LPCAMM2 is here!

Associate
Joined
21 Mar 2023
Posts
303
Location
Hertfordshire

The future is here for laptop memory, better in pretty much every way and a nice alternative to this new soldered on trend!

What are your guys' thoughts on this?
 
RAM design really hasn't changed that much in decades as it's been a long thin slot in piece for so long so going to CAMM2 is huge even for the desktop. I really like it as it can provide very fast speeds and tons of capacity all on a single piece that's still interchangeable. I really hope this becomes the new standard as it's definitely an improvement. Really can't see any downsides whatsoever with this plus being a thin board it can be very nicely hidden into the board design with a nice heatsink for an even cleaner look.

No worrying about running in dual channel, mismatching modules and capacities, no more large heatsink clearance issues with air coolers, shorter traces for much faster speeds so hopefully better compatibility/reliability and all on a single small board that can still provide well over 100GB+ capacity right now, what's not to love about it I'd honestly change my board today for this tech if it was an option and I've only had it two weeks.
 
Last edited:
Hope they do a clean break for this with DDR6 and come up with a tool free install mechanism.
Yeah I do wish that the manufacturers would move entirely to a new format and forgo trying to make backwards compatibility a thing.

High end DDR5 is already proving too high speed for SO-DIMM so a lot of manufacturers are just soldering memory onto the motherboard to enable increased speed and reduce cost.

The DDR5 SO-DIMM speeds also tend to be much slower than the full sized modules which negates some of the benefits.
 
Could be interesting on graphics cards and tiny form factor PC’s

Less and less sensible on laptop>desktop>workstation>server.
 
RAM design really hasn't changed that much in decades as it's been a long thin slot in piece for so long so going to CAMM2 is huge even for the desktop. I really like it as it can provide very fast speeds and tons of capacity all on a single piece that's still interchangeable. I really hope this becomes the new standard as it's definitely an improvement. Really can't see any downsides whatsoever with this plus being a thin board it can be very nicely hidden into the board design with a nice heatsink for an even cleaner look.

No worrying about running in dual channel, mismatching modules and capacities, no more large heatsink clearance issues with air coolers, shorter traces for much faster speeds so hopefully better compatibility/reliability and all on a single small board that can still provide well over 100GB+ capacity right now, what's not to love about it I'd honestly change my board today for this tech if it was an option and I've only had it two weeks.
My thoughts exactly

just have to hope this tech gets adopted quickly and becomes the norm very soon
 
My thoughts exactly

just have to hope this tech gets adopted quickly and becomes the norm very soon

The big question is why?
The applications for a format like this are limited so unless it provides a strong value proposition away from laptops I don’t see much of a future. Its not a solution to anything that is real issues.

Graphics cards or add in boards with upgradable memory is solution people have been asking for a long time. The question is would AMD, Nvidia and Intel get behind something like this?
 
Saw another preview of Mini CUDIMM DDR5 which looks really out of place on a ATX size board.
Only just realised this is something completely different than LPCAMM2: https://overclock3d.net/news/memory/msi-and-intel-showcase-prototype-mini_cudimm-ddr5-memory-tech/

Looks like it's trying to have the same advantages as LPCAMM2 apart from everything being in one module, which does mean it can still be easily expanded. But it could also have the same issue as current DIMMs if it's too tall and gets in the way of CPU coolers.
 
The big question is why?
The applications for a format like this are limited so unless it provides a strong value proposition away from laptops I don’t see much of a future. Its not a solution to anything that is real issues.
We're well past the point where the length of traces from RAM chips to CPU place a significant limit on memory speed (both bandwidth and latency). Moving from DIMM to CAMM helps a lot with that. It reduces power consumption too, since it takes quite a bit of energy to move each bit across the memory bus.
 
We're well past the point where the length of traces from RAM chips to CPU place a significant limit on memory speed (both bandwidth and latency). Moving from DIMM to CAMM helps a lot with that. It reduces power consumption too, since it takes quite a bit of energy to move each bit across the memory bus.

I feel PCB traces are a small price to pay for the current density.
 
Last edited:
High end DDR5 is already proving too high speed for SO-DIMM so a lot of manufacturers are just soldering memory onto the motherboard to enable increased speed and reduce cost.

Not a chance soldered memory was brought about to increase speed or reduce cost.
100% about increasing profits both in terms of the reduced part count, but also then upselling machines with better overall specifications by artificially segmenting them i.e want 32gb then you have to buy the top spec that has more storage than you'll need and an overpowered graphics you'll never use.
It's the same tactic that worked so well for Apple.

While I accept CAMM has benefits e.g. faster due to shorter trace lengths, it's still a consumer unfriendly model. If you want to upgrade, then you can't just add another stick or pair of ram sticks, you have to replace the module. But then who is going to want all of those lower capacity modules if they can't be installed in 2's or 4's?
 
I feel PCB traces are a small price to pay for the current density.
Even when it gets so bad that laptops are actually faster than desktops? The difference in density isn't that big anyway - looks like it's currently 128GB vs 192GB (4x48GB). If it's possible to to fit two CAMMs on a motherboard (maybe put one underneath?) then the advantage would be the other way.
 
Even when it gets so bad that laptops are actually faster than desktops? The difference in density isn't that big anyway - looks like it's currently 128GB vs 192GB (4x48GB). If it's possible to to fit two CAMMs on a motherboard (maybe put one underneath?) then the advantage would be the other way.

This form factor is questionable in laptops, but a laptop will never beat desktop. 4x48gb isn’t a limit of DIMM.
 
This form factor is questionable in laptops, but a laptop will never beat desktop. 4x48gb isn’t a limit of DIMM.
If desktops stick to DIMMs, then the difference in acheivable memory bandwith will become larger over time, eventually limiting overall performance.

4x48GB is absolutely a DIMM limit. The only way to fit more than that in for a given size of memory chip is to add more DIMMs, and you can't do that without adding load on the memory controller, which means lower clock speeds and higher power consumption.
 
If desktops stick to DIMMs, then the difference in acheivable memory bandwith will become larger over time, eventually limiting overall performance.

4x48GB is absolutely a DIMM limit. The only way to fit more than that in for a given size of memory chip is to add more DIMMs, and you can't do that without adding load on the memory controller, which means lower clock speeds and higher power consumption.

I’m sorry but you’re wrong on all counts.
 
I’m sorry but you’re wrong on all counts.
Could you please explain to me what I'm missing? In order to fit more than 4x48GB, you need one of the following:

1) Bigger chips. This isn't possible because until they become commercially available, at which point you can use them on a CAMM too, so there's no difference.
2) More chips. You can't fit more on a DIMM because there physically isn't space.
3) More DIMMs. You can't just add more because the memory controller can't drive them.

Is one of those wrong, or or is there a 4th option I haven't thought of?
 
Could you please explain to me what I'm missing? In order to fit more than 4x48GB, you need one of the following:

1) Bigger chips. This isn't possible because until they become commercially available, at which point you can use them on a CAMM too, so there's no difference.
2) More chips. You can't fit more on a DIMM because there physically isn't space.
3) More DIMMs. You can't just add more because the memory controller can't drive them.

Is one of those wrong, or or is there a 4th option I haven't thought of?

Where you’re re going wrong is making assumptions.

A DIMM can fit from 2 to 32 IC’s and possibly more. IC’s range from 1-32gb each, possibly more. CPU’s support upto 24 DIMMs at 8000 or more mega transfers. All of these parameters will increase with time.

I currently have systems that officially support memory capacities of multiple terabytes. IIRC Sandybridge-E platforms supported memory sizes of 256-512gb.
 
Where you’re re going wrong is making assumptions.

A DIMM can fit from 2 to 32 IC’s and possibly more. IC’s range from 1-32gb each, possibly more. CPU’s support upto 24 DIMMs at 8000 or more mega transfers. All of these parameters will increase with time.

I currently have systems that officially support memory capacities of multiple terabytes. IIRC Sandybridge-E platforms supported memory sizes of 256-512gb.
I see. The difference in opinion is because I was talking about desktops, but you're comparing HEDT/workstation/server machines. That doesn't really change the analysis though - there's still a hard limit on how much RAM you can install based on the chip size, physical module size and number of modules the memory controller can drive.
 
Last edited:
Back
Top Bottom