The engineers behind ULTRARAM have made another breakthrough in bringing the new type of computing memory closer to mass production. Researchers from Quinas Technology, a spinoff company from Lancaster University, and IQE plc, a leading supplier of advanced wafer products, have collaborated to bring ULTRARAM another step closer to volume production. The new technology seeks to combine the best of both worlds of DRAM and NAND by providing non-volatile, high-speed, high-endurance memory. Additional funding from Innovate UK is assisting both companies to move forward in discussions with foundries and other industry manufacturers for pilot wafer production.
With data retention of up to 1,000 years and up to 4,000 times the durability of DRAM, ULTRARAM has a lot going for it. The year-long collaboration has resulted in a new world’s first process of gallium antimonide and aluminum antimonide epitaxy. Per Blocks & Files, “Epitaxy is the process of growing a crystalline layer on a crystalline substrate in a regulated manner, with the deposition layer adopting the substrate’s crystal structure.”
“We have successfully achieved our goal of developing a scalable epitaxy process for ULTRARAM, a milestone towards industrial production of packaged chips. This project represents a unique opportunity to bring the next generation of compound semiconductor materials to life in the UK.”
-Jutta Meier, CEO of IQE
There are numerous advantages of ULTRARAM. From data retention w/o power to speed and endurance of DRAM, to reduced power consumption, it poses to be a highly disruptive technology vs current memory solutions. James Ashforth-Pook, Quinas CEO and co-founder, stated that “at 100 ns at energies of <1 femtojoule, which is lower than any known memory device. As a candidate for ‘universal memory’, ULTRARAM has the potential to radically transform the entire digital technology landscape, from small autonomous Internet of Things (IoT) devices, through to smartphones, laptops and datacenters.”
While DDR and NAND overwhelmingly dominate the consumer front in desktop and mobile storage, guised in various form factors, with HBM and GDDR in others, similar but more powerful incarnations exist at the datacenter levels. ULTRARAM has the potential to uproot all of them, but only time will tell if it can live up to its promise and gain adoption in industries that are already so embedded with other established technologies.
Discussion (16 replies)
Join Discussion →If this catches wind... just expect someone like Samsung or TSMC to buy them... patent the ever loving crap out of it. Then just keep making DDR for a while, until they decide the cost to produce is worth it to lead the market. Or DDR ram just blows it so far out of the water as it be a non factor.
I mean if they can make it cheap and dense enough it might be nice to have a non volatile storage solution for long term backups...
Or if it's fast enough we can make all the people who never understood the difference between ram and storage right and have them be the same thing.
A sliding scale... do you need storage, or ram... just get a 8 TB supermem system and you'll have all of both!
I will say right now this product feels very much like snake water.
Yep this does seem like a too good to be true kind of thing and even if it isn't, I'm not sure how they'll be able to break through to sell it.
I guess we hear about all kinds of new RAM types that never really took off, innovation doesn't always mean it will catch on.
Fill me in though, what is the intended use of this, replace flash? replace DRAM? or sit in-between?
After reading the article... It's more like.
We got something!
What's it do?
It's fast... ummm like RAM!
is it RAM fast?
Uhhh... maybe!!
Ok so it replaces ram?
It could... but we think it will hold data for a THOUSAND years!
Yea.. like the crystal storage technology they were talking about before?
Well we do grow crystals for this!
Oh so it's the crystal storage tech?
We think it may work like that...
What do you mean?
So we think it's storage and ram..
Sounds amazing!
Sure... I mean yea it is... Can we have some money?
If it's so great don't you have a ton of private equity investors?
Well.... no?
Oh... ok sure have some government money... we have to fund moonshots too.
Prediction: RAMBUS purchases the company, proceeds to sue all the other DRAMURAI for IP infringement related to these patents until licenses are sold.
so many .... so so many.....
This will depend on final characteristic and price.
Im guessing storage side not DRAM side.
Unless cost is really good, Im guessing local accelerators or some front end for some storage systems.
Honestly it sounds like Intel Optane.
Actually that is a fair point. I'm wondering if density and cost to produce just means it can't be that good for mas market.
If Optane's now-dead future was any indicator, it seems like what they're going to wind up with is persistent RAM that's not as fast as actual DRAM. Which would probably be the worst point about it. And yes, cost is usually an issue for anything that's new and novel due to economies of scale.
I was actually think this yesterday as well.
This is exactly what I was thinking when I was reading it.
I am sad that Intels PMem Optane memory is no more as I could really use it for one of my setups. Maybe this will fill in the void.
I didn't read the whole thing, but hopefully this will use the existing DIMM slots for memory just like PMem does for more direct access to the CPU. I need this.
So what makes them different when all the others have failed?
Oh in this case... absolutely nothing.. But it's an interesting idea for something that COULD be good.
I think the only reason the others (Intel Optane) failed was due to costs. Intel couldn't get the prices down on those devices and they cost a fortune when you could get near identical performance with far cheaper hardware else where.
and by near identical I mean something that the users (for personal/desktop/workstation) couldn't really notice.
NVME really took the wind out of the sails of the Intel Optane product lineup.
If they could get the costs under control that for comparable sized products (in terms of storage capacity) for this technology in same or similar storage capacities of NVME and other storage devices then they'd probably have a lot more success.
I've been trying to find some Intel Optane PMem modules in 256GB sizes for one of my servers for a few months now. I've ordered from three different eBay sellers already. One wouldn't even sell it to me as they claimed my server didn't support it. The other two cancelled the orders because they didn't actually have any to sell me.