I wouldn’t buy (but which tons of other people do), like cars and holiday trips, because I don’t see the value.

Well, I don’t drive and I don’t travel, and I for one don’t feel like I am entitled to any of them for cheap. I work hard and earn a decent salary and if there is something I really want, I pay for it and get on with my life. Life is just too short to be arguing over saving a couple of hundred dollars for 8gb of ram over a lifespan of 6-7 years?
 
I think people are complaining about it a decade later because currently Apple isn't including enough at each price point. At least that's where I stand anyway. And the benefits of adding more RAM or storage down the road have not gone away.
Nor have the benefits of using optical drives, optical drives are useful for several things. But most people don’t use them, just like most people don’t tear apart their laptop and upgrade their own RAM. The point is that after over a decade of MacBooks having RAM that is soldered, it doesn’t make any sense to expect that they’d regress back to upgradable RAM to satisfy a couple hundred nerds. I just don’t see this as being a reasonable expectation.
 
Maybe it’s a “them” problem then. There are a ton of things around me that I feel cost too much (to me) and which I wouldn’t buy (but which tons of other people do), like cars and holiday trips, because I don’t see the value.

Well, I don’t drive and I don’t travel, and I for one don’t feel like I am entitled to any of them for cheap. I work hard and earn a decent salary and if there is something I really want, I pay for it and get on with my life. Life is just too short to be arguing over saving a couple of hundred dollars for 8gb of ram over a lifespan of 6-7 years?
Yes, exactly. It’s literally still cheaper than the base model MacBook Pro was last year, even configuring it up to 16GB of RAM. I didn’t see this much whining about last years base-model MacBook Pro, I don’t see why people are complaining about this one which is cheaper… 🤷🏼‍♂️
 
Then, again, nobody’s forcing you to buy the configuration with 8GB of RAM… 🤷🏼‍♂️
Maybe it’s a “them” problem then. There are a ton of things around me that I feel cost too much (to me) and which I wouldn’t buy (but which tons of other people do), like cars and holiday trips, because I don’t see the value.

Well, I don’t drive and I don’t travel, and I for one don’t feel like I am entitled to any of them for cheap. I work hard and earn a decent salary and if there is something I really want, I pay for it and get on with my life. Life is just too short to be arguing over saving a couple of hundred dollars for 8gb of ram over a lifespan of 6-7 years?
Obviously nobody is being forced to buy anything and nothing in Apple's lineup is cheap. Not sure how those points add to this discussion at all.
 
Nor have the benefits of using optical drives, optical drives are useful for several things. But most people don’t use them, just like most people don’t tear apart their laptop and upgrade their own RAM. The point is that after over a decade of MacBooks having RAM that is soldered, it doesn’t make any sense to expect that they’d regress back to upgradable RAM to satisfy a couple hundred nerds. I just don’t see this as being a reasonable expectation.
Yes, I understand that. I don't expect upgradable RAM to come back to Macbooks either. Although it would certainly benefit more than a couple hundred people.
 
Yes, exactly. It’s literally still cheaper than the base model MacBook Pro was last year, even configuring it up to 16GB of RAM. I didn’t see this much whining about last years base-model MacBook Pro, I don’t see why people are complaining about this one which is cheaper… 🤷🏼‍♂️
So I guess there is a statute of limitations on computer talk or something like that? The internet let it go once so that's that?
 
Obviously nobody is being forced to buy anything and nothing in Apple's lineup is cheap. Not sure how those points add to this discussion at all.
People are complaining about the price, but if they don’t like the price and like the price on other things better, than they can just get that instead of dragging out a forum thread for over 1400 comments. It’s $200, and it’s still cheaper than the option that existed last year, so basically what this boils down to is some people want to complain about an option that’s the cheapest a 14” MacBook Pro has ever been is too expensive…. I think that’s just a little extra…. Here it’s cheaper, but people are freaking out and complaining about it being too expensive, or not good enough, or this that and the other thing.
 
So I guess there is a statute of limitations on computer talk or something like that? The internet let it go once so that's that?
The point is that people are creating an artificial controversy about this, even though it’s cheaper (even with 16GB of RAM) than it was last year. So the point is that most people (at least the content creators fueling this artificial scandal) were fine with the base price last year, in fact, many of these content creators were saying “it’s a great value”, and now that there’s a cheaper bridge option which starts with less RAM, suddenly it’s a big scandal, and content creators and other people need to trash it because they can’t edit their long 6K videos on it, or this, that, or the other thing…
 
Again... people asserting to know what their opponents believe better than they do. The thing about tech is that it's not fixed, so one can reasonably assume you'll get better machines for a comparatively lower price as time goes on. What the 'pro 8 GB' commenters here keep stressing is that we must continue to judge present specs and prices by the exact same standards of tech ~5 years ago. For the love of god stop telling us what we should be believing!

One of the most outrageous elements I had with the updated iMac is that its specs (beyond updated processor) is exactly the same as it was three years ago. No RAM boost, nor even the potential to upgrade it beyond 24 GB as it was exactly three years ago. Storage likewise hadn't been updated. Up until Apple starting building non upgradable machines they progressively added more storage and starting RAM without raising prices as a gratuity. The argument isn't so much that they're not forcing customers to buy more than they need, but that they're bottlenecking critical specs for the sole purpose of banking on upgrades or customers buying new computers much sooner.

When the retina MacBook of 2013 came out with a starting default of 8 GB Apple had moved heaven and Earth to keep the benchmarks as low as possible for the explicit purpose of gouging customers that aren't entirely sure if they need more. Storage can at least be mitigated but making the RAM non upgradable means you essentially have to buy the most advanced version of the computer up front because you know you won't have the option to upgrade later.

There is no artificial controversy. We know exactly what Apple is doing, and it's unfortunately working. But the simple fact is that it's literally impossible to argue against 16 GB being good enough when such people are arguing that 8 GB is good enough for most people. Apart from hindering Apple's profit margin on consumer notebooks, why should consumers be complaining about having too much memory? Considering Apple computers are essentially premium models, so shouldn't some of the basic and cheap features like memory and storage reflect that?
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Back
Top