SSDs Have Bleak Future, Says Researchers

[garyshome]

Wait this is a Grad Student? Am i missing something here? Well maybe if this came from a manufacturer it might carry more weight. Student? Not Engineer or Professor or anybody with any practical or real life experience. Come on Tom's.

 

[yumri4]

@dimar um huh? why would a RAM drive need to be in a PCI-E 3.0 slot as that is slower then the RAM slots as the latency is longer. and a RAM drive is a drive made out of RAM so you would want to make a card in a PCI-E 3.0 slot which takes DDR3 RAM? firstly why decrease the speed to only 200MB/s as DDR3 can go a heck of alot faster then just that. then like just get a PCI-E 2.0 x8 SSD drive as it will have what you want and it will go at least double that speed or at least they claim to but cost several grand each as mainly they are a couple normal high performance SSDs in RAID 0 or RAID 0+1 or RAID 1 depending on which you are looking at and how you set it up.
So a better solution to you will be to get a computer with like a LGA 2011 or LGA 1366 processor and put in 20+GB of DDR3 1600 or faster in it then make the RAM drive out of the system RAM cheaper faster and probably the entire system will go faster because of that solution and the kind of CPU you will have to use in it.
But that is a RAM drive not a SSD slightly different technology and the controller works differently from what i understand anyways.

 

[A Bad Day]

And that's why we need to research and develop new non-volatile memory systems. Sticking with Flash memory past 2020 is like clinging onto DRAM2 for GPUs when your competitors have already moved onto GDDR6 and DRAM4...

 

[drwho1]

[citation][nom]yumri4[/nom] @drwho1 " With 10-12, 16TB drivers on ONE tower, who needs a NAS " well most NAS servers have a lot more than just a little 16TB amount of disk space also ussualy include RAID and still have a larger amount. Yes i know not all do but many mroe expenive ones do some even have over 1EX of disk space which i cannot think of any reason why a single company and/or person would need so much space unless they do not delete anything for a few years including DNS records and back-ups of servers.[/citation]

I'm NOT a company, I do have a tower with 14TB across 9 drives, FULL of my movies/TV shows.
This is nowhere enough to accommodate my whole current and growing library of DVD's.

I don't own a NAS, but I have been thinking (and probably will build one) in the future.

When?
When (A) the hard drives have reached at least 5TB or bigger, and (B) when said drives cost no more than $100 a piece. Then I would put an 8 drive NAS together, as long as I can run/connect multiple of same to my main PC.

With that said, this is the second article in the same week that mentions 16TB drives, in other words, this size drives might already be in the works, and I wouldn't be surprised at all if we get to see them within the next decade.

Add a few short years after that, they will be priced "right" for the mass market.

I'm more concerned with what kind of "bios" we will need for this drives.
Will UEFI recognize this drives, OR will it be something totally different?

 

[dimar]

[citation][nom]yumri4[/nom]@dimar um huh? ... [/citation]

I meant as long as SSDs are about 200MB/s, it's fine with me. But I'm having second thoughts now. I don't want SSDs to be less than 500MB/s read/right actually

Now, I'm looking at ADATA XPG Gaming v2.0 Series DDR3 specs "Blazing speed at DDR3 2400+, the transfer bandwidth can reach up to 19,200 MB/s (PC3 19200). With dual-channel design, the memory bandwidth can reach up to 38 GB/s".

PCI-E 3.0 x16 can reach 16 GB/s, so even that won't be enough.
According to wikipedia, PCI-E 4.0 is 2GB/s per lane, which makes it 32GB/s at x16. So even that won't be enough Is this all correct? or am I getting things wrong?

 

[yumri4]

@drwho1 my point wasn't that haveing 16TB was a company thing my point was that a 16TB NAS is/was all ready around is all it is just price of it and stuff as 16TB right now of HDD for SATA, SAS, iSCSI or whatever you prefer to use for your hard drives is going to be alot of money to do in a NAS. Also with that SSDs ( which this article is about ) will probably NOT get to 16TB any time soon even though tape storage and HDDs probably will either by expanding the amount of platters inside of the casing and/or making each platter able to store more data on its surface.
@dimar the thing about RAM Drives which is bad is once you turn off the computer the drive is wiped so you will have to remember to back it up completely before turning the computer off or having a back up protocol interjected into the shutdown routine of your machine for the RAM drive.

 

[bison88]

On the other side of the argument that Laura Grupp and her partners are discussing, is that NAND memory will hardly be the leading technology by that time, that's if things work out for the best with future technologies such as MRAM and PRAM. There are at least a dozen technologies still being looked at or re-looked at with a good half-dozen getting some serious attention and backing. The quest for that all-in-one memory technology is far from over.

While I'm not a huge advocate for SSD's consumer wise, I'd like to see them compete with traditional HDD technology, if not just for the simple fact the HDD market has went from a couple dozen down to a handful (believe 3 now) and SSD's still have that freshness bringing about dozens of companies producing there own products since just about anyone can order memory chips by the bulk and a controller without having to create dreary fabs in a research environment. Problem I see personally is the limited write endurance. It honestly scares me how its limited with SLC and goes down drastically with MLC and TLC. I know it'll last years even with the massive stress I put on my drives for read/writes but it really does concern me and in a way I'm glad these folks are at least bringing that issue to the forefront once more. TRIM helps but it doesn't solve all of the problems by a long shot.

While these researchers are kind of stating the obvious for most people who have done there research, I do believe there is something to be said about the overall endurance of the drives themselves. After all it would be damn near perfect even with the price of SSD's if they didn't have at least one negative thing about them. Then theres that quest for the all-in-one memory design again. I feel this story and the researchers skipped out on covering that believing NAND will be the only technology in 2024. Many are expected to marketable well within the end of this decade, again, if all goes well.

 

[belardo]

And yet Hard Drives and CPUs have PASSED their limits as well... okay, there is only so much we can shrink tech... but hey, maybe we can get sub 1nm CPUs in 10 years?

2024... thats 12 years from now. The PC (not clones) of 1980 are nothing like those made in 1992.
In 12 years, a lot can change... and most likely, we will not be using NAND... who knows?

 

[teodoreh]

Dears sirs,

I made a very reliable research under the assumption that all R&D labs working on memory storage will stop researching new technologies, and will only focus on lowering the scale of circuits.

Yours truly,
Someone unknown.

LoL

 

[yumri4]

@bison i agree with you on that and really i do not see much use of a SSD in the consumer space except for gamers and enthusiasts. For the main reason of the HDD is bigger and fast enough for the general use of the consumer of Internet, word processing, and light graphics usually from a .jpg, . gif, and/or .png file which doesn't really stress the hard drive that much if at all thus the performance gains from a SSD is not that much noticed by a consumer. with the thing of the ever going and probably never stopping search for the all-in-one memory design me personally i do not think one exists but one design per spefic function for example one design for a CPU but a different one for RAM and a different one for a disk storage device and etc.
Again 2024 is a long time from now in the computer world and SSDs will probably already have started to be replaced by then if not already replaced by another technology.
@teodoreh that is a stupid assumption to make when doing research as it most likely will not happen and even though yes the scale of the circuits will continue to shrink to the physical limits the power, performance, and room for allocation space gains will most likely scale up also.

 

[punahou1]

This is one of many problems associated with silicone based technologies. Bell Labs is currently researching organic based technologies which will resolve this and all other issues associated with miniaturizing silicone. As a result, the SSD will most likely evolve into a different form. Nothing to loose sleep over but something to definitely consider in terms of a very long term investment strategy...

 

[yumri4]

@punahou1 yes but how long have they been research an organic alternative and nothing has turned up as yeah i remember this research from 4 or 5 years ago and nothing has happened which i can tell but they did find out alot of what does NOT work but barely anything which does work.

 

[daship]

Buy 2024 SSDs and nand will be a fond memory. Technology will advance to something even better.

 

[Guest]

12 years are really way too long a time to predict any trend in computing storage in reliable way. How many different technologies have come and gone in the last 30 years? Tape, floppy, optical drive, ZIP/Jaz, CD, DVD, Bluray, flash drive. 12 years ago, no one has known we will be carrying a thumb drive so small that has way more storage than the HDD used at that time. Who would have foreseen the demise of ZIP/Jaz while it became popular in the 90s?

 

[freggo]

If you look at what various companies admit they are researching you can -or can not- imagine they stuff they are keeping very, very quiet about !!

 

[warmon6]

[citation][nom]4745454b[/nom]Farthermore I'm sure we've heard this kind of talk all the time about similar and other things. Can't go above 1GHz, can't go above 640kB, can't go above 4GBs of ram, etc. There are always plenty of can't and shouldn't people out there. 2024? Serious? Any idea of what we will develop between now and then? Think of it this way, what did we use for computers 12 years ago back in the year 2000? Did you ever think we would have 4 and 8 core CPUs with 24GBs of ram, 3TB hdds, and the ability to use 3 GTX580 class GPUs together? Care to guess what we will have 12 years from now?[/citation]

Well the GTX 580's couldn't been though of in 2000 but we did already have SLI (from 3DFX) at that time,

http://en.wikipedia.org/wiki/Scan-Line_Interleave

 

[Wamphryi]

[citation][nom]Tomfreak[/nom]Sometimes I wonder, desktop HDD are not limited to power consumption, but why arent the HDD manufacturer push 10K RPM HDD down to mainstream to "slow the SSD" adoption . Surely the latency gap would have improved significantly, 8.9ms vs 4.0 seek time. It may not reach the SSD ones, but it is still better to delay the SSD getting into mainstream. While many of us like the capacity of HDD, but we I would prefer having slighly smaller 500GB with boost speed over the 2TB ones. The size of the mainstream SSD are still less than 64GB, win7 64bit takes up almost half of it, so the leave not much room for games/other thing.[/citation]

I have two Velicoraptors in RAID 0 and they cant compete with my Crucial M4 in terms of performance. The Raptors are very useful for video editing scratch drives and indeed there is more capacity on them but that is really the only advantage they have over the Crucial M4. I put a Crucial M4 in one laptop and a Vertex 2 in another laptop and the difference was night and day. A 128 GB is more than enough for applications and games on a well managed system and 60 GB is enough for a well managed laptop imo.

SSD = OS, applications and games

10000 RPM HDD = Video scratch drives, encryption work in RAID 0

7200 RPM HDD = Mass storage of data in Raid 1 with reasonable performance

5400 RPM HDD = External HDD use due to lower power consumption and less heat. Hot swapping for daily back ups.

 

[vic20]

You mean in the future someone uses my time machine to determine the fate of SSDs? GREAT SCOTT!!! I'm going to detroy my time machine right now!

Seriously though, if latency becomes an issue there is RAID 0 on machines and even the drives themselves for a reason.

 

[Rock_n_Rolla]

[citation][nom]4745454b[/nom] Did you ever think we would have 4 and 8 core CPUs with 24GBs of ram, 3TB hdds, and the ability to use 3 GTX580 class GPUs together? Care to guess what we will have
12 years from now?[/citation]

- 12 years from now? Umm... By that time, Intel had released their 4ghz 10nm 16-24 core 4ghz
processors. AMD are now on its 11900 something series Radeon video cards with GPU clocking at 2.4
ghz with Gddr ram capacities of 8gb @ speeds of 8+ghz. Smartphones and Tablets now have quad
core i7 like processing power and desktop like performance and now capable of processing hi end 3d
applications and games like BFD3. Laptops can have the processing and rendering power of a very hi
end 8 core sandy bridge E Xeon i7 x79 server/desktop. HD 1080p TVs are now a thing of the past
and the 4K resolution (taken from Red Epic cameras) are now the "de facto". PC monitors now
supports 4K resolutions with higher refresh rates for ultimate clarity and viewing performance.
Digital cameras can now take pictures with the performance and clarity of a hi end DSLR and DSLRs
can now take pictures ranging from 30-50 megapixel with less noise. On the streets you'll see lots of
hybrid and electric cars and regular sedans and pickups now uses Carbon fiber as their standard
material. ISPs now offers 300-500 megabits of internet connection. File sharing and video streaming is
much much easier and faster. Graphics of FPS PC games are now so life like.

Tom's hardware would still Tom's hardware and SOLID STATE DRIVES uses better technology and
innovation and CHEAP as cakes AND more reliable than hard drives...

 

[gallidorn]

Any story that starts by referencing a study performed by a Graduate Student, shouldn't be taken too seriously.

They are basing the so-called limits on a very narrow point of view. The study doesn't take into consideration the out-of box thinking that creates advances in technology. It only references what they know or assume at this time. It is impossible to used today's technology to establish limits in the future, because there will surely be many substantial advances in memory technologies over the next 12 years!

Look at the advances over the past 10 years and now imagine what we have to look forward to.

 

[BattleshipLorenzen]

Are any of you academics? Experienced graduate students are as variable (good and bad) as professors in their research output and quality. There are plenty of Ph.D.'s doing terrible research. This is not only a genetic fallacy, but a bad one. Her work would have been overseen by a Ph.D., whose influence ranges from modest to "nothing gets out without her/his approval". Many TH readers appear capable of evaluating the reasoning used in coming to the conclusions; stick to that. Knocking research because it comes "from a graduate student" is just silly.

Anyway, the researcher's statements in the linked article do surprise me. I wonder what the academic paper actually said. Usually a reviewer will gut you for saying something like "the end of SSD's" when you really mean "this particular SSD technology appears to have this limit, given these assumptions." Perhaps the article took the statements outside of their context (e.g. the limited scope of the implications may have been obvious in the paper). Alternatively, maybe they really were overhyping their research to the media. That gets frowned upon by peers (but not all).

Agreed with those who say "fast enough!". TH linked an article somewhat recently about a research group who thought they might have a technology that uses lasers to write to HDDs. Maybe those will be implemented by that time.

 

[BattleshipLorenzen]

Rather, I should have said "Are any of you familiar with academia?" I did not mean to imply that being an academic was, in any way, a wise life decision.

 

[DSpider]

By 2024 flash storage will be analogous to floppy disk magnetic storage.

 

[starcore]

Ill worry about it in 2024, but it is nice to know. I want to thank all the researchers and tom's hardware for doing the work and posting it.

 

[Guest]

OMFG.... So for 12 years I get superb performance compared to HDD, with prices dropping and size increasing every year until some other technology comes along costing $10 per GB and the uptake fo that decreases our dependence on SSD

BOO HOO HOO!!!!!