DRAM Pricing Jumps 50%, Only 70% of Orders Getting Filled

“TSMC is Reportedly Considering 3D DRAM Production​

by AleksandarK Today, 10:55 Discuss (0 Comments)
The world's largest chipmaker, TSMC, is reportedly considering entering the memory sector with the manufacturing of advanced DRAM technologies. According to a report from The Information, which cites sources close to the plans, TSMC is contemplating starting the production of 3D DRAM. As we know, there is a desperate need for more DRAM output from the three major memory makers: SK Hynix, Samsung, and Micron. New manufacturing capacity is being developed, but even when it comes online, there might still be a supply imbalance. According to The Information, TSMC is considering 3D DRAM manufacturing at the request of its largest customer, NVIDIA.

Conventional DRAM is currently produced by the big three, with plans to advance into 3D DRAM production beyond what HBM currently offers. To effectively scale memory, the 3D approach is the only way forward, as Micron recently confirmed. Micron has decided not to bring 4F² DRAM to market, even though it claims to have invented the cell design and has been researching it for around two decades, opting instead for the path to 3D DRAM. As one of the most experienced silicon manufacturers globally, TSMC is seemingly evaluating this path as well.“
 
mullet said:
Could Intel take some of their FABs and make memory?
Click to expand...
I would not think, apparently memory wafer is a complete different tech than logic-compute, intel does not have the up to date memory know how and even a samsung that do both could not convert their compute wafer fab into a memory one, without a massive makeover, they are complete independant division for a good reason.

Memory wafer are way denser (and now even more so as they are vertically stacked)
 

"CPU-Z Data Shows 16 GB RAM Gaining Ground While 32 GB and 64 GB RAM Shares Shrink​

by DrUsmanSaleem Today, 19:03 Discuss (1 Comment)
PCs with 32 GB and 64 GB of RAM are becoming less common, at least in the latest CPU-Z Validator statistics. According to Doc TB, one of the developers behind CPU-Z, this is the first time RAM capacity on PCs has declined year over year. 32 GB is still the most popular configuration at 37.2% of validations, though its share is down 2.5% compared to a year ago. 64 GB took a much bigger hit, falling 18% to 7.3%. Lower capacities are moving the other way, with 16 GB up 12.9% to 33.0% and 8 GB up 5.5% to 11.0%, while 96 GB and 128 GB setups dropped by more than 26% each. Memory type tells a similar story, with DDR4 now making up 54.5% of validations, up 6.0%, while DDR5 slipped 2.9% to 32.9%.

Of course, these figures only cover systems that users actually submit to the validator over the last three months, so they aren't a full picture of the PC market. Steam's September survey, for example, showed 32 GB overtaking 16 GB as the most common configuration among gamers, but that promptly flipped with the October survey. Still, the timing makes sense. Memory prices have climbed sharply this year, and Team Group expects the worst of the shortage to land between Q4 2026 and Q2 2027. DDR4 platforms may also get a second wind, with DDR4 motherboards getting released for unannounced LGA1700 processors expected in early 2027. It looks like plenty of people are sticking with older DDR4 systems or settling for 16 GB instead of paying current prices."
 
erek said:

"Memory shortage spurs smartphone makers to melt used memory off old devices, then re-solder it onto new products — recycled memory to keep budget phones affordable; 'reballing' adoption to double by 2027​

​

Recycled chips gain momentum."

https://www.tomshardware.com/pc-com...fordable-reballing-adoption-to-double-by-2027
Click to expand...
Great just what the tech worlds needs. shits really going to die left and right now because you know this will expand to every thing other then Data centers before long
 
On package HBM cannot be desoldered so they cannot do this that way... but they will keep those H100 around forever as a way to "recycle" it, they will recycle more than the ram they will keep the whole thing and thinker with it, the useful datacenter life cycle of those HBM TPU-H100 will be way longer than cheap phones.

We are talking 10+ years datacenter hardware versus fast 2-3 years cheap phone cycle, they made 800 millions a year of those super cheap found for a long time (with hundreds of millions a year of the mid-high end has well) versus very few of the former that about never stopped to be used yet to have something to try to salvage-recycle from (v100 that will turn 10, still running, same for early generation of TPUs).
 
Last edited:
LukeTbk said:
On package HBM cannot be desoldered so they cannot do this that way... but they will keep those H100 around forever as a way to "recycle" it, they will recycle more than the ram they will keep the whole thing and thinker with it, the useful datacenter life cycle of those HBM TPU-H100 will be way longer than cheap phones.

We are talking 10+ years datacenter hardware versus fast 2-3 years cheap phone cycle, they made 800 millions a year of those super cheap found for a long time (with hundreds of millions a year of the mid-high end has well) versus very few of the former that about never stopped to be used yet to have something to try to salvage-recycle from (v100 that will turn 10, still running, same for early generation of TPUs).
Click to expand...

“AMD Unveils MEXT Memory Solution: Leverages Flash To Mitigate DRAM Costs While Expanding Capacities & Retaining Similar Performance Even With 50% Cut In DDR5​

Hassan Mujtaba
•
Nov 10, 2026 at 12:00pm EST
Add Wccftech on GoogleComments
AMD-MEXT-728x359.jpg

AMD MEXT is a brand new memory solution that utilizes Flash to mitigate DRAM costs and supply concerns across EPYC CPUs.​

AMD MEXT Aims To Address The Ongoing DRAM Crisis Through An Innovative "Flash-Based" Memory Solution​

The memory supercycle continues to rage, and supply/demand continues to worsen. Today, the DRAM cost alone makes up 75% of the total server cost, and DRAM manufacturers face shortages that show no sign of ending.”
https://wccftech.com/amd-mext-memory-solution-flash-repaces-dram-same-performance/
 
erek said:

“AMD Unveils MEXT Memory Solution: Leverages Flash To Mitigate DRAM Costs While Expanding Capacities & Retaining Similar Performance Even With 50% Cut In DDR5​

Hassan Mujtaba
•
Nov 10, 2026 at 12:00pm EST
Add Wccftech on GoogleComments
View attachment 836020
AMD MEXT is a brand new memory solution that utilizes Flash to mitigate DRAM costs and supply concerns across EPYC CPUs.​

AMD MEXT Aims To Address The Ongoing DRAM Crisis Through An Innovative "Flash-Based" Memory Solution​

The memory supercycle continues to rage, and supply/demand continues to worsen. Today, the DRAM cost alone makes up 75% of the total server cost, and DRAM manufacturers face shortages that show no sign of ending.”
https://wccftech.com/amd-mext-memory-solution-flash-repaces-dram-same-performance/
Click to expand...
AMD is announcing today that MEXT is currently in Phase 1, which means it's ready for deployment across AMD EPYC CPUs with abundant cores and servers with dedicated NVMe drives or bays.

In the Phase 2, MEXT will be available to license, and will come with RAM View pre-installed.

MEXT features three main components:
  • Linux Driver: Lightweight, low-overhead kernel module that works with any Linux OS
  • AI Engine: Ensemble of AI models (including transformer-based LLMs) — can run on an Epyc core
  • RAM View: Observability platform illustrating real-time memory utilization
 
  • Like
Reactions: erek
like this
Marees said:
AMD is announcing today that MEXT is currently in Phase 1, which means it's ready for deployment across AMD EPYC CPUs with abundant cores and servers with dedicated NVMe drives or bays.

In the Phase 2, MEXT will be available to license, and will come with RAM View pre-installed.

MEXT features three main components:
  • Linux Driver: Lightweight, low-overhead kernel module that works with any Linux OS
  • AI Engine: Ensemble of AI models (including transformer-based LLMs) — can run on an Epyc core
  • RAM View: Observability platform illustrating real-time memory utilization
Click to expand...

When paired with MEXT, Physical DRAM can net a 3x return in capacity at much lower costs than what it takes to add DRAM.

In Agentic AI particularly,
  • using a 50% DRAM solution with 50% MEXT Flash nets:
    • a 90% increase in perf/$ within a 5-Agent workflow, and
    • a 70% uplift in memory-hungry agent use cases.
  • A mix of 25 percent DRAM and 75% MEXT nets
    • a 3.4x increase in the same 5-agent concurrency task, while
    • a memory-hungry agent nets a 3.5x better perf/$ ratio.

In EDA (Synopsys PrimeTime on 6th Gen EPYC CPUs),
  • you can retain up to 95% of total DRAM performance even with a 50% cut in DRAM using MEXT, leading to a 1.9x perf/$ ratio.
  • A 75% cut mixed with MEXT nets you 80% of the performance and a perf/$ of 3.2x versus the standard all-DRAM configuration.

On Azure VMs running Spark TPC-DS (5th Gen EPYC), MEXT enables up to 99% performance on a 3-node cluster when running a 50% DRAM/MEXT split versus an all-DRAM baseline configuration.
 
  • Like
Reactions: erek
like this
DDR4 for me: 5995WX + board + 256GB 2933MHz ECC RDIMM cost me around $4500. Versus $6-$8K for that much DDR5 RDIMM alone. I just can't wait around for DDR5 to be available again. At this point I'm not expecting it until DDR6 is well into production.
 
You must log in or register to reply here.
Back
Top