Discussion in 'Article Discussion' started by Gareth Halfacree, 6 Aug 2013.
Crossbar Arrays proved in the fab.
Although the first steps have been made, there still remains arguably the most important part that is never addressed in these breakthroughs: Yields.
What are the yield amounts in these products. What is the possibility of mass production?
AMAZING!!! Can't wait to see the first batch of storage products based on this new RERAM technology.
They ain't saying - but the production process is based on existing technologies for CMOS, which are well-established - meaning that, in theory, yield should be high. In theory.
Bear in mind that the yields for flash were initially tiny. The capacities were were also minuscule (I remember having just 8mb to store my first digital photos on). It's taken a decade or so but manufacturing has stepped up to deliver orders of magnitude more flash than there was capacity for in the past just because the demand was there.
RRAM sounds like it could be just as transformative as flash was but with the bonus that manufacturing processes for it are more or less already in place. It could be as soon as 2015 that we start to see RRAM based cameras and phones which would be pretty quick compared to previous technologies.
Sounds interesting and really good for people who want large storage is a small amount of space.
Could this mean the likes of 4TB pen drives being sort of just around the corner, or even 100TB SSD type hard drives as well.
Certainly going to be interesting times in a few years.
Goodbye overly expensive SSDs.
The more interesting potential is that RRAM means that your SSD and RAM will be combined into one slot and you'd have instant boot because it's all non-volatile!
Given that each chip can be 1TB and the typical number of chips on a stick of DRAM, that's some pretty exciting possibilities right there!
Also think about the data integrity potential as well; if your RAM is non-volatile, power cuts will/may no longer result in lost data...
Quick revolutionize all the RAM!
Now that is seriously interesting, imagine having a card which has 100TB on it and imagine being able to choose yourself how much you want as system ram and the rest is storage.
I have a sneaking feeling we may not see this tech in consumer space for quite a while, and it's not to do with yields and production. If those claims are to be believed, ReRAM products could profoundly affect the server markets, after all, greater than HDD storage at greater than SSD speeds with lower power draw is phenomenal. Not to mention the numerous advantages of being able to house terabytes of RAM with such a small footprint.
The thing is though, this so far outstrips what the average consumer has a need for. The storage is great, that would definitely be useful, but most people certainly don't need hundreds of gigs of RAM, the software simply won't need it for a long time. Also, if the speeds are to be believed, as was said above you could combine both the SSD and the RAM into one device that slots into a motherboard. That's basically removed an entire product line, either traditional SSD storage, or DRAM. Companies would clearly want to prolong the life of their traditional RAM products, or risk going out of business as it's taken over by another product. This isn't usually the case given the incremental improvement nature of PC tech, but the gap here is so big, all sorts of things could happen. The last thing I'd like is to see memory further slip in a similar fashion to graphics, with only 2 massive "competitors" (not far off as it is).
The moment they make anything successful and profitable, 50 patent troll companies will appear to sue them into oblivion.
Just because you can fit 1TB into the size of a postage stamp doesn't mean that this will be the only units sold; as you say there will still be the market for (e.g.) 4GB of RAM for general use. The benefit of ReRAM here is that this 4GB will be in the form of a single chip rather than 2 or 4 (relatively) large cards. Think of the benefits for laptops (and tablets of course) where the bulky and fiddly SODIMM cards and the irritating mount can be replaced with a single chip card the size of an SD card... similarly the space saving reducing the size of an SSD or HDD down to a postage stamp...
Alternatively as already suggested the single chip memory can simply be partitioned into a RAM section and a storage section - hell do away with "RAM" completely and only use the Windows Page-File or Linux Swap partition on the one volume. With the speed of ReRAM you don't even need the completely separate memory hardware to get the same effect so you might as well do it all from some sort of swap file/partition
That is an interesting thought, but what I find a little more interesting is how system RAM in general could be obsoleted. In other words, you read from the disk, dump to the CPU cache, and go straight to calculating. One of the only reasons RAM exists is because there wasn't any realistic way to skip it while getting things to process in a timely manner. Even in the days when people only had 1MB of RAM total, it could probably take days to calculate anything useful without RAM. But if reading from the disk is equally as fast as RAM, I would think all you'd need is a specially designed CPU with a little bit more cache than usual. Skipping out on RAM altogether would also mean a re-write of OSes, or at least some dirty hack to emulate RAM temporarily, much like swap space.
Expanding upon the idea of removing RAM from the system, there's also the idea that maybe 512MB of regular SDRAM could be soldered onto the motherboard, where it could be used for IGPs and getting a modern OS to boot so you can create a paging file or swap partition. Basically, this RAM would act as a transition until OSes could phase out RAM entirely. Interestingly, it would also allow you to run your computer with basically just a CPU and a power source.
Sit's wondering how long it will be before Apple claim to have invented it.
In all fairness the scaling down of this to 4-8GB chunks or say 16GB chunks would lower costs massively providing this can be made. Even a 32GB module would be hypothetically easier if this technology were to be bought up (which if it does work will, whether or not it'll be used is up for debate).
What this means is that although we will not get 1TB modules, we can definitely aim lower and still increase our NVRAM density by at least four to eight-fold without potentially sacrificing quality or speeds.
If this works.
Yes I clearly understood the benefits, as shown by the rest of my comment. My point is more to do with how because it's simply so much more than we need, it will be artificially limited by the companies involved. This is a product that if it lives up to the hype, will no doubt remove RAM from the market. That's a lot of lost sales right there, after all, the same companies all produce both SSDs and RAM and sell both, by removing one, the sales of the other won't magically double.
By the time these 1TB chips will be viable (which by the sounds of things isn't that far) what's the point in producing 4Gb chips? So your standard computer will no longer have that need, they'll just chug along with their terabyte super RAMdisk effectively. Now on our end that's fantastic, more storage, less power and more speed, what more could you want? Exactly. People are already reaching the point where upgrading to recent top notch gear is no longer even worth considering. My parents only use their computers for emails, word documents and the occasional bit of video streaming, they won't need to get new systems for yonks. The same goes for offices, why upgrade a whole system if actually there's no real benefit? This could no doubt lead to a further slowing in the desktop market, especially when compared to the mobile market, which would continue to see incremental advances for quite a while and therefore strong sales.
Could be interesting to see how this all plays out really.
I agree that the initlal impact of this tech will likely be on the server side. Imagine what Google must think of a tech like this. Now imagine the rest of the large / mid-sized companies that have servers everywhere, being accessed constantly. Heck, it might not be out of the realm of possibility for companies to start creating diskless environments again, with all work being done from a large server farm (via the VAX and IBM clusters of yester year ...)
As far as consumers needing space ... hell, if you build it, we will fill it. I don't *need* terabytes of storage at home, but I have it, and I'm filling it, and constantly having to buy more storage for backup / recovery / imagery / video / etc.
I'm thinking future systems will no longer be "buy mobo / chip / RAM / HDD / psu / gpu / case." Designed properly, no only does this tech remove the HDD / RAM into a chip that can be onboard, you also have MUCH better integrated graphics capabilities when you combine the performance / heat benefits nVidia's new 'Logan' chips are going to have. Just SLI a couple of those chips and you've got low heat, great performance, and heavy speed of the RRAM tech. All running from a 9 volt battery ...
I'm guessing a RAM maker will buy this tech if possible and then release it scaled down. For market purposes. -_-
Not holding my breath to see this technology being manufactured and sold within the next decade.
There was alot of companies with such claims before.
Separate names with a comma.