Symetrix/CEFE Labs versus 4DS ...................
- As manyknows, CEFE Labs is a joint venture company owned by ARM and Symetrix that has been set up to developp a new non-volatil memory called CeRam. Ceram has been invented by Professor Araujo of Colorado Uni, more than ten years ago.
- After morethan five years of exchange and research with ARM, Professor Araujo, who is theSymetrix CEO, finaly licenced the technology to ARM and they decided to create CEFE Labs to bring the CeRam technology to commercial level.
- Beside being(apparently) a serious competitor to 4DS, the CeRam venture presents manyfeatures interesting to understand the 4DS situation/strategy.
- First, where4DS like to stay under the radar, The CEFE Labs creation was announced throughan organised PR campaign. The ARM CEO made his announcement october 2 2020, thesame day lots of papers were published in the technical media. The memoryGuy published a very long and interesting paper on Ceram this same day. A paperlike this will need a few days of research, even for them, but it wasbupblished on october 2.
- So, solid PRstrategy and money in place. But the most enlighting differences between 4DSand CEFE Labs is their developpement strategy.
o Where 4DSkeeps her cards close to her chest, ARM and Symetrix gave the technology to betested by many Fabs and Labs (among wich, Imec that made a few wafers forSymetrix). So a larger and immediat validation of basic principles.
o CEFE Labs (asSymetrix), is a private company. They dont have to make CR and fight/explainthings to pesky shareholders. The parent companies : ARM and Symetrix gavehundreds of patents and millions to CEFE Labs.
o CEFE Labs istechnicaly more independant. They can make their own 300 mm wafers at home (orat parent Symetrix Lab). ARM gave them the equipement (wich means also Cleanroom, equipments for ultra pure water and other chemicals storage and disposal=> 10 Millions, 20 millions ?)
- The otherbig difference is the targeted market. CEFE Labs will first target niche and embededmarket and 4DS want to reach the mass storage and Data centers market. But CEFELabs also said they will leave it to Fab licensees to advance the technologyfor mass storage… Or even S-RAM. Big difference with 4DS who seems to want to giveFabs a complet and detailled recip to produce stacked layers of cells almostimmedialtly, the only things to add being in-house CMOS and packaging.
-Onesimilarity between the two projects is that neither CEFE Labs nor 4DS want tomaufacture their product. They will leave it to large manufacturers.
Nowabout the respective technologies
- There are morequalified people here to compare the two technologies, but I will try to do it,expecting to be corrected by others.
- At firstglance, the CeRam technology seems to be impressive and well advanced. Theswiching mecanism is ultra fast : about 100 femtoseconds or less. So muchbetter than 4DS (in nano seconds). But the real speed will be less than thatbecause it will be limited by the CMOS speed, and CMOS is needed to allow thememory to work (the same logic could works for 4DS, though).
- The CeRamtechnology seems more elegant, with bulk change and no filament or oxygen atoms exchange. It seems closer to Optane solution Phase change memory (the applied current will change the sandwiched layer from amorphous to cristalin/ conductive and back to amorphous) except that CeRam solution will be much faster.
- The CeRamsolution works at high temperatures (Keeps information up to 400° C)
- CeRam isbulk change. All the intermediate part between the top and the bottom of thecell will become conductive. So no filament. Professor Araujo made it clearthat CeRam is completly distinc from what he calls the "filamentaryparadigm".
- 4DS seems tobe in the same situation as it has no filament, but it is unclear if it couldbe assimilated to bulk change.
- It is alsounclear at what level of developpment the CeRam technology is.
- The ARM CEOspeaks of 18 month to commercial level, but the CEFE Lab CEO speaks about 18month "to prove that the device with increasingly small geometry operateas bulk switches and then verify with commercial manufacturers the reliabilityand repeatability of the device in fab-compatible integration at fabdimensions".
- As explainedin the Memory Guy paper there are people that says that in fact CeRam do createa filament. That point doesn't seem to be clarified at the moment : As perEenews analog paper ...nonetheless, with decresing geometry it isdifficult to be absolutely sure of the bulk switching effect. And CEFE Lab CEOsaid "Until we have something definitiv like X-Ray confirmation, we keeptouching different part of the elephant and every time it is confirmed an elephant".
- +++thememoryguy.com
https://thememoryguy.com/ceram-moves-front-and-center-on-the-nv-memory-stage/
- +++ spectrum.ieee.orghttps://spectrum.ieee.org/nanoclast/semiconductors/memory/arm-spinout-details-correlatedelectron-memory-plans
- So, theCeRam solution might not be as advanced as some says. In many ways what ispublished about CeRam looks like the datas 4DS published on the early days(2014)
- A paperposted on EET Asia in 2014 (at a time when 4DS was still giving data the media) gave the first comparisons between the twotechnologies. It says that 4DS also claims that the operation of it's memory isbased on strongly correlated electron-effect Mott Like transition. The CeRamoperation rely also on the same Mott effect. The difference being that theactive material (between the top and the bottom of the cell) is praseodymiumcalcium metal oxyde or praseodimium doped with calcium. The CeRam activematerial is nikel oxyde doped with carbon (they are testing other materials).
- +++eetasia.com
https://archive.eetasia.com/www.eetasia.com/ART_8800697963_499486_TA_b9a57999.HTM
https://archive.eetasia.com/www.eetasia.com/STATIC/ARTICLE_IMAGES/201405/EEOL_2014MAY14_STOR_TA_01Fig1.jpg
Witchone will win the race
- If at firstglance the CeRam solution seems to be well advanced, the fact that they arestill trying to figure out what is happening at smaller geometry (Bulk effect ornot), indicate that still have some work to do.
- And there noclear information wether CeRam will stack easily. Because they said that theywill first target niche market an embedded, it seems to indicate that they willleave multy layered mass memory for later or more probably for bigmanufacturers licencees to developp it.
- All in all4DS seems to have at least two years advance on CEFE Labs CeRam. But CEFE Labshas other products cooking at the moment and they may also like some secrecy andeception. They may surprise everybody with another discovery or a spectacularprogress. Be sure that this will be announced with a powerfull PR operation.
Conclusions
-The moststriking difference between the two is the power and money behind CeRam. 4DSlooks poor and vulnerable when compared with CEFE Labs. This underline the factthat 4DS can't do it alone (imo). 4DS must rely on HGST, SanDisk and WDC (Imo, 4DS isalready part of WDC sphere and will be absorbed, one way or another).
-Thecommercial strategy to impose the thechnology seems more aggressive on the CEFELabs side. They might not be close to a commercial product, but be sure thatanyone will look first on their direction for futur SCM.
-That mightnot be enough to win the race. If WDC comes out with an ultra fast commercialproduct based on 4DS ReRam, it might take some very strong positions on SCMmarket before CeRam emerge on this market.
-The finalquestion is : what will be the choice ofWDC for their futur SCM ?
-We know thatSanDisk had a ReRam solution developped with HPE before being bought by WDC (2layers, 32 GBT, 24 nma).
-We know thatHGST had a PCM solution (phase change, like Optane). We don't know wich one isthe most advanced at the moment, but we know that ReRam is the prefered choiceof WDC because of feasability and commercial reasons (could be deployed inrack-scale design and have an ecosystem support).
-The WDCsolution was supposed to be revealed in 2017.
theregister.com
https://www.theregister.com/2016/08/16/wd_says_resistance_is_not_futile/
Gutfeelings = 4DS people have a real and serious advance, but now a big dog is running andbarking loudly after them.
The jury is still out to says if the dog bites.