> In a universe where you want high-performance energy efficient RAM [.....] You simply can’t have this with socketed RAM.

You make it sound like a prerequisite. I'm pretty sure the reason LPDDR is Low Power is not due to (currently) being unavailable in socketed modules - that would seem to be a side-effect of the original target market (smart phones).

They are just some BGA packages, I see no reason they couldn't be soldered onto a board with a socket. If you scan through the features that make LPDDR LP, they have nothing to do with sockets, or anything affecting sockets: lower operating voltage, more power states, partial refreshing etc:

https://en.wikipedia.org/wiki/LPDDR

LPDDR is low power because it is a more sophisticated device with more complex protocol compared to regular DDR, because it has a higher pinout allowing it to have a more efficient electrical connection and because the computer manufacturer can optimize the connection with specific chips in mind instead of having to support a wide range of devices.

I think the number of pins is a practical limitation. Current DDR DIMMs use 260-280 pins if I remember correctly, while an average LPDDR4X BGA package is 400 balls or higher. Apple uses chips with 877 balls. It is definitely possible to design modular RAM with high-pinout, but would it be economically feasible?

You can't compare balls on the BGA package to signal pins. The vast majority of these are for power and ground. LPDDR4 actually has fewer data lines than DDR4.

It is if it's a standard that everyone decides to support...

You can fit FAR more signals on the bottom of a chip and in a multi-layer PCB than you can on any socket you might think of. There is a practical limit to how small the pitch of the pins on a socket can be if it's going to be field-replaceable; at high pin counts, the length of the socket itself becomes a driving factor.

As an example, even moving away from edge-connectors (which are just a non-starter for 800+ pins) Samtec's highest-density high-speed connector, the NovaRay line [0] advertises 224 signals per square inch, so an 800 pin module would require 4 square inches of connector space alone, not to mention making the system PCB assembly thicker and compromising on speed, power and thermals.

[0] https://www.samtec.com/connectors/high-speed-board-to-board/...

So design it like a CPU socket? Laptops usually don't have socketed CPUs because they also need to mount a heatsink, but RAM doesn't have that issue.

So an FM2 socket (to pick one of the more reliable ones), which has ~900 pins has a 4.5in^2 area and adds ~.25" in thickness, while reducing the maximum speed you can get through the chip at the very low voltages necessary to run LPDDR, which is what I meant by compromising speed and power.

Not to mention that socketting parts reduces your production yield over SMT, increases your assembly cost, and cuts reliability (it's fine on a desktop because you almost never move your desktop, and there's usually a big thermal assembly bolted on top).

Pentium m used to be socketed, pretty reliable, and available for thin laptops. I'm not trying to dismiss the difficulty, just saying they could perhaps be overcome

Hardware engineering is ever thus: Difficulties can always be overcome, but every cost adds 5x that cost to the MSRP.

That's the thing, who is everyone? Such technology would only be relevant to premium-level ultraportable notebooks. Desktops, gaming laptops as well as budget laptops would stick with the cheaper and more ubiquitous DDR. Smartphones and tables will continue to solder RAM on due to space constraints. Apple (who is the driving force in this market segment) won't be interested since they are going fully custom and I doubt that an LGA socket will work with their on-package custom-build wide RAM interfaces. So in the end we are left with things like Dell XPS 13, MS Surface and few others. These are all popular and impactful brands, but they are just a drop in the bucket compared to the total PC sales. And I doubt that technology reserved to only selected premium laptops is going to be cheap.

If LPDDR is faster won’t desktops and gaming laptops also want it?

One possible other compromise could be soldered LPDDR RAM, but with a DIMM expansion slot for DDR as well. I wonder whether the DDR could be switched off and the baseline power draw eliminated if the RAM usage was below needing it.

So I will preface this with that I'm not an expert. But as I understand it, multicore systems are already frequently NUMA, but to a degree that most stuff (I am eliding categories of code here, but you get the idea) don't have to really care. From my understanding of how all this works, the difference of speed and bandwidth for LPDDR versus DDR seems like risking significant complexity for an expandability that few people would use and in turn cause an everything-or-nothing problem, where either high-performance code is written assuming anybody might have this, or nobody assumes it and you get a pile of stalls.

> If LPDDR is faster won’t desktops and gaming laptops also want it?

The desktop version of LPDDR4 is there, it's just called DDR5.

We have cpus with lots of pins, nobody is stopping ram from switching to a cpu socket layout..

Laptop CPU is onboard for the same reasons RAM is. This is not about the amount of pins, it's about power requirements for signal clarity.

I've never done anything beyond amateur electronics, so sorry if I'm way off base here, but it at least seems intuitively plausible to me that soldering the components to the board allows them to work reliably at lower power. With a pin and socket connection, you've got to try and be at least reasonably resistant to corrosion or other sources of less-than-perfect connection.

That said, I don't know that RAM chips are the hill I want to die on, anyway. Could we get to a better place if we dispensed with the whole, "just the one big motherboard," design, and looked to the old backplane-style physical layouts? Then you could have a module that contains the CPU and RAM and suchlike all in a single package, and move other things that may not want to be quite as tightly integrated off to another package.

Up until relatively recently, a lot of laptops and even phones would put the connectors on separate PCBs connected by a ribbon, so that damaging those wouldn't mean a $500 motherboard replacement.

I definitely remember ordering a replacement headphone socket board for an old MacBook and just putting it in myself, and I've done similar with, eg, the connector port board on a PS4 controller.

But I guess this is one of the hazards of Apple having their AppleCare programme— they can do the math and calculate that saving a dollar on the cost of making and mounting and connecting that extra component still makes them money even if they have to eat the occasional full motherboard in the first three years of ownership.

Not just for the easier replacement, but also to avoid translating the mechanical stresses directly to the motherboard. The separate PCB can mount in a way that lets it move a little.

If I am not mistaken the ports on recent Apple machines are on a separate, easily replaceable board.

Seems like this stuff maybe comes and goes— like here's a repair guide for a 2010-vintage MBP which definitely shows all the connectors being right on the logic board:

https://www.ifixit.com/Guide/MacBook+Pro+15-Inch+Logic+Board...

Doesn’t appear to be the case on the 16” MBP: https://www.ifixit.com/Teardown/MacBook+Pro+16-Inch+2019+Tea...

Looking at this photo from that page it appears that they are: https://guide-images.cdn.ifixit.com/igi/uavG2A3vFrVmwv6J.hug...

You can see the ports still in place after the motherboard has been removed.

Interesting, didn’t notice that. It looked like the logic board went all the way to the edges.

Part of it comes down to signal drive strength.

If you design to a socketed DIMM package, you actually have two trace domains, the CPU to the socket, and the socket to the RAM chips. For that to be reliable, you have to adhere to standards for DIMM (or SODIMM) so that multiple manufacturers can build compliant interchangeable parts.

If you solder the chips down to the main board directly, you only have once domain for the traces, CPU directly to RAM. At that point you can tweak the design for several goals, speed, power, etc.. And you can tune the drive strength on the signalling the same.

That is how you end up with more efficient soldered designs over socketed designs.

This sounds conspicuously like an Intel NUC, I think?

Yes, as I wrote somewhere below, LPDDR4 signal is actually more robust than DDR4 signal, and LPDDR4 controllers have more features to fix degraded signals.

DDR5 has more roots in the LPDDR4 than in the regular DDR4.

ribit and others are correct. While this may have been possible in the past, at modern memory frequencies and timings, it's not really plausible to deviate from what the platforms are designed for: LPDDR being soldered and placed a specific layout area relative to the CPU, and for socketed DDR, the sockets being placed in one of a few placements relative to the CPU, with a little layout wiggle room.

Edit: I guess to clarify, the plausibility challenge is that whether or not it is theoretically possible, the platform owner (Intel, AMD, etc) or the memory vendors (Samsung, SK Hynix, Micron) are unlikely to support the lark.

There’s no technical reason not to (I’m no EE but I’m sure it could be done somehow), but there’s no demand for removable memory. The entire consumer PC market is already a rounding error for the semiconductor industry, and the enthusiast market that would be interested in this is a very small portion of that market.

Like it or not, repairable/upgradable laptops are a thing of the past. There’s just not component support or sufficient market demand to make the parts removable anymore.

Way overstating your case. PC's might be on the decline but they are still far far from being a "rounding error for the semiconductor industry".

Laptops also make up a major component of that still very alive market so no, I disagree that there is not enough market demand for having laptops that can be repaired or upgraded. It's the industry themselves that have worked to hamper the right of repair while giving nebulous pseudo-technical rationalizations for it.

Modern laptop motherboards can easily be about the size of a cell phone. Most of the case is battery and the main board is a shrinking percentage of the BOM cost. The entire motherboard is the replaceable component. It’s less “right to repair” than the additional components and design required to make them repairable are more expensive than just replacing the whole board.

Seems like a deliberate design choice, not some inevitable consequence of modernity. Why exactly does the board need to be cell phone sized when the actual form factor of even a netbook is substantially larger?

To fit more battery. Other than my Mac mini, every computing device (including watch and phone) I own is mostly battery inside.

Which you could also do by simply making the device a little thicker and have a battery run the entire width of the device. Point is past a certain point, thinness becomes more of a drawback than a benefit, especially given the trade-offs involved.

You might want that trade off, but consumers in general have been leaning towards thinner and lighter for years. Just look around at complaints about the thickness of early android watches being too big and chunky. Or the huge popularity when the first MacBook Air came out.

Given the option between thicker w/ a bit more battery life or thinner and lighter, consumers are choosing thinner and lighter.

Of course engineering is always a trade off, and it is possible to go too far in one direction. But it's not all or nothing, Apple is the one feeding us that false dichotomy.

> ...consumers in general have been leaning towards thinner and lighter for years.

> ...consumers are choosing thinner and lighter.

A) I'm guessing actually lighter more than thinner.

B) Are consumers really clamouring all that intensely for "Thinner! Thinner!", or is that a) because that's all that's offered, and b) the relentless hype from manufacturers -- above all, bendy phone maker Apple -- amplified by rarely-critical media?

B. Definitely B. I’d happily choose a phone that’s chonkier if it was an option.

Battery is a big part of mobile devices of all kind, but it's not “mostly battery” for most of those.

For instance, in a iPhone 11, the battery occupies less than half of the surface of the phone[1]. On a Macbook pro, it's about the same.[2] [1] see this ifixit video if you want to see it by yourself: https://www.youtube.com/watch?v=Feyqwf3cYtw&t=161s

[2] https://guide-images.cdn.ifixit.com/igi/noPvQ2CLs1jptX2p.hug... You can see that the motherboard is much, much bigger than a smartphone.

Looking at the picture from [2]. it looks like the motherboard is really not much much bigger than a smartphone. There's a lot of space dedicated to ventilation (note the two large fans and the black cooling ducts under the hand in the picture. Some fraction of that motherboard is also the solid state storage. The picture here of the internals of an iPhone 11 https://www.zdnet.com/article/heres-whats-inside-your-new-ip... (I don't do videos) sure looks like it's mostly battery. If I discount the empty space, the insides are more battery than not battery.

It is mostly only a deliberate design choice in the sense that consumers have leaned towards designs that are smaller, lighter, and with longer battery lives.

It doesn't matter that the rest of the form factor is larger: there's not a lot of empty space in these, so a larger main board will still mean a larger and heavier device.

I don't know what kind of cell phone you have, but it's not an usual one: here is a picture of an open modern thinkpad https://gearopen.com/wp-content/uploads/2019/12/internals-15...

That’s a laptop from 2 years ago; the new M1 MacBooks are more indicative of where things are headed in the future. Things like memory and any custom ICs are just going to be integrated on the same silicon as the CPU in most cases.

Here you go: https://eshop.macsales.com/blog/wp-content/uploads/2020/11/M...

Still bigger than a smartphone (it occupies the entire width of the laptop (when counting the heatsink, 2/3 of the width otherwise), and around one quarter of the whole area).

Also, your original sentence was

> Modern laptop motherboards can easily be about the size of a cell phone.

Here we see that the one and only best laptop in this regard comes close to the cell-phone-sized motherboard. That's far from what “modern laptop can easily be” means. Most laptops sold today have a much bigger motherboard than this.

(BTW, as much as I'd like to see non x86 laptops going mainstream, software compatibility is huge issue for an ecosystem as diverse as the Windows one (and Roseta 2 being too much of a coinflip isn't encouraging…), we might get there at some point, put it's gonna be slow and painful.)

So just...put LPDDR on a DIMM? Or some other kind of removable slot?? I see no technical reason you couldn't.

Or, instead of soldering chips, make little LGA slots for the individual chips. I can't imagine the tiny added resistance from an extra millimeter of wire and pin surface contact would impact much of anything and even if it does, I'd rather take that than have no options when the next Windows update makes my machine a paperweight (obviously exaggerating, but it's getting there).

The reason is signal integrity. Running a high speed signal (never mind a bus of 64+ of them) through a connector is hard and takes a lot of power because the connector can seriously degrade the signals. LPDDR can be low power because it doesn't have a "memory stick" abstraction, just memory chips and controllers. It's not the resistance that kills the signal, it's the inductance and the impedance mismatching.

LPDDR3 DIMMs exist. I assume everyone is talking about LPDDR4, which seems to use 0.1V less.

But the majority of people couldn't give a toss about how much of their laptop ends up in a landfill. They're using paper/glass straws, you see.

Perhaps electronics landfills should be local again.

I think you're referring to DDR3L DIMMs, which are just a down-volted version of DDR3. LPDDR has never had DIMMs because of signal integrity issues. They take power to resolve.

For something running at 5-6W while idle, this is not a small difference.

> Modularity is not a free lunch

Hold on. This is true for excessive modularity.

Yet, most of the time, lack of modularity (e.g. bundling) creates lock-in reducing customer's choice and ultimately increasing prices.

When PCs became standardized ( https://en.wikipedia.org/wiki/IBM_PC_compatible ) prices dropped incredibly.

I would really prefer that a laptop was just a tiny bit thicker to fit a 100WH battery to compensate for that. My T480 gets 12-16 hours with normal use, I imagine it could easily hit 20 with current processors.

High-end 15" laptops already have 100Wh batteries, and 100Wh is the practical maximum. So abandoning LPDDR would lead to a reduction in battery life on the high end, which probably isn't going to fly.

> isn't going to fly.

Literally - 100Wh batteries are maximum you can take on planes

In US, on non international flights, you can take 160wh ones if they are replaceable.

Strangely, the rule comes from ICAO.

Some countries too follow the 160wh standard, but most airlines simply don't bother enforcing anything, but the lowest common denominator.

I promise that was an unintentional pun. But yeah, that's the limiting factor.

> a 100WH battery

a hand grenade typically releases 250kJ, or about 70Wh

you're asking for the equivalent of 1.5 grenades in your lap :)

A chocolate tablet typically stores 4000kJ of energy. Yes that's 15 hand grenades, what are we waiting for: ban chocolate tablets now!

And don't get me started on flour, not only it stores a ton of energy (50 grenades per kg) but it's also highly explosive! [1]

[1]: (only with the right air to flour ratio of course, but accident happen every once in a while: https://www.grupa-wolff.com/a-tragic-flour-explosion/)

The chocolate bar doesn't start a violent fire if pierced. If it would we could power our laptops with chocolate :)

The flour conditions of exploding are far more difficult to achieve than a LiIon battery conditions - orders of magnitude difference.

> The chocolate bar doesn't start a violent fire if pierced. If it would we could power our laptops with chocolate :)

Neither does a NiMH battery. There's no link between vulnerability to external damage and ability to power a laptop.

In fact, there's no fundamental reason why we couldn't run our laptop on chocolate[1], just that it's very far away from current technology.

[1]: http://large.stanford.edu/courses/2014/ph240/ho2/

I read Lion battery (lion as in the chocolate bar) and couldn't help but find it hilarious

It’s just that modern laptops run on Li-ion batteries.

True, but not that hard. I had some “dangerous experiments for kids” book back in the day and by far the most fun thing in it was aerosolizing flour with a straw to get the right mixture near your Bunsen burner for your own glorious short lived fireball.

In a world, somewhere, Flowers and Lions power laptops.

1.5 hand granades worth of LOW VOLTAGE ELECTRIC energy released into A CIRCUIT over A DAY or more and protected by multiple safety cut-off circuits with thermal and other sensors.

vs

A grenade releasing that as KINETIC energy into YOUR FACE over the span of LESS THAN A MILLISECOND and protected by a spring and a keychain ring.

Even if a lithium battery of that size experienced a critical failure right in your lap, you'd get away with a few scars and some serious but treatable burns. And I've only heard of one commercial product in the last 15 years that had serial critical battery failures...

Airlines tend to cap lithium-ion batteries at 100Wh, so it’s quite common for laptops and power banks to go to or just shy of that. https://www.apple.com/newsroom/2019/11/apple-introduces-16-i... speaks of a 100Wh battery. My latest laptop has about a 90Wh battery.

Hopefully your laptop and hand grenade have radically different power discharge profiles.

> Hopefully your laptop and hand grenade have radically different power discharge profiles

Li battery fires are not unheard of, unfortunately, so the profiles may start to resemble sometimes.

Lithium batteries burn, quite quickly but it still takes a few seconds, that can be pretty dangerous indeed but that's several orders of magnitude slower than a grenade.

There's a subtle difference. One is designed to release that energy in a fraction of a second in a frangible metal casing, the other is designed to release it over a number of hours. Yes, batteries can go wrong, just like any tech, but comparing them to hand grenades is just ludicrous.

> comparing them to hand grenades is just ludicrous

Next, let's calculate the energy input required to heat a small cup of coffee and compare that to a hand grenade. (Hint: The coffee wins easily.) I wonder how many hand grenade equivalents a 10 minute lukewarm shower equates to?

In some universe there's an internal combustion shower powered entirely by an operator ripping out pull rings with their teeth and throwing armed grenades into a heating chamber.

Preferably yelling 'fire in the hole!' each time.

To keep your peace of mind, refrain from looking up the energy content of a pizza!

Is it any worse than having 0.75 of a hand grenade on your lap? If it starts to smoke, you throw it away all the same. Besides, this capacity is standard in larger laptops, but not in smaller ones for some strange reason.

> If it starts to smoke, you throw it away all the same.

No. If it starts to smoke, you commit & push --force, then you throw it away.

You can save a couple seconds with git-fire.

https://github.com/qw3rtman/git-fire

The hand grenade is arguably better at releasing that energy at an instant.

Please forgive the hi-jack...

FWIW, the book Design Rules: The Power of Modularity systematically explores the tradeoffs.

https://www.amazon.com/Design-Rules-Vol-Power-Modularity/dp/...

Written by two economists, it's the only prescriptive definition of architecture that I've managed to find. Influenced me deeply.

So why are you holding back on us? :) What are the main trade-offs?

Heh. Just what you'd guess. Interfaces allow separation of concerns, parallel dev efforts, competition. Integration allows optimization, at greater risk.

Their core contribution is modeling the options with net present value (NPV) to remove the guess work.

--

Tangentially...

They also explain the dependency structure matrix (DSM), a tool for modeling and visualizing architectures.

One big takeaway for me was their explanation of "complexity catastrophe". The crossover point when the cost of change exceeds the benefit. A metaphor superior to most interpretations of "technical debt", IMHO.

So when you have a DSM, these kinds of problems pop out, can't be ignored.

We believe that socketing memory is a good tradeoff to enable greater longevity through memory upgrades, and also a more useful upgrade path by being able to keep your memory during certain mainboard upgrades.

In the DDR5 timeframe, there is less of a power delta as well between DDR5 and LPDDR5.

Can you please point us to some technical report or something where they document the energy saving of moving from dimm slots to soldered ram/ssd ?

I don't think you'll find any. The way I understand it, the savings come from the new LP variant of DDR chips and nobody ever bothered to put them on DIMMs.

And even if the DIMM design (it's more than just a slot) doesn't make sense anymore, we could easily make a new "dumb" memory slot standard that would allow us to swap out the chips directly - like we have for CPUs.

> we could easily make a new "dumb" memory slot standard that would allow us to swap out the chips directly - like we have for CPUs.

You could, but it would take up a lot of space, given the higher pin count, and would most likely increase power consumption.

By the way, check out reviews of the Framework Laptop for a first-hand illustration of this effect. They seem to be getting 40-50% lower battery runtimes than laptops with comparable specs. That's example the problem — of you design for modularity, you have to leave some optimizations on the table.

The battery impact for us at this point is likely more a factor of additional firmware and driver tuning that we need to do. We've tested to 10-11 hours in MobileMark 2018, but have actually seen video playback runtime (which should be a lighter load) drop below that due to some audio driver issues that we're working through.

There is certainly some impact to using DDR4 instead of LPDDR4, but we believe it is still the better choice for most consumers by allowing for greater product longevity at lower cost.

How about standby time?

That doesn't in any way tease out the effects of sockets it seems impossible that that doubles power usage.

Doubtful, the Framework laptop is pretty much like any other USB-C-based laptop, except that it reserves some space inside the chassis so that the adapter dongles are flush with the case.

Unfortunately, I am not aware of any document like that. There are some papers comparing everything efficiency of different types of RAM, and there is of course indirect evidence of premium notebooks having significantly improved the battery life after moving to LPDDR.

LPDDR does many, many things to reduce power consumption (the current version even does DVFS, which I think is a first for main memory; graphics cards have been doing interface DFS for a long time, but a high end graphics card will burn around 30-70 W in the memory chips, and uses extremely high speed PHYs running at around 20 GBit/s per pin in the current generation).

The only thing that really changes between DIMM and soldered on is a somewhat increased trace length. That accounts only for a very small increase in power.

> That accounts only for a very small increase in power.

What are you basing this statement on?

PHY is only part of the DRAM chip, and trace length is only a part of what determines energy per bit, so increasing trace length by x % wouldn't increase overall power by x %, but a fraction of that. Thinking about it, there is a bigger problem than just energy/bit scaling; since LPDDR switches dynamically between high-speed and lower speed modes, the high speed mode is likely not used a lot. However, the low speed modes rely on being able to turn the termination off, which saves significant amounts of power. A socket may make that difficult to achieve, and if you have to run with ODT in the, I'm assuming, mostly-used low power mode, you might end up with a significant increase in power.

Of course there are multiple factors at play, but that observation doesn’t justify the conclusion that any particular factor makes only a ‘very small l’ contribution. Do you have even a back-of-an-envelope calculation to support this claim? The knock-on effects of increased trace length can be quite complex.

Why is it that you can't have this with socketed RAM? Is there a physical limitation?

As far as I understand, a number of things make a difference:

- Socketed RAM is limited by the number of pins, DDR SO-DIMM for example has 2-3x fewer pins than your average LPDDR4X module. Having more pins helps running the chips on lower power as it was explained to me by an engineer (something to do with grounding and lower voltage, I am not really sure how it works but I trust the guy)

- Socketed RAM is a least common denominator, there will inevitably be some variance between different modules, which leaves you with less possibilities for optimizing the electrical connection

- Current LPDDR is much more sophisticated than regular DDR, the protocol is different etc. So you can't just stick LPDDR chips on a DDR SO-DIMM and expect it to work

I am sure one could design a modular system based on LPDDR, but the question is whether such system would be feasible. It would be limited to premium laptops only (as modular system would take too much space in a smartphone), it would probably need a higher pinout (which would mean a complex mounting bracket of some sort) and it would require laptop manufacturers to agree on a certain standard.

There are other considerations as well, especially if one wants to break the current performance boundaries. For example, Intel has announced that they will integrate HBM with their upcoming Xeons. How do you intend to make that socketable, that's 2048-bit memory bus. Or consider Apple's new in-house SoCs. By mounting RAM directly onto the package substrate they can potentially deliver high-bandwidth RAM in an energy efficient package without increasing the cost and the complexity of the mainboard. E.g. their upcoming hardware is widely expected to use 256-bit RAM, something that has so far been reserved for workstations. As you make the interface wider and wider, modularity becomes more and more expensive. Why don't GPU's offer modular RAM for example?

RAM could be packed together with a CPU on a replaceable board.

Not much. LPDDR4 signalling is actually more robust than that of regular DDR.

It's not physically impossible to make DIMMs with LPDDR4, just nobody really wanted to come with a standard.

I imagine you may have some difficulty getting the pins to make contact if they’re too small.

Modern CPUs are an order of magnitude more complex and usually built on a smaller process, yet we have no trouble putting them on intermediary PCBs to spread out contacts and allow larger pins. I see no reason we couldn't make LGA sockets for DRAM chips in the same way.

The beauty and success of DIMM modules is that they are a simple slide-in to install technology. It's easy to operate, cheap, robust, and it is very space efficient. Imagine RAM modules in an LGA socket instead... how much space would it require? How robust would it be? What will be the price? I am quite sure all of these issues can be solved, but is it worth it? Let's be honest, users who are interested in having upgradeable RAM in their ultraportable laptop are such a small minority that the practical interest in developing modular low-powered RAM is close to zero.

The practical interest might be close to zero because in the past memory upgrades a number of years down the road were one of the most common upgrades and added tremendous value and another few years of life. Similarly many older laptops got a second life by being upgraded to SSDs. Accidents like that, which prevent sales of new units, can't happen if storage and memory are soldered on.

I don't think this is the driver of these decisions per se, but it is undeniably a bonus for the manufacturer; non-upgradable devices become obsolete faster, necessitating new replacements.

Also even when RAM was replaceable, many manufacturers put low limits on its size, I guess for the same purpose.

I don't know about other brands, but for ThinkPads (both in the IBM and Lenovo eras), there was often a stated maximum RAM that was based only on the capacity of SODIMMs available at the time of manufacture.

But it wasn't an artificial limitation. In practice, once larger capacity memory became available with the same technology and form factor, it would work fine. I have upgraded several ThinkPads with memory beyond what the original datasheets said was possible.

Yeah... sadly CPUs in laptops have been soldered for a while now.

The same CPUs that have a socket equivalent for desktops. It's a travesty.

I have absolutely no idea, but I guess the bus to the processor and other periphery could indeed be a weak point because of EM-interference. We have absolutely crazy data rates here and therefore very high frequency signals so proximity between components might indeed matter.

That said, manufacturers have huge ambition to not allow for modular design.

> In a universe where you want high-performance energy efficient RAM

You are cherry-picking an example that fits only a specific category of laptops.

Even so, CPU and RAM could be bundled together into a socketed cartridge that is easy to upgrade.

The real reason is all about planned obsolescence and profit margins.

RAM energy efficiency is not important because very small fraction of the energy consumed by other components.

It only somewhat matters for power consumption in sleep mode when the RAM is the only component being powered, but at least on Windows that’s solvable with a fast SSD and hibernation.

RAM energy efficiency is extremely important because it is a significant part of the baseline power consumption. Power-hungry components like CPU and GPU can be run in the lowest power state most of the time, so it really comes down to RAM and the display. If you have a 60Wh battery and want to get 15 hours battery life, you need to get your idle power consumption under 4 watts. So even 0.5 watt in baseline RAM consumption makes a huge difference.

I recall underclocking the RAM on my video card a few years ago when in desktop mode. Reduced the heat in the PC case by several degrees.

Performance difference between system RAM and VRAM is an order of magnitude. Specifically, DDR4-3200 peaks at 25.6 GB/second per channel, modern high-end GPUs peak at 500-900 GB/second depending on the GPU model.

VRAM consumes substantial power to deliver that performance. On many GPUs, VRAM chips are actively cooled. System RAM doesn’t even need passive heatsinks, uses too little electricity to care.

GPUs do use fairly fine-grained dynamic frequency scaling on their memory chips for precisely this reason: having a top-end GPU run at around 15-20 W under light loads simply wouldn't be possible otherwise.

Yeah, but at the time my card only went down half-way, to 1200 MHz IIRC, while I made it go down to 800 Mhz in desktop mode. In gaming it was closer to 3 GHz. This was before my NVIDIA GTX 970, some AMD model.

Just really blew my mind how much heat the VRAM contributed.

I love the idea of needing to replace the laptop to upgrade RAM. /sarc

I find this to be less of an issue than the noise it generates. If you maximise ram at purchase, the limitations of a laptop seem to be battery and storage more than anything else. A 8 year old MacBook is just fine for a lot of usage, except for the dead battery and limited storage. Having these easily repairable would make my life a lot easier - this laptop looks great.

Modularity is a great feature, but modularity is a mobility killer.

Modular laptops, phones, etc.. will be always penalized in terms of either performance or battery life for a given size.

>You simply can’t have this with socketed RAM.

[Citation needed], there is no real electrical engineering reason why you couldn't.

The problem is not that the chip is soldered, but that you can't buy it. Soldering BGA is not that difficult.

I would have to leave it to a professional, I can barely solder connectors on my RC car power packs

Then make the CPU+RAM user-replaceable.

Where there's a will there's a way. The problem is there's no will.

1 watt active is low. But so is 2 for dimm ddr3 [1]. They switched to make upgrading more difficult.

"Typical Power Consumption of PC Components - Power Draw in Watts" https://www.buildcomputers.net/power-consumption-of-pc-compo...

You've repeated this in a few places in this thread with this same sort of conspiratorial air, but when you are talking about south of 10W total consumption (sometimes south of 6W!), a watt is a very significant amount and has a very large impact on overall battery life.