LPDDR6: New Beginning of AS: 12GB Standard

[Confused-User]

After non showing of any M4 in WWDC, I am pretty sure Apple is going all in with LPDDR6 in upcoming M4+ series by the end of the year; now supported by Ross Young. Clearly, @Confused-User is confused now cause he predicted Apple will launch Mac Studio before end of the year is delusional.

You don't get to call that delusional before, y'know, the actual end of the year.

I will similarly refrain from calling you delusional about Apple using N3B for the M4 in Macs (yeah, it's not going to be called "M4+") until they ship. Though it's a real effort. The notion that they are going to redo layout for all the IP blocks that should be a straight re-use from the M4 is bizarre.

If you did the math, you'll see that the M4 is ~ 13% larger. But a lot of that extra space is going to 2 more E cores and 2 more Thunderbolt controllers. And maybe even a third display controller. That makes sense as the vast majority of the chip area is using 2-1 FinFlex, meaning it's not much less dense than N3B.

Going back to N3B would sacrifice both power and clocks, which is even nuttier, as we can see that Apple has decided to push clocks somewhat aggressively (costing power, which tradeoff would likely become prohibitive on N3B).

Also, if you're so convinced they're going to support LPDDR6 this year, why wouldn't you assume that the new memory controllers on the shipping M4 *already* support it? There's plenty of precedent for that in other CPUs (as recently as a couple generations ago in Intel world, where one generation supported both DDR4 and DDR5).

BTW, which memory manufacturer do you think is going to be able to ship in such qty this year? And have they spoken about this coup in their last quarterly report?

 

[Chuckeee]

I will similarly refrain from calling you delusional about Apple using N3B for the M4 in Macs (yeah, it's not going to be called "M4+") until they ship. Though it's a real effort.

I’m impressed with your restraint. I avoided replying since I doubted my own limits

 

[tenthousandthings]

You don't get to call that delusional before, y'know, the actual end of the year.

I will similarly refrain from calling you delusional about Apple using N3B for the M4 in Macs (yeah, it's not going to be called "M4+") until they ship. Though it's a real effort. The notion that they are going to redo layout for all the IP blocks that should be a straight re-use from the M4 is bizarre.

If you did the math, you'll see that the M4 is ~ 13% larger. But a lot of that extra space is going to 2 more E cores and 2 more Thunderbolt controllers. And maybe even a third display controller. That makes sense as the vast majority of the chip area is using 2-1 FinFlex, meaning it's not much less dense than N3B.

Going back to N3B would sacrifice both power and clocks, which is even nuttier, as we can see that Apple has decided to push clocks somewhat aggressively (costing power, which tradeoff would likely become prohibitive on N3B).

Also, if you're so convinced they're going to support LPDDR6 this year, why wouldn't you assume that the new memory controllers on the shipping M4 *already* support it? There's plenty of precedent for that in other CPUs (as recently as a couple generations ago in Intel world, where one generation supported both DDR4 and DDR5).

BTW, which memory manufacturer do you think is going to be able to ship in such qty this year? And have they spoken about this coup in their last quarterly report?

Well, there is a chance Apple will change the process node for M4 Pro/Max/Ultra, but not backward to N3 (= “N3B,” although TSMC has stopped using that name), instead forward to N3P.

So A18, A18 Pro, and M4 would be on N3E, while the systems that require active cooling would be on N3P. The window for the start of N3P production opens in less than two weeks. Apple is likely to lead the way, as usual. [Edit to add additional NOTE: This scenario might support an earlier launch of the M4 Ultra (and thus Mac Studio/Pro), due to the lack of production pressure on N3P from iPhone/iPad, not to mention iMac and MacBook Air. I’ll keep my bet on March 2025, but you never know, Apple has proven repeatedly now that they are agile in exactly this way.]

What’s more, TSMC has said the opposite of the OP’s speculation: “We have also successfully delivered N3P technology. It has passed qualification and yield performance is close to N3E. [N3P] has also received product customer tape outs and will start on production in the second half of this year. Because of [PPA advantages] of N3P, we expect the majority of tape outs on N3 to go to N3P.” [Anandtech]

Nonetheless, M3 will get a moment in the sun, as it were, when the betas for Sequoia reveal it is on par with A17 Pro in terms of Apple Intelligence performance, and M3 Pro/Max are even better.

 

[Confused-User]

Well, there is a chance Apple will change the process node for M4 Pro/Max/Ultra, but not backward to N3 (= “N3B,” although TSMC has stopped using that name), instead forward to N3P.

Sure. I've mentioned that before. It's very likely *if* they don't bring out M4 Macs until the fall. I hope it's earlier, but every day that goes by without M4 Macs makes that more likely.

It's not very interesting from the perspective of the discussion we were having, because N3E -> N3P is easy. You don't have to redo layout. That's very different from N3B -> N3E, where you do.

Despite TSMC's preference, I continue to write N3B instead of N3 because it's the only concise way to avoid ambiguity.

 

[MayaUser]

fast forward some months...still no 12 gb ram standard
Try again next year

 

[Populus]

fast forward some months...still no 12 gb ram standard
Try again next year

I thought the thread was about the faster LPDDR6 memory, which could bring its own benefits regardless of the memory size.

12GB of RAM wasn’t expected for this years iPhone, and probably next years’ iPhone 17 will remain with 8GB, with the exception being maybe the 17 Pro Max. Hopefully the iPhone 18 lineup will finally come with 12GB of base RAM, although maybe for 2026 we’re already dreaming about iPhones with 16GB of RAM.

I think some of us are a little obsessed about RAM lol

 

[hovscorpion12]

As far as Macs, LPDDR6 could be coming with the M4 Pro/M4 Max. The 6500 MT/s speeds are now a joke compared to the 8000 MT/s on existing computers. We're definitely going to be eating good next month.

As far as 12GB of RAM, that is a rumor. There is no proof that 12 GB will be the new base RAM going forward. The tear down of the M4 iPad pro 8GB model does show Apple is shipping 12GB of RAM, however on the outside it is only 8GB. As far as consumers are concerned. They're buying 8GB RAM models.

The obsession of RAM is insane. Those who are angry at Apple's pricing are proof of the Apple walled garden effect. Apple is not the only manufacture of ARM computers. Shop around.

 

[Populus]

As far as Macs, LPDDR6 could be coming with the M4 Pro/M4 Max. The 6500 MT/s speeds are now a joke compared to the 8000 MT/s on existing computers. We're definitely going to be eating good next month.

As far as 12GB of RAM, that is a rumor. There is no proof that 12 GB will be the new base RAM going forward. The tear down of the M4 iPad pro 8GB model does show Apple is shipping 12GB of RAM, however on the outside it is only 8GB. As far as consumers are concerned. They're buying 8GB RAM models.

The obsession of RAM is insane. Those who are angry at Apple's pricing are proof of the Apple walled garden effect. Apple is not the only manufacture of ARM computers. Shop around.

I would really happy if Apple shipped M4 Pro Mac mini with faster LPDDR6 RAM, however, if that’s not the case, then maybe I should wait until the M5 Mac minis? I think It would add to the longevity of the computer…

 

[hovscorpion12]

I would really happy if Apple shipped M4 Pro Mac mini with faster LPDDR6 RAM, however, if that’s not the case, then maybe I should wait until the M5 Mac minis? I think It would add to the longevity of the computer…

Based on Teardowns.

- The 2023 M3 Pro/M3 Max ships with LPDDR5 RAM.
- The M4 iPad Pro ships with LPDDR5X.

Obviously, one is a tablet where the other is a computer. We have a 50/50 chance that the 2024 M4 Pro/M4 Max ships with either LPDDR5X or LPDDR6.

 

[MayaUser]

For macs LPDDR6 Ram is more important than base Ram because you can upgrade to 16gb ram on the mac...
But for the iphones you cannot no matter what...even in the ipads pro for some time you can by going higher storage

But i dont have high hopes for lpddr6 in the next macs next months, but i was surprise that Apple brought wifi 7 to their devices so early. So that "lpddr6 base 12gb ram new beginning" probably will happen next year

 

[theorist9]

I was wondering if they could use LPDDR6 with the M4 architecture, given that the M4 iPad uses LPDDR5X. But then I remembered that the M1 was initially released on the Mini with LPDDR4X, but that the M1 Pro/Max/Ultra featured LPDDR5.

This tells us that each generation of M-series chip is not necessarily restricted to a single generation of RAM. Hence it's at least possible the M4 Macs could feature LPDDR6.

 

[Pressure]

Based on Teardowns.

- The 2023 M3 Pro/M3 Max ships with LPDDR5 RAM.
- The M4 iPad Pro ships with LPDDR5X.

Obviously, one is a tablet where the other is a computer. We have a 50/50 chance that the 2024 M4 Pro/M4 Max ships with either LPDDR5X or LPDDR6.

The M4 iPad Pro's are just using LPDDR5.

The RAM chips clearly show the 5 digit FBGA code on them (D8DNV and Z8DMS).

Insert those in the Micron FBGA lookup table here and you get the following part numbers:

5 digit FBGA code	Part number	Memory type	Density (Gbit)
D8DNV	MT62F1G64D4AS-026 XT:C	LPDDR5	64 Gbit (8GB)
Z8DMS	MT62F768M64D4AS-026 XT:B	LPDDR5	48 Gbit (6GB)

 

[hovscorpion12]

The M4 iPad Pro's are just using LPDDR5.

The RAM chips clearly show the 5 digit FBGA code on them (D8DNV and Z8DMS).

Insert those in the Micron FBGA lookup table here and you get the following part numbers:

5 digit FBGA code	Part number	Memory type	Density (Gbit)
D8DNV	MT62F1G64D4AS-026 XT:C	LPDDR5	64 Gbit (8GB)
Z8DMS	MT62F768M64D4AS-026 XT:B	LPDDR5	48 Gbit (6GB)

Incorrect. The M4 iPad Pro uses LDDR5X RAM. This information comes directly for iFixit

iPad Pro 13" Chip ID

This guide contains a selection of photos and chip identification summary for the iPad Pro 13. Check out our teardown here. Special thanks to...

www.ifixit.com

Apple uses Micron MT62F768M64D4AS-026 XT:B 6 GB LPDDR5X SDRAM memory

 

[Pressure]

Incorrect. The M4 iPad Pro uses LDDR5X RAM. This information comes directly for iFixit

iPad Pro 13" Chip ID

This guide contains a selection of photos and chip identification summary for the iPad Pro 13. Check out our teardown here. Special thanks to...

www.ifixit.com

Apple uses Micron MT62F768M64D4AS-026 XT:B 6 GB LPDDR5X SDRAM memory

The image I posted is directly from the iFixit teardown 🧐

It clearly says Z8DMS as the FBGA 5 digit code on the memory chips.

The part number for Z8DMS is MT62F768M64D4AS-026 XT:B, which is what iFixit clearly has right but that is LPDDR5-7500 and not LPDDR5X.

 

[DaniTheFox]

With this settled, I predict the M4 Mac will come with 2x 6GB base memory (the question is still why the iPad Pro can only use 8GB out of this).
So, back to the original question. The use of LPDDR6 does not give us 12GB base memory. It is the decision of Apple to use 6GB LPDDR5 chips that does.

 

[EugW]

With this settled, I predict the M4 Mac will come with 2x 6GB base memory (the question is still why the iPad Pro can only use 8GB out of this).
So, back to the original question. The use of LPDDR6 does not give us 12GB base memory. It is the decision of Apple to use 6GB LPDDR5 chips that does.

I too had predicted 12 GB RAM for the new Mac mini and MacBook Air based on the iPad Pro teardown, but the pundits are predicting 16 GB actually.

However, I don’t expect LPDDR6.

 

[Confused-User]

I am a little embarrassed to admit I enjoyed collecting the quotes for this post. It's an immature emotional reaction to another person's incredibly smug ignorance. Oh well, I accept my flaws.

The amazing thing is that he kept on talking about this for months. Even after the iPad M4 shipped. Crazy.

Question: Have you wonder why Apple rumored to launch M4 series by the end of the year? This round Apple will launch M4 from top to bottom including Mac Mini within a quarter period, why?

Mark Gurman can't tell you, but I can. The key ingredient missing is LPDDR6.

It's still missing.

This is extremely unlikely. The standard won't be released until Q3. It would be extremely atypical for Apple to support both LPDDR5 and LPDDR6 with its next memory controller, which normally they'd implement once and then use everywhere (Ax/Mx chips of the same generation).

The standard may not even be out until well into 2025.

Let's say upcoming iPhone 16 Pro series will occupy 40% of total volume; that's mean iPhone 16 Pro will need 20 million memory chips per quarter. Yes, that's a lot compared to 8G4 which I estimated will need 3 million memory chip per quarter. Yes, iPhone volume is crazy huge but if Samsung has been making LPDDR6 since 2023, I think they will provide enough volume for Apple.

Maybe I should have posted this after the iPhone 16 came out. But really, his main point was about the M4, so I waited.

Thanks for detailed lineup, it's time for my analysis with updated information. As I am more convinced that 12GB LPDDR6 will be standard in upcoming M4 SoC.

Fail.

FYI, upcoming Blackhawk's IPC/PPC improvement is true. It is biggest bump of ARM Cortex X series. Maybe it is due to inclusion of SME, but the threat to Apple Silicon is real.

Hahaha. It can't even get to M3 levels, with dramatically higher power draw. (Though they did get closer than I expected.) Same for the QC SD8g4.

And yes, please use your brain to think why Apple will continue using N3B for M4 Max with more cores, NPU, Ultra-fusion connector and so on. Definitely upcoming M4 Max will break 100 billion transistors even though the amount of SLC and memory bus will be cut.

@Confused-User is confused and delusional about Mac lineup. I have listed some evidence in the page before, yet he still insists I am delusional. Here is my speculation of iPhone and Mac lineup by the end of year:

1. JEDEC going to announce LPDDR6 standard before iPhone event.
2. iPhone 16 Pro will feature 12GB LPDDR6.
3. By Q4 of the year, Apple will announce M4+ family (or whatever name Apple marketing feels suitable) with starting RAM of 12GB LPDDR6.

@Confused-User If you still think I am delusional, why not you write down your prediction here so that we can compare results.

A very faily fail. He got *everything* wrong. Well, we haven't seen the Max yet, but I'll be happy to come back here tomorrow and eat crow if that's on N3B.

After non showing of any M4 in WWDC, I am pretty sure Apple is going all in with LPDDR6 in upcoming M4+ series by the end of the year; now supported by Ross Young. Clearly, @Confused-User is confused now cause he predicted Apple will launch Mac Studio before end of the year is delusional.

Still with the LPDDR6 and M4+. Yes, it looks like I was wrong about the Studio; the difference between us is I wasn't handing down wisdom to the unwashed. I made a cautious prediction, making clear that there was plenty of uncertainty.

You don't get to call that delusional before, y'know, the actual end of the year.

I will similarly refrain from calling you delusional about Apple using N3B for the M4 in Macs (yeah, it's not going to be called "M4+") until they ship. Though it's a real effort. The notion that they are going to redo layout for all the IP blocks that should be a straight re-use from the M4 is bizarre.

If you did the math, you'll see that the M4 is ~ 13% larger. But a lot of that extra space is going to 2 more E cores and 2 more Thunderbolt controllers. And maybe even a third display controller. That makes sense as the vast majority of the chip area is using 2-1 FinFlex, meaning it's not much less dense than N3B.

I was so right about this. Even about the display controller, as it turns out. Apple spent some area putting in a third controller (M1-M3 have two).

So, @TigeRick, are you ready to own your mistake, and perhaps show more humility in the future?

 

[Confused-User]

So, @TigeRick, are you ready to own your mistake, and perhaps show more humility in the future?

In case it helps, the M4 Pro is here, and it too uses LPDDR5 (presumably 5X) at ~8533MT.

Come on, you know you can do it!

 

[Populus]

The memory bandwidth is 273GB/s on the M4 Pro vs the 120GB/s on the M4 which already uses LPDDR5X if I recall correctly.

Can memory bandwidth be more than doubled just by raising the clock speed? I’m just asking, not affirming anything, because that jump in mem bandwidth looks pretty big to me.

I will only read respectful replies, thank you.

 

[leman]

The memory bandwidth is 273GB/s on the M4 Pro vs the 120GB/s on the M4 which already uses LPDDR5X if I recall correctly.

Can memory bandwidth be more than doubled just by raising the clock speed? I’m just asking, not affirming anything, because that jump in mem bandwidth looks pretty big to me.

I will only read respectful replies, thank you.

It is consistent with a 256-bit bus and newer LPDDR5X-8533. The RAM itself is 13% faster and the wider bus would double the throughput.

I also hit confused first, because that looks like quite a jump, but it makes sense.

 

[Confused-User]

The memory bandwidth is 273GB/s on the M4 Pro vs the 120GB/s on the M4 which already uses LPDDR5X if I recall correctly.

Can memory bandwidth be more than doubled just by raising the clock speed? I’m just asking, not affirming anything, because that jump in mem bandwidth looks pretty big to me.

I will only read respectful replies, thank you.

@leman's answer is correct. The direct answer to your question is "no", you can't do it just by raising clocks. Nor would you want to, as your imaginary silicon would have a very very high power budget...

The Pro has double the bus size of the plain M4, just like in the M1 and M2 generations. That gets you to 240GBps; the 13% clock bump gets you the rest of the way.

BTW, I don't remember the M4 using 5X, just 5, but I may be mistaken and that doesn't matter either way.

 

[Populus]

@leman's answer is correct. The direct answer to your question is "no", you can't do it just by raising clocks. Nor would you want to, as your imaginary silicon would have a very very high power budget...

The Pro has double the bus size of the plain M4, just like in the M1 and M2 generations. That gets you to 240GBps; the 13% clock bump gets you the rest of the way.

BTW, I don't remember the M4 using 5X, just 5, but I may be mistaken and that doesn't matter either way.

Yeah, that makes sense, thank you.

It is consistent with a 256-bit bus and newer LPDDR5X-8533. The RAM itself is 13% faster and the wider bus would double the throughput.

I also hit confused first, because that looks like quite a jump, but it makes sense.

Yeah, that’s a logical explanation.

 

[altaic]

@leman's answer is correct. The direct answer to your question is "no", you can't do it just by raising clocks. Nor would you want to, as your imaginary silicon would have a very very high power budget...

According to Wikipedia, they could have gone all the way up to LPDDR5X-10700 (where LPDDR6 will start). But yeah, all the power data for this stuff is confidential, and who knows what sort of production volume those parts are at. I’m happy with LPDDR5X-8533 for the Mini, but I’m hoping for 9600 or 10700 for the Ultra.

 

[Boil]

...9600 or 10700 for the Ultra.

9600 for the Ultra, 10700 for the Extreme...! ;^p

 

[Confused-User]

So, @TigeRick, are you ready to own your mistake, and perhaps show more humility in the future?

<crickets>

Yeah, that's what I thought.