Will SSD's make HDD's extinct?

[MongGrel]

Who knows. Maybe they will.

Or maybe a breakthrough in read-write graphene/crystalline media makes them both irrelevant.

Or maybe a breakthrough in quantum computing or biologically based data retention makes the whole idea of binary storage with transistors a pleasantly old-fashioned idea.

Has wondered where a few of those concepts had been simmering at myself.

They seemed to be beyond conceptual a few years ago, but really had not followed it myself.

 

[Viper GTS]

There isn't nearly enough NAND production capacity to replace HDD any time soon. Even if SSD became magically cheaper than spindle (ignoring supply/demand issues that would arise) there simply isn't enough flash to go around. And it's not even close.

It's going to be a while.

Viper GTS

 

[BFG10K]

Are you talking about the completely worn out SSDs?

Nope. The fact is unpowered SSDs are unproven for cold storage given they haven't been around as long as magnetic media, which we know works.

All we have are manufacturer ratings, but then manufactures were telling us SSDs are inherently more reliable than HDDs, but that proved to be false.

Worn out HDs dont keep their data long either.

A "worn out HDD" is a nonsensical concept. HDDs don't have write cycles.

 

[beginner99]

HDDs will eventually disappear but it will IMHO be at least a decade. Most consumers are clueless an will keep buying the cheapest laptops. The will take a 1 TB 5400 RPM hdd over a 256 gb ssd. Even if they don't need the space.

Also amount of users really needing Terabytes of storage is rather small. Most people don't have huge movie collections on disk. They stream or use bluray.

I think the tipping point will be when 1 TB SSDs get as cheap as a 1 TB hdd. Then it will be enough storage for most users. Then less HDDs are sold meaning price will go up and they will slowly disappear.

Unless some magically break-though happens that makes SSDs a lot cheaper over night. But I think Intel would be dumb not to milk 3DXpoint, eg. keep it expensive for some time for enterprise use only. Maybe it actually is expensive to produce, more expensive than flash, I don't know.

 

[ShintaiDK]

Nope. The fact is unpowered SSDs are unproven for cold storage given they haven't been around as long as magnetic media, which we know works.

All we have are manufacturer ratings, but then manufactures were telling us SSDs are inherently more reliable than HDDs, but that proved to be false.


A "worn out HDD" is a nonsensical concept. HDDs don't have write cycles.

I am sorry, you just make up BS then when you talk about days :|

A worn out HD is a real thing btw. The platters simply get their magnetic substance grinded away over time. The less material, the less magnetic charge.

HDs are not a long term storage solution. There is only 1 single media for that. And that is tapes.

 

[Puffnstuff]

When I can afford it I will replace every single mechanical drive in my systems with ssd's. My laptop and tablet are already all ssd and in the not too distant future my desktop will join them. The only exception might be those Hitachi helium drives if they can actually sustain high speed writes and drop in price and I'd still use ssd's to run my programs.

 

[zebrax2]

I am sorry, you just make up BS then when you talk about days :|

A worn out HD is a real thing btw. The platters simply get their magnetic substance grinded away over time. The less material, the less magnetic charge.

HDs are not a long term storage solution. There is only 1 single media for that. And that is tapes.

That is not a worn out HDD that is an HDD that failed. Unless my understanding of it is wrong the drive head shouldn't be touching the platters at all.

 

[BFG10K]

I am sorry, you just make up BS then when you talk about days :|

A worn out HD is a real thing btw. The platters simply get their magnetic substance grinded away over time. The less material, the less magnetic charge.

LOL.

If you think a head touching the surface of a HDD platter is part of normal operation and wears out the magnetic layer eventually, you really don't understand even the most basic concepts of a HDD.

Here's a hint: a HDD is not a vinyl record. The head rides on a cushion of air and never touches the platter under normal operation.

Any head that touches the platter is classed as a catastrophic hardware failure/defect.

Now run along and do some basic Google searches to educate yourself and stop embarrassing yourself further.

 

[VirtualLarry]

LOL.

If you think a head touching the surface of a HDD platter is part of normal operation and wears out the magnetic layer eventually, you really don't understand even the most basic concepts of a HDD.

Here's a hint: a HDD is not a vinyl record. The head rides on a cushion of air and never touches the platter under normal operation.

Any head that touches the platter is classed as a catastrophic hardware failure/defect.

Now run along and do some basic Google searches to educate yourself and stop embarrassing yourself further.

LOL. Yep.

 

[ShintaiDK]

LOL.

If you think a head touching the surface of a HDD platter is part of normal operation and wears out the magnetic layer eventually, you really don't understand even the most basic concepts of a HDD.

Here's a hint: a HDD is not a vinyl record. The head rides on a cushion of air and never touches the platter under normal operation.

Any head that touches the platter is classed as a catastrophic hardware failure/defect.

Now run along and do some basic Google searches to educate yourself and stop embarrassing yourself further.

I never said the head itself causes the issue. Any dust particle or particles from already lose material will do it as well.

Just because you made a fool out of yourself saying that SSDs would lose data in a matter of days, doesn't mean you have to get aggressive.

 

[BFG10K]

I never said the head itself causes the issue.

LOL, wut? If the head's not involved, how does something grind the platter?

Any dust particle or particles from already lose material will do it as well.

Back pedal fail. External particles won't get in because of the filter. Furthermore, Helium drives are hermetically sealed so they're airtight.

Particles falling from the internal assembly are also classed as hardware failure/defects. That's not normal operation for an HDD, unlike SSD write cycles which is inherent to all NAND.

So no, HDDs don't "wear out" like SSDs do.

Just because you made a fool out of yourself saying that SSDs would lose data in a matter of days, doesn't mean you have to get aggressive.

Don't be mad because all three consoles ended up using an AMD APU.

 

[ShintaiDK]

Wow, its obviously too personal for you.

So this is my last post on the subject. The chemical substrate itself also degrade over time. And if we are to believe you, that the hard drive casings are 100% sealed and clean. Then bad sections should essentially never happen. But they do, over and over.

 

[BFG10K]

Wow, its obviously too personal for you.

I'm not the one calling people fools.

So this is my last post on the subject. The chemical substrate itself also degrade over time.

Everything degrades over time, and everything develops potential failure points with age and/or from manufacturing imperfections.

But hardware failure is completely different to NAND write cycles. The former is not supposed to happen while the latter is an accepted tradeoff inherent to the design.

And if we are to believe you, that the hard drive casings are 100% sealed and clean. Then bad sections should essentially never happen. But they do, over and over.

Likewise we were told SSDs couldn't fail because they have no moving parts. But they do, over and over. But you don't see anyone running around claiming bricked SSDs are "worn out".

 

[Topweasel]

LOL, wut? If the head's not involved, how does something grind the platter?


Back pedal fail. External particles won't get in because of the filter. Furthermore, Helium drives are hermetically sealed so they're airtight.

Particles falling from the internal assembly are also classed as hardware failure/defects. That's not normal operation for an HDD, unlike SSD write cycles which is inherent to all NAND.

So no, HDDs don't "wear out" like SSDs do.


Don't be mad because all three consoles ended up using an AMD APU.

So what would be the total data transfered a 4TB 40nm V-Nand MLC drive before you start to worry about writes?

Even on the best HDDs you do eventually start generating bad sectors. You do have to deal with mechanical and logical failures. 3d nand specially when used like it is on the 850pro erases "range anxiety". The biggest reason though that we won't see HDDs go away is cold storage. But depending on the development of 3d nand I could see it pushing the market for general storage in the next couple years.

 

[Puffnstuff]

I'll take an ssd over a spinner any day of the week.

 

[coercitiv]

So what would be the total data transfered a 4TB 40nm V-Nand MLC drive before you start to worry about writes?

Even on the best HDDs you do eventually start generating bad sectors. You do have to deal with mechanical and logical failures.

What BFG10K may have suggested, and he's welcome to correct me if otherwise, is that when placed in enterprise environments SSDs tend to reach comparable annual failure rates as HDDs.

The only two studies I found related to the subject are the ones related to Faceebok SSD usage (2015) and Google HDD usage (2007). Other data points can be observed from Backblaze, Intel or OCZ. If anyone alse has other/better data to share, especially for consumer HDD failure rates, please do, this is a very interesting subject in my view!

As far as I could tell, when placed under heavy server duty roles, SSDs may tend towards similar failure rates as HDDs, simply because they eventually reach high wear states. Keep in mind though, HDD also exhibit high failure rates under heavy loads, albeit due to mechanical failures (early defects in the first 3-6 months or old age after 5+ years - see Google study).

However, when looking at consumer loads, the data landscape may look completely different:

• annual failure rate for HDDs can be considered as higher than 1%, with 2% being a good value
• annual failure rate for SSDs can be considered as lower than 1%, with <0.5% being a good value

Here's how some data from Intel looks like:

 

[Topweasel]

Server usage will always be the great equalizer. That said my comment still stands. Take Samsung, their plan for 3d nand is to to stack layers till at least 2020 on the 40nm process with drives getting to 10+ TB. On the 14nm process on a 256gb drive they got to nearly a PB of writes before failure and that was on a TLC unit. A MLC drive at 40 times the size, at a process nearly 4 times larger is going to have a write capacity so epically large that it would mitigate server load based wear failures almost completely.

On top of that you have Intel/Micron's super Nand and SLC (if using 3d tech as well) that would be better SDD tech for high write server tasks. Both of which allow for almost exponential increase in write wearing numbers. So much so that Intel is targeting the the Xnand (can't remember the name might be wrong) as basically a secondary memory solution with near infinite writes.

I am not a SDD die hard. I once got told off here for supporting the idea that a guy might not want to trade his striped Raptors with an SSD about 2 years ago here. But we are now entering an era where HDD's are going to struggle to keep up with capacity of SSDs and new tech basically eliminates the wear anxiety that might hold people back.

 

[WhoBeDaPlaya]

Likewise we were told SSDs couldn't fail because they have no moving parts. But they do, over and over. But you don't see anyone running around claiming bricked SSDs are "worn out".

At least spinning rust drives fail gracefully / gradually (most of the time).
One "oopsie daisy" in an SSD's translation tables and you're toast.

 

[Denithor]

Okay, let's get some real data into this discussion.

http://royal.pingdom.com/2011/12/19...-drive-charting-hdd-and-ssd-prices-over-time/

Dated, but shows the historical trend nicely. Circa 2012 HDD was $0.05/GB while SSD was $1/GB.

http://www.storagenewsletter.com/ru...when-will-ssd-have-same-price-as-hdd-priceg2/

This one, from mid-2013, shows pricing falling into the $0.50 - 0.75/GB range for 250/500GB drives.

Now, in late 2015, I personally was able to buy a BX100 250GB drive for $64, about $0.25/GB. And that wasn't the best deal during the sales week, just one I happened to see.

Currently the cheapest HDD in this size is around $15 (quick price check on Newegg). The cheapest non-refurbished 1TB drives are about $40. So directly comparing same size, SSD is a bit over 4x more expensive than HDD. Considering instead the "standard" size HDD used these days, SSD is more than 6x higher cost. These are for desktop drives. In laptop space, it's even closer: 250GB laptop drive is $35 ($0.14/GB), 1TB is $50 ($0.05/GB). So for same size, SSD is less than 2x the price while for a 1TB it's 5x more expensive.

The cost/GB for SSD seems to be halved about every 12-18 months. HDD prices have been completely stable for years and years. I'm thinking we should see price parity by around 2020 assuming no disruptive factors (sudden increase in HDD platter density, 3D XPoint, etc).

 

[Denithor]

Oh, and regarding the "wearing out" of HDDs: They don't wear out from the heads grinding down the platters like a vinyl record. Instead, air resistance inside the drive slowly wears out the bearings, causing minute vibrations in the platters, bringing them closer and closer to the read heads as the wobble becomes more pronounced. Eventually, either the bearings completely fail or the heads physically touch the platter (or even just cannot read properly due to being a different height from the platter) and the drive fails.

This is the entire reason for use of helium in drives today. Platters spinning through helium have a lower coefficient of friction than those spinning through air. So there is less resistance and the bearings last longer.

http://www.extremetech.com/computin...b-hard-drive-helium-filled-shingled-recording

 

[moonbogg]

They should fill high performance HDD's with compressed hydrogen and name them "Fire HD Pro" series.

 

[Puffnstuff]

At least spinning rust drives fail gracefully / gradually (most of the time).
One "oopsie daisy" in an SSD's translation tables and you're toast.

I used to subscribe to that theory until several months ago when my veloraptor suddenly decided that it no longer wanted to be a valued member of my pc and crashed the head onto the platter without warning. The abrupt end of life event happened for me just like all of the failed sandforce 12xcx ssd's I've had die in the past only it made a horrible sound as it passed on to the hard drive afterlife (aka my garbage can).

 

[hectorsm]

I'm thinking we should see price parity by around 2020 assuming no disruptive factors (sudden increase in HDD platter density, 3D XPoint, etc).

You can count on the HDD platter density to continue to increase in the next few years. I don't see SSD replacing HDD by 2020. It's too early to say for sure but most likely be closer to 10 years from today.

 

[StinkyPinky]

I wish companies like Dell would move to get rid of HDD's from their client pc's. I've been arguing with the guy who does procurement at work for years now about not buying optiplex pc's with shitty hard drives inside. Why get a 1tb hdd when all the data is stored on file servers, when you can get a 120g solid state that actually allows for a better user experience and efficiently. The guy is a luddite though so i expect we will be getting HDD's for years to come, and that's why Dell still offers them.

 

[hectorsm]

I wish companies like Dell would move to get rid of HDD's from their client pc's. I've been arguing with the guy who does procurement at work for years now about not buying optiplex pc's with shitty hard drives inside. Why get a 1tb hdd when all the data is stored on file servers, when you can get a 120g solid state that actually allows for a better user experience and efficiently. The guy is a luddite though so i expect we will be getting HDD's for years to come, and that's why Dell still offers them.

120GB is cutting it close even for many business. I work for a state agency and I'm seeing some PCs run out of space with 160GB drive. Most of our data is on the network so most of the space is been eating up by large apps. Our division has been upgrading these HDD to 500GB.

For consumers the situation is a lot worse. A 120GB will not cut it for most people now days. Larger OSes, larger applications, larger jpeg files, more HD video streaming, etc is becoming more and more common which will require drives in the Terabyte size. The majority of people that don't need large storage are using tablets or phone anyway since they are sufficient for their needs.