SK Hynix Throws a Jab: CAMM is Coming to Desktop PCs

[TheLostSwede]

None of that is in the same realm of desktop, ddr5 up to 7000? No mention of timings (which are likely abysmal)?

Actually, it's up to 8000 for now and will most likely go faster in the future. That was just the one thing that was posted here on TPU about it.

The cooling scenario is entirely different with parts consuming a fraction of what desktop parts use all while having everything in a laptop strapped to a unified heatpipe/vaporchamber cooler with blower fans making your ears bleeds as soon as you put a load thats going to max the available tdp.

Format is a terrible idea for desktops, heat will undoubtably be an issue. Comparing DDR5 7000 with loose timings c48+ at low 1.1-1.2v isn’t the same thing as a desktop setup. Go put a gen4 nvme drive on the back of an itx board and see what happens to temps.

Whatever dude, you clearly don't understand the benefits and believe the PC desktop is a fixed thing that has never changed over the past 40+ years...
Also, PCIe 4.0 NVMe drive thermals ≠ DDR5 8000 thermals, but again, whatever, you have clearly made up your mind, so not point continuing this discussion.

 

[rv8000]

Actually, it's up to 8000 for now and will most likely go faster in the future. That was just the one thing that was posted here on TPU about it.

Whatever dude, you clearly don't understand the benefits and believe the PC desktop is a fixed thing that has never changed over the past 40+ years...
Also, PCIe 4.0 NVMe drive thermals ≠ DDR5 8000 thermals, but again, whatever, you have clearly made up your mind, so not point continuing this discussion.

I’m just not naive enough to believe slapping a bunch or memory ic’s on the rear side of a motherboard with 100-250w cpus radiating heat directly to passively cooled parts with no direct airflow isn’t going to severely limit frequencies, voltages, and can potentially negate proximity and trace length benefits without significant improvements to ic efficiency.

Go play with some desktop ddr5 at 8000 and above and come back to me about cooling limitations.

 

[evernessince]

At CES, some motherboard manufacturers showed boards with all the power connectors on the back of the board.
These require new cases, so I'd expect such cases to change to accomodate cooling for rear-mounted RAM if that's the way things go.

That was my first though as well. You could place the RAM, cables, and M.2 slots all on the back of the mobo.

If there are benefits to using this new form factor I don't see any reason to not adopt it. Preferably for desktop they'd have a tool-less design. I already hate installing the tiny screw for M.2s.

I’m just not naive enough to believe slapping a bunch or memory ic’s on the rear side of a motherboard with 100-250w cpus radiating heat directly to passively cooled parts with no direct airflow isn’t going to severely limit frequencies, voltages, and can potentially negate proximity and trace length benefits without significant improvements to ic efficiency.

Go play with some desktop ddr5 at 8000 and above and come back to me about cooling limitations.

It would be possible to cool backside components with a lateral fan. assuming they do end up putting memory near the CPU. There would be an added bonus to that in that it would keep the CPU cooler as well. Of course it's also possible that cases adapt and include fan mounts that can hit the back of the motherboard or they can add large perforations similar to what you have on the outside of most cases to the motherboard tray to increase airflow and heat dissipation. Depending on how much material they remove from the tray for the perforations, they could offset any loss of structural rigidity by increasing tray thickness. In fact, I think in general a perforated tray would be a good idea as it would help keep things cool regardless of back mounting or not.

 

[rv8000]

That was my first though as well. You could place the RAM, cables, and M.2 slots all on the back of the mobo.

If there are benefits to using this new form factor I don't see any reason to not adopt it. Preferably for desktop they'd have a tool-less design. I already hate installing the tiny screw for M.2s.



It would be possible to cool backside components with a lateral fan. assuming they do end up putting memory near the CPU. There would be an added bonus to that in that it would keep the CPU cooler as well. Of course it's also possible that cases adapt and include fan mounts that can hit the back of the motherboard or they can add large perforations similar to what you have on the outside of most cases to the motherboard tray to increase airflow and heat dissipation. Depending on how much material they remove from the tray for the perforations, they could offset any loss of structural rigidity by increasing tray thickness. In fact, I think in general a perforated tray would be a good idea as it would help keep things cool regardless of back mounting or not.

That’s where this borderlines on being an impractical change to standards; similar to the 12vhpr connector or rear mounted connectors on motherboards where there’s little to no benefit in implementing either.

Implementing this in almost any modern case is rather dumb as the cable management is done on the rear side of the motherboard tray and will block most if not any airflow to be provided in a case that isn’t dual chamber.

There are SO many standards and designs that would have to be changed across the industry to accommodate it, for little to no benefit. Desktop memory frequency, latency (in relation to timings) and capacity has aways been superior to mobile/laptop implementations because heat and power requirements are much different. Theres no current benefit aside from companies making billions by forcing new standards in the industry.

 

[Nioktefe]

I think where this might come is for strix halo. The rumors talk about 256 bit lpddr5x bus which is twice what is on am5 (source : https://www.techpowerup.com/311895/...d-to-feature-16-cores-improved-io-die-and-gpu )

In this case CAMM is the only real way to bring it to desktop since lpddr5x is not possible on a DIMM

 

[Scrizz]

so.... If I need 512GB of RAM on my desktop/workstation, this is going to have a much larger footprint that the equivalent DIMMs.
I can see the benefits for laptops and space constrained systems, but for desktops/workstations this seems like a very bad idea.

 

[R0H1T]

The rumors talk about 256 bit lpddr5x bus which is twice what is on am5

Well it's not going to be a problem, Apple has what 1024bit(?) bus width on the M2 ultra IIRC.

 

[thestryker6]

Why not do this on all motherboards? You could place one on each side of the CPU socket and move the CPU socket slightly further away from the power regulation circuitry. No need for the CPU coolers to have clearance for the RAM any more.

...

Why? RAM doesn't get very hot and it would be super easy to put a heatsink on the CAMM modules, just like on normal DIMMs.

This only works if the voltages for low latency with high bandwidth comes down a lot as running 8000+ already gets shaky without active cooling. I certainly want the change to CAMM to happen as several industry businesses have spoken of it enabling more bandwidth I just don't know if putting them on the back will work. I'm also curious what the DDR5 capacity is going to look like on them, and what the module sizes end up being as it looks like Samsung and Micron have only shown LPDDR5/X modules.

I think where this might come is for strix halo. The rumors talk about 256 bit lpddr5x bus which is twice what is on am5 (source : https://www.techpowerup.com/311895/...d-to-feature-16-cores-improved-io-die-and-gpu )

In this case CAMM is the only real way to bring it to desktop since lpddr5x is not possible on a DIMM

CAMM modules for DDR and LPDDR have different pinouts so they cannot be used interchangeably. That means either moving all CPUs over to LPDDR or making custom ones for these SKUs neither one of which seems particularly likely.

 

[Chaitanya]

Why not do this on all motherboards? You could place one on each side of the CPU socket and move the CPU socket slightly further away from the power regulation circuitry. No need for the CPU coolers to have clearance for the RAM any more.

You can see it in this video, although he doesn't remove the "shim".

ATX and mATX arent starved for real estate, maybe in future if these modules can bring down latency and increase capacity significantly the surely will be adopted across industry. No more restrictions on memory clearance is a big plus for this form factor. There are a always handful of SFF boards and system with SO-DIMMs majority of them still have slots perpendicular to board leaving the memory restrictions on coolers in place.

 

[Xajel]

Wow, so ATX motherboards will finally have a single PCIe slot now!!

Seriously, They have to do something, people are using MATX instead of ITX or ATX because they might need extra PCIe Slot, but still need a compact system, the additional M.2 slots wasted so much space that there's a MATX or two that has only a single PCIe slot!! it's like moving from ITX to MATX just because they needed more space for M.2 slots !!

They have to extend the idea of using vertical M.2 mount, like some ASUS motherboards, but this can happen somewhere else, two M.2 drives can be mounted vertically in the same space occupied by a single horizontal M.2 slot, with plenty of space to add heatsinks as well.

How on earth they will add these CAMM? I guess a single module can take less space than 4 slots, yet it can do dual channel, but how can they add two? stack them over with some gap between? this actually can work, especially for compact quad channel memory system (MATX Threadripper coming back?) Or we might finally get Threadripper Pro in ATX formfactor, EATX no more.

Motherboards are getting more features that makes them bigger, the market for compact system is endangered because most of these features are implemented horizontally, some had to innovate by having some logics in a separate daughterboard (VRMS) or moving to the back of the motherboard (M.2 slots), but it keeps getting more crowded with ATX 12VO requiring extra power circuit on the motherboard to provide 5V & 3.3V (with extra space for the connectors as well).

 

[remixedcat]

The latency isn't about the chips themselves, but rather between the memory module and the CPU socket. Two different things. Sorry if that wasn't clear.

and this will help reduce DPC latency which musicians desperatly need! RN we gotta do all kinda kernel tweaking and CPU governor changes and all kinda stuff and reducing any BG apps and all kinda things and I am all for anything that makes it less work we have to do as musicians that need more time for creating and less time having messing with stuff!

 

[TheLostSwede]

How on earth they will add these CAMM? I guess a single module can take less space than 4 slots, yet it can do dual channel, but how can they add two? stack them over with some gap between? this actually can work, especially for compact quad channel memory system (MATX Threadripper coming back?) Or we might finally get Threadripper Pro in ATX formfactor, EATX no more.

They can't be stacked, that's the whole point of the design. It's going to require a change to your typical ATX style moboard layout.

 

[onemanhitsquad]

How are you going to see your "RGB show" if the memory is on the back?!

 

[thestryker6]

They can't be stacked, that's the whole point of the design. It's going to require a change to your typical ATX style moboard layout.

That depends as the spec supports both stacked and not depending on whether or not you're using a single or dual channel implementation:

This is one of the main reasons I'm curious about DDR5 capacity on CAMM as desktops might be forced into single channel mode to maximize capacity, but that would make cooling harder.

 

[TheLostSwede]

That depends as the spec supports both stacked and not depending on whether or not you're using a single or dual channel implementation:
View attachment 330624

This is one of the main reasons I'm curious about DDR5 capacity on CAMM as desktops might be forced into single channel mode to maximize capacity, but that would make cooling harder.

Hmmm, the initial point of the CAMM was to get away from stacked modules, so it seem like JEDEC has added things I wasn't aware of.
So we might see single channel CAMMs for desktop and dual channel for mobile then... urgh...
Thanks for sharing this, as I don't think it's that widely known. Is it your name and email address on the first pic?

 

[thestryker6]

Thanks for sharing this, as I don't think it's that widely known. Is it your name and email address on the first pic?

Yeah I was too lazy to pull it off since it's what I use on tech forums and whatnot (even though I still. probably should have ). It's off the free to download JEDEC spec sheet.

Hmmm, the initial point of the CAMM was to get away from stacked modules, so it seem like JEDEC has added things I wasn't aware of.
So we might see single channel CAMMs for desktop and dual channel for mobile then... urgh...

Yeah I was confused by the wording in the JEDEC announcement where they were talking about stacked CAMM so I went and registered so I could check the spec sheet.

 

[remixedcat]

How are you going to see your "RGB show" if the memory is on the back?!

You'd have to have easy side panel access to even replace or upgrade modules.