P4-630Post me a few 14"/16" laptops that take 64/128 GB RAM.

Are you kidding?

lexluthermiesterAre you kidding?

Ok I see some 64GB there are.. (ms surface)
Anyways I will still say that 16~32GB RAM is enough for the average user IMO.

P4-630Anyways I will still say that 16~32GB RAM is enough for the average user IMO.

For a system that has a dedicated GPU with it's own VRAM, sure 16GB would be ok. But for an IGP based system? 16GB is just not enough these days. 32GB? Yeah ok, that's good and will be for a while.

john_Exactly. RAM and storage should remain unsoldered. There are ways to do it without having to sacrifice the thin form of a laptop. So-dimm slots and SSDs can be arranged as extensions of the PCB for example, instead of being put on top of the PCB. Modern laptops don't need a huge motherboard and usually only come with 2-3 ports, so the motherboard doesn't need to be big, meaning there is plenty of space to put the motherboard, the battery and a few horizontal expansion connectors, meaning the RAM and one or two SSDs.

Don't bring common sense with you next time! How are we supposed to sell CAMMs now that you've said that?

lexluthermiesterFor a system that has a dedicated GPU with it's own VRAM, sure 16GB would be ok. But for an IGP based system? 16GB is just not enough these days. 32GB? Yeah ok, that's good and will be for a while.

I'm going to get backlash for saying it, but this is where Apple has done it right. No, not the part where they still sell a $1000 laptop with 8GB RAM and charge hundreds of dollars more for an increase to a bare minimum usable config. It's the part where they increase LPDDR capacity and bandwidth on their "Pro" and "Max" targeted SoCs. It's only logical that you will want more and faster DRAM attached to the SoC if you choose a higher performance model or else you squander the advantages of that extra compute power by choking it in bandwidth or capacity constraints. They also allow configurations up to 128GB in a 14" laptop which... Who else is doing that currently? I'm sure some other brand offers that, but it's definitely not a common configuration option even today.

FouquinIt's only logical that you will want more and faster DRAM attached to the SoC if you choose a higher performance model or else you squander the advantages of that extra compute power by choking it in bandwidth or capacity constraints.

AMD could theoretically deal with it with their infinity cache, but they never tried & they probably never thought a big IGP could sell at a (massive?) premium. Apple of course changed that & here we are! For that AMD should definitely thank Apple & we as consumers will have more choices now, hopefully leading to better/lower prices o_O

WirkoDon't bring common sense with you next time! How are we supposed to sell CAMMs now that you've said that?

That's kind of the idea. CAMM/CAMM2 needs to stay away from the consumer sector.

FouquinI'm going to get backlash for saying it, but this is where Apple has done it right.

Not from me. While I'm no fan of Apple, I agree with you, lots of system RAM is one of the things they do right.

lexluthermiesterI don't give a flying rats bum what people think is enough. I care about being able to upgrade and expand.

lexluthermiesterCAMM/CAMM2 needs to stay away from the consumer sector.

So since these two statements contradict each other I have to ask: why? You want to retain the ability for consumers to upgrade and expand their machines, but you want CAMM to stay out of consumer machines; removing the ability to upgrade and expand. SODIMM is a dying standard for many reasons. One of which is bandwidth and signal integrity limits and the other is mechanical constraints. CAMM/LPCAMM, as it has been demonstrated, is a higher bandwidth, higher integrity, and lower mechanical footprint alternative to SODIMM that consumers can use to upgrade their machines. So, why do you advocate that it stays out of the consumer sector? Unless you're advocating for a yet finalized/developed third solution that I'm unaware of? Genuinely curious.

FouquinSo since these two statements contradict each other

Not really.

FouquinSODIMM is a dying standard for many reasons.

That's an opinion. There is nothing wrong with SODIMMs. If it's not broken, don't fix it. CAMM/CAMM2 is not the right way forward.

lexluthermiesterFor ultra portables, sure. But not all laptops are ultra portable. Some of us expect upgradeability regardless of form-factor. You discounting that doesn't change the expectation.

No not all laptops are ultra portables, the normal laptops will be getting Gen 15 arrowlake which will come with bog standard ram, ultra portables will be getting lunar lake with on package ram

lexluthermiesterNot really.

Yes, really. Quote: "I care about being able to upgrade and expand," and, quote, "CAMM/CAMM2 needs to stay away from the consumer sector," contradict each other as statements on the grounds that a standardized solution that allows you to achieve user configurable expansion with increased bandwidth and signal integrity such that CAMM/LPCAMM provides does not currently exist. So you'll have to present new solutions to these fundamental problems and work to have it standardized in place of CAMM. Or perhaps you are simply virtue signaling.

lexluthermiesterThat's an opinion. There is nothing wrong with SODIMMs. If it's not broken, don't fix it. CAMM/CAMM2 is not the right way forward.

Thank you for reiterating your own opinion, but you did not answer my question. Though my own opinion on SODIMM is very much an opinion, it is one shared and published by JEDEC, Samsung, Micron, AMD, Intel, and many many more. I asked for what you advocate to solve the clear and present issue of retaining configurability without halting progression. If you advocate for SODIMM, I'm sorry to inform you that it is an impossibility.

I'm seeing the signs of your heels digging deeper into the sand on this topic, so we'll need to cut this off nicely before things turn nasty. I expect a rebuttal but I won't respond further, so feel free to get the last word in on this. I won't contend.

FouquinOr perhaps you are simply virtue signaling.

I think you're reading into things that aren't there.

Fouquinbut you did not answer my question.

That was deliberate.

Fouquinit is one shared and published by JEDEC, Samsung, Micron, AMD, Intel, and many many more.

If everyone jumped off a cliff and then told you to join them as it was for your own good, would you follow them?

Fouquinso feel free to get the last word in on this.

Ooo, witty.

FouquinI won't contend.

Good. You're learning.

kondaminNo not all laptops are ultra portables

True.

kondaminthe normal laptops will be getting Gen 15 arrowlake which will come with bog standard ram

While that's a fair point, the one I'm trying to make is that upgradeability should be an option engineered into even ultra-mobile platforms.

kondaminultra portables will be getting lunar lake with on package ram

Yeah, this isn't a terrible idea, if done right.

lexluthermiesterNot really.


That's an opinion. There is nothing wrong with SODIMMs. If it's not broken, don't fix it. CAMM/CAMM2 is not the right way forward.

Nothing wrong? Only if you agree with staying with a form factor that struggles going beyond JEDEC speeds with DDR5. Speed is the same reason that CAMM2 is being considered for desktop use when DIMMS will start having a hard time with the newer standards.

"If it's not broken don't fix it" is starting to get abused as an argument (probably fueled by the recent fumbling of the 12VHPWR). AGP was also fine when PCIe was out—single-digit improvement, even with GPUs released in 2006. You could have argued that PCIe was never needed for GPUs back then. Keep saying it loud enough, and more companies might just be tempted to not try anything new at all, and releasing the same stuff, but with ever higher margins.

OLED ? MicroLED ? why when IPS could do just fine as the end game for monitors. Sure the tech isn't without fault, but It's not broken and fairly accurate. OLED also got issues that are annoying to fix, and require heavy R&D money. Miniled? it's improving some stuff, but also got it's own set of issues, let's keep edge-lit, and not spend any R&D money into making miniled better.

NoyandNothing wrong? Only if you agree with staying with a form factor that struggles going beyond JEDEC speeds with DDR5. Speed is the same reason that CAMM2 is being considered for desktop use when DIMMS will start having a hard time with the newer standards.

That has not been conclusively proven. However, even IF there is merit to that claim, CAMM/CAMM2 is NOT the way forward. CAMM/CAMM2 does NOT have the flexibility and configurability of standard DIMMs. It might be able serve mobile to reasonable effect, but not desktop/workstation/server applications. Especially not as designed. It's a concept that needs to be shit-canned because that's where it belongs, in the garbage.

Noyand"If it's not broken don't fix it" is starting to get abused as an argument (probably fueled by the recent fumbling of the 12VHPWR).

Abused? Or used properly to proper effect? Narrow-minded moronic designs need to be called for what they are.

usinameEVERY single CPU in this list is 4P + 4E cores, which is hilarious.

How many cores do you use on your ultra slim and light laptop? These aren’t for workstation laptops.

lexluthermiesterThis is an instant deal-breaker. Not acceptable at all.


Moose muffins!

How much RAM do you need in you ultra slim, ultra light weight laptop? These aren’t for running CAD software!

Tech NinjaHow many cores do you use on your ultra slim and light laptop? These aren’t for workstation laptops.

I think he was referring to the fact that we are calling a 4P+ 4E i9, i7, i5. Which is an egregious change from the past where they would all have different core counts. Now they are using clock speeds, memory and 1 less CU to differentiate them. There should be no iX label at all in this case something like Core Ultra 285V, 275V, 265V would be more appropriate

lexluthermiesterThat has not been conclusively proven. However, even IF there is merit to that claim, CAMM/CAMM2 is NOT the way forward. CAMM/CAMM2 does NOT have the flexibility and configurability of standard DIMMs. It might be able serve mobile to reasonable effect, but not desktop/workstation/server applications. Especially not as designed. It's a concept that needs to be shit-canned because that's where it belongs, in the garbage.


Abused? Or used properly to proper effect? Narrow-minded moronic designs need to be called for what they are.

You can also root for Cudimm
CUDIMM Standard Set to Make Desktop Memory a Bit Smarter and a Lot More Robust (anandtech.com).

Those are all JEDEC standards, nobody is going to get royalties for pushing those news connectors. If anything the industry as a whole would have an easier time to just give up and solder everything.

For DDR5, SO-Dimms are really behind DIMMs when it comes to speed. Even the sweet spot 6000Mhz CL30 doesn't exist in that format. The absolute majority of the sticks use JEDEC, with Kingston being the only brand selling a single SKU with timings a bit tighter. Even G-Skill of all brands is only selling JEDEC ripjaws. Buy a ROG, Alienware, Clevo, XMG, Puget laptop, the best that you'll get is 5600mhz, with loose timings. (unless you buy the 18" Alienware) In the DDR4 era, they were selling far more laptops with XMP memory.


Yes, abused in a discussion about laptops. Camm was developed to fix laptop problems...which it does. 4 dimms laptops are an insane rarity in 2024, and still cannot match the 256GB maximum allowed with camm2. And you know what else it solve? All those poor laptops being sold in a single-channel memory setup. The solution to have the so-DIMMs as PCB expansion might work for the laptops using a smaller battery, but what about the other? lots of laptops are 50% battery 50% PCB now. The other side of the PCB isn't empty either. What about the trace length? quality ? CAMM even managed to make LPDDR5 something not soldered anymore. Remember that most laptop makers have bad faith, to begin with: Asus Zephyrus laptops with Zen 5? still using good old soldered LPDDR5. The easy way out. That's the path that they will always take. Unless someone else proposes an easy solution for them. (any connectors isn't btw, it's always going to be more expensive for them than soldering, LPDDRx camm2 being used is a favor)

JEDEC is the one who started to look at it for desktop use because they think that DDR7 and up might be troublesome. But as you can see, there are multiple solutions developed to fix that problem.

If JEDEC is full of shit about the future specification of memory, then how do we know that everything managed by them isn't also a web of lies ? I don't understand what they have to gain by doing this though.

Tech NinjaHow much RAM do you need in you ultra slim, ultra light weight laptop? These aren’t for running CAD software!

None of that is for YOU or the makers of the systems being made to decide. That is for the user/owner to decide.

NoyandYou can also root for Cudimm
CUDIMM Standard Set to Make Desktop Memory a Bit Smarter and a Lot More Robust (anandtech.com).

Those are all JEDEC standards, nobody is going to get royalties for pushing those news connectors.

That seems ok. That design maintains the ability for easy user upgrades.

lexluthermiesterNone of that is for YOU or the makers of the systems being made to decide. That is for the user/owner to decide.

Come on, while I agree in principle, this is a U class apu, it's not a workstation so 32GB is pretty decent IMO. You'll have Arrow Lake H/HX if you need more memory, more cores and with higher TDPs targeting those that need the power. Even 32GB would be fine for photo editing and some video editing (maybe not 8K).

usinameEVERY single CPU in this list is 4P + 4E cores, which is hilarious.

I must be blind. Where does it say this?

Luke357I must be blind. Where does it say this?

Here;
www.techpowerup.com/review/intel-lunar-lake-technical-deep-dive/3.html

The Lunar Lake-MX CPU complex has a total of 8 CPU cores, four of these are the new Lion Cove performance cores (P-cores), and the other four are Skymont efficiency cores (E-cores).

The reasoning for dropping HyperThreading support is also discussed on that same page. I'm ok with that development as I haven't been using HT in years due to the Spectre/Meltdown and other side-channel/branch-prediction vulnerabilities, most of which depend on HT in some form.

Luke357I must be blind. Where does it say this?

The source - Videocardz