Monday, August 25th 2025

ULTRARAM Nearing Production to Bridge the Gap Between Memory and Storage

Quinas Technology and IQE plc announced this week that ULTRARAM, an experimental memory designed to combine DRAM speed with NAND-like persistence, had cleared a significant manufacturing hurdle and is moving toward pilot production. After a year of close collaboration, the partners have developed an industrially scalable epitaxy process for the compound semiconductor layers that ULTRARAM relies on. Both companies now report discussions with foundries and other partners about pilot wafers and packaged chips. Innovate UK funding and recent intellectual property recognition helped push the project forward, and the teams describe the result as a key step from university lab work to commercial memory products.

At the technical level, ULTRARAM relies on resonant tunneling within precisely grown layers of gallium antimonide and aluminium antimonide. The new epitaxy passage provides engineers with atomic-layer control, allowing standard photolithography and etching to form reliable memory cells. In lab testing the design has shown extremely fast switching, very low switching energy, and data retention measured in years rather than hours. If those characteristics can be reproduced at scale, systems could eliminate DRAM refresh power in servers, cut boot and resume times for client devices, and enable always-on features. That hope comes with a dose of realism, because moving from a scalable epitaxy method to profitable, high-yield manufacturing will require strong supply-chain partnerships, careful yield optimization, and system-level integration. All of that will be demonstrated in a pilot run, which will serve as the final true test.
Add your own comment

12 Comments on ULTRARAM Nearing Production to Bridge the Gap Between Memory and Storage

#1
RejZoR
It's gonna be cool when we'll just put 10TB of this into the system and it will be both storage and operational memory all while being non-volatile. System crashed? It'll boot up in exactly the same state it crashed at (or shall I say before).
Posted on Reply
#3
Unregistered
I'm polishing my balls for the coming bright future!!!

Hey People!!!
Let's not forget about RaceTrack memory, this one is coming soon too!
Posted on Edit | Reply
#5
DeathtoGnomes
user556It's not a RAM substitute, endurance is only estimated to 10 million erases. It'll never replace DRAM. en.wikipedia.org/wiki/UltraRAM
Endurance is not yet proven, 10m is, after all, still in theory at this point. Its targeted for Enterprise servers, not PC, that could change quite a few things like power consumption and WaaS eventually.
Posted on Reply
#6
user556
DeathtoGnomesEndurance is not yet proven, 10m is, after all, still in theory at this point. Its targeted for Enterprise servers, not PC, that could change quite a few things like power consumption and WaaS eventually.
It's kind of irrelevant whether the value is 5m or 20m. None of the ranges being talked about are vaguely close to DRAM, which is practically unlimited. MRAM is the only non-volatile capable of being used as a RAM for real. All the others, including this "ULTRARAM", and Flash memory, are better classed as a ROM tech.
Posted on Reply
#8
user556
10^7 is 10 million, so no change in that claim. 10 M writes to DRAM can be done in just a few seconds. Retention is only relevant for ROM functions.
Posted on Reply
#9
Unregistered
user556so no change in that claim.
You don't need to change any of your claims.
I'm just saying that: In the entire world there's no ROM capable of stationary retention for about 1000 years.
So the final question is: - Who says that UltraRam in it's final version won't be capable to play as RAM too?
#10
user556
Until that day comes, claiming it to be a DRAM replacement is just plain lying.
Posted on Reply
#11
Unregistered
user556Until that day comes, claiming it to be a DRAM replacement is just plain lying.
I think that in it's final form this memory is going to be deployed :
as replacement for Ram based on capacitors, as replacement for SSD based on flash/nand and only after some years, they will find a way to unify everything in one storage (maybe in 2030?).
#12
user556
In the server market, it has to compete with Flash price and density. That's a tall order for any new storage! Most new entrants with excellent retention have made inroads better in hash environments like space/transport/farming/military.
Posted on Reply
Add your own comment
Nov 14th, 2026 18:53 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts