Thursday, January 29th 2026

Qualcomm and AMD to Adopt SOCAMM2 Memory in AI Products

Qualcomm and AMD are reportedly considering the integration of SOCAMM2 memory into their AI product lines to enhance memory capacity and speed. Following NVIDIA's "Vera" CPU, which features LPDDR5X memory on the SOCAMM factor, offering 1.2 TB/s of memory bandwidth and supporting up to 1.5 TB of LPDDR5X memory, Qualcomm and AMD aim to pursue a similar strategy to achieve high-speed, high-capacity AI systems. These systems require fast memory close to AI accelerators, and SOCAMM can complement the extensive HBM memory on the accelerators. It acts as a fast memory pool with sufficient capacity to store entire models in memory, reducing the need for data transfer to SSD flash storage.

NVIDIA's SOCAMM approach involves surrounding the CPU with multiple SOCAMM modules, a strategy Qualcomm and AMD are expected to adopt in their AI solutions. AMD's Instinct MI accelerators, paired with EPYC CPUs, might be the next candidates for this integration, or AMD may develop a completely new system. Qualcomm already offers AI solutions like the AI200 and AI250 inference accelerators, which include up to 768 GB of LPDDR5 per card. These could soon be complemented by SOCAMM, allowing for standardized memory expansion based on the SOCAMM standard rather than soldering individual memory modules. This approach would enable manufacturers to offer various system versions and solutions by simply adding or removing SOCAMM modules, without the need for extensive PCB soldering work.
Source: Hankyung
Add your own comment

12 Comments on Qualcomm and AMD to Adopt SOCAMM2 Memory in AI Products

#1
azrael
I'm starting to wonder if, were there a switch/filter to filter out all articles related to AI, would I be looking at a blank front page? Just a general remark, not exclusively related to this article.
Posted on Reply
#2
bonehead123
azraelI'm starting to wonder if, were there a switch/filter to filter out all articles related to AI, would I be looking at a blank front page? Just a general remark, not exclusively related to this article.
Just an FYI/PSA: If you filter out the word "AI", the "AI" will filter you out, hahahahaha :D

But yea, I get ya....seems as if stuff isn't related to AI, there would be hardly anything to write about anymore......

Maybe the mods can create a subforum/category called "everything else except AI crap" :D
Posted on Reply
#3
rv8000
azraelI'm starting to wonder if, were there a switch/filter to filter out all articles related to AI, would I be looking at a blank front page? Just a general remark, not exclusively related to this article.
I’ve honestly thought about putting in a request/feedback post about that very idea. I don’t expect it would be received well by staff though for the normal reasons.
Posted on Reply
#4
john_
While this news might look like AI related only, when I read "LPDDR5X memory on the SOCAMM factor" I understand one thing. Get ready for very very VERY expensive laptop memory in the future. So, while it looks unrelated to us, it's not. It's completely related and alarming.
Posted on Reply
#5
LX_forReal
SOCAMM2 memory is very suitable for SOCs that require high bandwidth like the Halo series, but now it gives way to AI (as an aside, is there really an enterprise using Qualcomm's AI series inference accelerators?)
Posted on Reply
#6
GenericUsername2001
john_While this news might look like AI related only, when I read "LPDDR5X memory on the SOCAMM factor" I understand one thing. Get ready for very very VERY expensive laptop memory in the future. So, while it looks unrelated to us, it's not. It's completely related and alarming.
Datacenter products using LPDDR5X will/are driving up the price of laptop RAM already, whether or not they are soldiered vs SOCAMM. Datacenter products going with SOCAMM instead of soldiering the memory down is actually the less bad option for us. Mass production of SOCAMM memory will make it more appealing for laptop makers to use vs soldiering stuff in too, and when those data center AI-GPU-BBQ Sauce devices get upgraded or go EOL it will be possible for the SOCAMM units to be pulled then sold on the used market, which is a lot better than them being stuck soldiered to board.
Posted on Reply
#7
_JP_
*Running around, waving arms*
SOCAMM2 was going to be the next step for the darn laptops! Now we won't even get that?!
Posted on Reply
#8
ymdhis
azraelI'm starting to wonder if, were there a switch/filter to filter out all articles related to AI, would I be looking at a blank front page? Just a general remark, not exclusively related to this article.
The front page would be a huge list of articles about Windows 11 updates breaking a feature, and "Genuine Software deals" ads.
Posted on Reply
#9
Tigerfox
GenericUsername2001Mass production of SOCAMM memory will make it more appealing for laptop makers to use vs soldiering stuff in too
I very much hope that. I would prefer SOCAMM2, but LPCAMM2 would be fine, too.
LX_forRealSOCAMM2 memory is very suitable for SOCs that require high bandwidth like the Halo series, but now it gives way to AI
Framework allegedly worked together with AMD to make LPCAMM2 work on Framework Desktop with Strix Halo, but signal integrity was too bad. If SOCAMM2 were to work with Strix/Gorgon Halo (and Intel PTL with 12 Xe³), that would make those SoCs much more apealing for me.
What do you mean "but now", the main benefit of Strix Halo is huge shared memory for CPU and GPU for AI.
_JP_SOCAMM2 was going to be the next step for the darn laptops!
SOCAMM2 was never intended for Laptops until now, but solely for Nvidia Workstations.
Posted on Reply
#10
LastDudeALive
LX_forRealSOCAMM2 memory is very suitable for SOCs that require high bandwidth like the Halo series, but now it gives way to AI (as an aside, is there really an enterprise using Qualcomm's AI series inference accelerators?)
Per their website, they're an option offered by a lot of cloud/data-center providers (AWS, IBM, Lenovo, HP, etc.). They're just another non-Nvidia (i.e. budget) offering.
Posted on Reply
#11
TheinsanegamerN
GenericUsername2001Datacenter products using LPDDR5X will/are driving up the price of laptop RAM already, whether or not they are soldiered vs SOCAMM. Datacenter products going with SOCAMM instead of soldiering the memory down is actually the less bad option for us. Mass production of SOCAMM memory will make it more appealing for laptop makers to use vs soldiering stuff in too, and when those data center AI-GPU-BBQ Sauce devices get upgraded or go EOL it will be possible for the SOCAMM units to be pulled then sold on the used market, which is a lot better than them being stuck soldiered to board.
Agreed. Form factor deoesnt matter, chips are chips.

This frankly is GREAT news if it sees adoption. The current generation of chips are gimped by memory, the 780m was fast enough to see a substantial ~30% performance bump from having LPDDR5x speed memory, and that was 12 CUs. The 890m in the HX370 is 16 cores and on the faster rDNA 3.5 arch, but it's stuck with the same slow arse DDR5 so dimms, or with LPDDR5x which is fast enough to feed it, but is soldered, therefore non upgradeable AND restricted to capacities offered by the manufacturer. Dell wont sell you 64GB of RAM on a HX370 unless you buy their precision line, for example.

Putting LPDDR5x on a removable form factor and making that the standard is the best of both worlds.
Posted on Reply
#12
Wirko
TigerfoxI very much hope that. I would prefer SOCAMM2, but LPCAMM2 would be fine, too.
SOCAMM, like HBM, is for those who need extreme density and can pay for that. 192 GB packed on a surface area not much bigger than four LPDDR chips means 16-high TSV stacking of 3 GB dies. Perhaps a base die is needed to make all of this work, I don't know. Nothing about SOCAMM (and HBM) is designed to be cheap.
Posted on Reply
Add your own comment
Nov 17th, 2026 14:06 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts