Although impossible to test I wonder if the trade-off in speed los would be worth it to allow us to upgrade the RAm and storage ourselves again?
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.
Is 8GB of RAM Enough for MacBook Pro?
- Thread starter MacRumors
- Start date
- Sort by reaction score
The scandal isn't the lack of memory, but what Apple are charging for a product that is ridiculously cheap at wholesale prices. Lets say you sell 1000 MBPs but only 10% of those customers bother to pay the $200 for memory upgrade. Now lets say you charge $30 but all the customers did it because its a no-brainer.
Exactly, I think a lot of the people saying 8GB of RAM isn’t enough on the M-Series chips haven’t used an 8GB RAM M-Series chip. I’ve noticed a trend of referencing “my 16GB Intel machine struggles with x, or “my 16GB Intel machine doesn’t do z”, but in my experience with both, Intel machines aren’t nearly as snappy or efficient with their RAM usage, even when they have more RAM.
The MacBook Pro is ridiculously cheap at wholesale prices? How do you know if they’re not just barely eeking out a profit on the base configuration MacBook Pro? The last two years, they started at a more expensive price point, they dropped the base price point quite a bit by offering this option, so I don’t think we can assume they’re making this huge markup on it… High-resolution Mini-LED displays aren’t really that cheap. I wouldn’t expect the cost of manufacture has dropped THAT far…
I recently did comparison research on Windows PCs RAM upgrade pricing on models that use soldered RAM. Many competitors are similarly priced to Apple. Most I’ve seen are at least $100 for that upgrade, and a lot are either close to Apple’s pricing, or in some cases, even double what Apple charges. And these PCs don’t use it as Unified Memory, so it’s not as efficient and performative as Apple’s Unified Memory system, so I think there’s also value-added by that. Cases of double can be seen with Microsoft’s Surface products, which is actually a very interesting comparison since Microsoft positions its Surface lineup as a premium lineup similar to Apple. And when we look at RAM upgrades on Surface devices, they’re generally either the same, or higher. The Microsoft Surface Pro charges $400 to upgrade from 8GB to 16GB, and the Microsoft Studio Laptop charges $730 for their RAM upgrade from 16GB to 32GB, an upgrade that’s again about half the cost with Apple at $400 from 18GB to 36GB on the M3 Pro chip. So from everything I’ve seen with my research into Windows PCs that make use of soldered RAM, the RAM upgrade pricing tends to be fairly comparable, and in some cases more. Where you get all the cheaper RAM upgrades tend to be with the slot-based RAM, which is also slower and less power efficient.
PS. Very good points at the beginning of your comment as well, this is what I’ve been trying to explain to people in the course of the debate, whether or not an 8GB RAM option should exist or not is a totally different argument from whether RAM upgrades are too expensive or not. A lot of these articles and a lot of people in these arguments are calling for the elimination of the 8GB RAM configuration entirely because it supposedly doesn’t belong in a “pro” lineup of devices, which is a totally different argument from whether or not it’s too expensive to upgrade.
Last edited:
Very interesting! I'm glad you did that work of comparison in the PC world because I tried just about an hour ago and I'll have to say I quickly gave up. Hahaha. It's largely because I found the navigation and choices difficult enough that I wasn't going to spend globs of time figuring it out.
At work I've been issued a Lenovo ThinkPad something-or-the-other. I went to Lenovo website, chose something in that trim level and well, never got very far. For one thing, the base model seems to come with 16GB, but their ThinkBook base model comes with 8GB. And I couldn't figure out how to increase the GB factory installed (as opposed to buying it after factory purchase). Anyway.... (boy the PC world seems really confusing, to me!)
Wow. I feel like going from 8gb to 16gb at the cost of $200...seems so high! Interesting to know this is comparable in the PC world. I wonder what it actually costs Apple to do that. Maybe like less than $10?
I also don't really understand the concept behind Unified Memory Architecture but I have heard from other sources, as you have also noted, that it tends to be a more efficient way to use RAM, but the main downside is it can't be upgraded after the fact.
Ya, Windows PCs configurators tend to be rather confusing, and sneaky, as I’ve found with some they’ll automatically switch specs on some things when you upgrade the spec on one thing like RAM, which then, of course, will make it even more expensive. In all of the comparisons I did, I made sure that it didn’t automatically switch anything else, that way it would be only the pricing for the RAM upgrade, and be a fair comparison. And in Microsoft’s case, they at least conveniently show what their RAM upgrade pricing is right there in the button. But definitely took a while, and is not nearly as simple and intuitive as Apple’s configurator on their site.
Unified Memory acts as one single pool of memory that both the CPU and GPU have direct, low-latency access to, so it makes it much faster than traditional RAM systems which require data to be duplicated and shared between system resources rather than them having direct access to the same pool of memory. Because data doesn’t have to be duplicated in order to be shared between system resources, it makes the system significantly faster. That’s at least the way I understand it. I hope that explanation is helpful. 👍🏻
Pricing can potentially depend on a variety of factors.
For example, the cost to go from 8GB to 16GB memory on an HP laptop can be as little as $70, to go from 16GB to 32GB can be as little as $80. Sometimes increasing memory automatically includes a storage size increase or other upgrades.
Yeah, it might be true in the PC world that increasing RAM could also increase other features. This doesn't seem to be the case in the Mac world though. You generally can increase RAM and SSD independent of each other. And, generally it seems that the increase from one level to another level is the same fixed cost. So for example, going from 8GB to 16GB is $200; going from 16GB to 32GB is another $200; and it's also the case that going from 8GB to 32GB is $400.
The way Apple has generally done this is very clear and easy to understand, regardless of whether we agree with the dollar COST of the increase in RAM
LPDDR of the same (and faster) speed is available to order for products being released next year... Like you say, it's impossible to verify the exact percentages, but it seems like performance degradation should be near zero, but a small energy loss would occur. Whether it's small enough to be meaningful in any way is almost impossible to know. I'd take the risk, but I can't see Apple doing it for the very many reasons mentioned in this thread- it's a bit cheaper to solder it so a couple of dollars are saved, etc etc.
So for you using a RAM-limited box that is paging to disk gives you 0 issue. You do not care that the design competence is being limited due to less than ideal RAM, because (today 2023) Mac OS makes it all just work to a performance level that you find adequate.
Responding to your needs/wants and millions of others are why Apple keeps base RAM at cheapest available level. And when those RAM-limited boxes present as too slow in a couple of years Apple sells you a new Mac. Apple wins again. ;~)
Wrong. Placing RAM much closer and on disk is faster. Basic physics. And read up on UMA.
Edit: Note that speed of RAM is not so simplistic as just what some LPDDR vendor advertises. If processes are much farther away and have to go through additional controlling steps the net speed is much slower.
IMO it would not be, because fast UMA design and chip architecture are a synergistic whole. UMA is part of why M-series chips are so competent. And they present a different approach than the (hot) Intel products.
Ya, one of the main advantages of Unified Memory is that it’s a shared pool of high bandwidth memory. If you replaced it with slotted RAM, I think you’d take a substantial hit on the power efficiency, and performance would take a dramatic hit, at least using current slotted RAM.
Much slower? Nope. I think you're the one that needs to do some research. You've been drinking the Apple cool aid.
No Kool Aid. Imagine two runners going out of the starting blocks at the same time. Runner #1 opens the gate and runs off the track, runs a km on the streets, goes to the gymnasium and opens the door, runs around the gym, opens another door, runs out and back to the track, opens the gate again and proceeds to the finish line. Runner #2 runs straight to the finish line 100 meters away. Runner #1's net speed is much slower.
Physical proximity matters. Physics. Plus UMA means less doors/gates/controllers to slow things down.
There are downsides to UMA of course. The most obvious of which is it becomes prohibitively difficult to reach the very high RAM amounts in excess of 128 GB per chip that a few apps want to see. Which is why I am so curious to see what Apple does with Mac Pro and M3. IMO Apple now owns the laptop market from a performance standpoint, but Mac Pro so far remains meh.
Last edited:
Very good explanation of the difference between the two systems, I couldn’t have come up with a better analogy for it. And as an aside about the comment you’re responding to, the ad-hominems that many on the other side of this debate like to hurl at us aren’t very helpful for a profitable debate or discussion.
Last edited:
There's a trouble with that analogy, electricity flows a LOT faster than runners. Some 270K K/s, so the distance between one part of a laptop and another is so vanishingly small that it really can't be perceived. Yes, it's more than one signal, but we're talking extremely small bits of time! Usually the bottleneck on a computer isn't the RAM, the CPU itself is slower, then there's storage and I/O. I doubt even a benchmark would show any significant difference with the same speed RAM, and faster and slower RAM are only going to make a little difference.
But to use your runners analogy, and the distances RAM signals travel, your 2 runners are going to get just as far as each other in that amount of time. (read, they wouldn't have moved at all)
While I don’t necessarily agree with you, but I think I can fix your analogy.
Think of the CPU as a door with a time lock on it in front of the finish line. It doesn’t matter how fast or how far the runners go if the door is going to be locked in front of them at the finish line. The old lady with a walker is just as fast as long as she gets to the door just as it opens.
Except that part of the UMA is having less such doors with time locks. Apple's architecture does not have the same restrictions as the RAM-in-a-slot approach. Which is why I had Runner #1 opening all those extra doors/gates in addition to running more than 10x as far. Runner #1 may even be a faster runner, but will take longer to reach the finish line.
Last edited:
Of course electricity flows a LOT faster than runners. It was an analogy. My understanding is that Apple gets the UMA RAM so close on-chip to the coprocessors involved that the huge distance delta as compared to slotted RAM does matter. Plus slotted RAM requires additional controlling.
And you correctly state "Usually the bottleneck on a computer isn't the RAM, the CPU itself is slower, then there's storage and I/O." But Apple is not in usually mode with UMA. They built a different-operating chip with M-series, and that includes RAM utilization. Thinking along lines of what was is wrong-headed thinking IMO.
I agree that " faster and slower RAM are only going to make a little difference." My point was regarding the net speed of RAM access/use, which I probably confused by claiming "faster."
Last edited:
The issue is trying to use runners as analogy on how a computer operates is limited at best. Cixin Liu tried to do it, in agonizing detail, in his novel The Three-Bottle Problem. It didn’t work particularly well there and only has very limited application for trying to explain unified memory.
I’m just gripping about the analogy.