HAMR or Heat-Assisted Magnetic Recording. 100 TB HDD by 2025.

snowhiker

Wannabe Storage Freak
Joined
Jul 5, 2007
Messages
1,424
#1
"The HAMR process consists, in short, in heating the magnetic media platters before writing the data to about 450°C using a laser with 810mm wavelength and 20mW power."

and

"Apparently, the HAMR technology is only a first step of development for HDDs since ASTC (Advanced Storage Technology Consortium) that consists of a group of multiple HDD manufacturers plans a further HDD development beyond the current 10TB+ storage sizes spanning until 2025 and beyond, promising storage sizes reaching 100TB+ in ten years from now."

Article here.

10 years is a long ways away. Will HDD be around 10 years from now or will SSDs have completely taken over the market?
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#3
10 years is a long ways away. Will HDD be around 10 years from now or will SSDs have completely taken over the market?
I think the hand writing is on the wall already. Samsung has shown that SSDs trump HDDs right now in terms of size with its 16TB drive. Had they bothered to make something similar in a 3.5" form factor, I've little doubt they could have achieved 50TB. SSD trumps HDD in speed. The final hurdle is cost. Right now it seems we're hovering around $350 per TB. That's perhaps a factor of 10 higher than HDD. However, 3D NAND is supposed to increase in density by something like a factor of 8 in the next 3 or 4 years. Assuming a chip costs about the same to make as now, or only marginally more, that puts SSDs within a factor of 2 in terms of cost. Even if HDDs get marginally denser, SSDs will only cost 3 to 4 times more. If you want to go out to 2025, sure, by then we'll reach the limits of 3D NAND scaling but we might be a factor of perhaps 50 denser than now (and also a factor of 50 less costly in terms of price per TB). That brings us to, say, $7 per TB. Most of what I read suggests any future increases in HDD density will come at the price of more expensive manufacturing costs. Therefore, price per TB may not drop in proportion to drive size. Their hypothetical 100TB HAMR drive might well cost over $500. Going by my numbers, a 100TB SSD might cost $700. At that point, it's hardly worth going with HDD. And then you have other technologies in the wings like XPoint. By then it may scale enough in terms of cost and density so as to be better than 3D NAND.

Granted, this is all speculation but honestly I wouldn't be buying any stock of HDD companies now. SSDs are well on their way to replacing HDDs in consumer applications. Arguably, once you get 1 or 2 TB drives under about $100 you're there. If Samsung is correct that'll happen in 3 years. Most people don't need much more storage than that, and $100 or so is about what you'll pay for an HDD anyway. Enterprise may take a bit longer, but the speed of SSDs could make paying a premium worthwhile. We saw this already with CRTs. Once LCDs only cost slightly more than CRTs you couldn't give away CRTs any more. We're seeing it with LEDs displacing all other light sources. We'll see it with EVs replacing ICE vehicles in the next decade. And we'll undoubtedly see the end of the traditional HDD. We're probably living in a time with more disruptive technologies coming to fruition at the same time than at any other time in history.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#4
I thought about 7 years ago you predicted hard drives would be dying out in 2012 or something, and they are still hanging in there. :) I hear that in some countries tape drives as still used for enterprise backups.
The dead end of SSD process shrink and requirements for stacked memory cells has limited some of those plans for super cheap SSDs.
I suppose SSDs will eventually take over completely, maybe for your grandchildren. I'll most likely be dead by then.
Almost all consumer vehicles may indeed include hybrid power in the 2020s, but only a limited number will be EVs for reasons already discussed ad nauseum.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#5
I thought about 7 years ago you predicted hard drives would be dying out in 2012 or something, and they are still hanging in there. :) I hear that in some countries tape drives as still used for enterprise backups.
For consumer stuff they're pretty close to dying out. I'm not sure exactly what I predicted 7 years ago. It may have been HDDs dying by 2015 or 2017.

I've little doubt in some third world countries they're still using 5" floppies as backup. But my point here is especially for consumer applications the HDD's days are numbered. I can't see any situation where I would need to buy another HDD. By the time my 2TB gets short of space, I'll probably be able to upgrade to a 10TB SSD for a reasonable price.

The dead end of SSD process shrink and requirements for stacked memory cells has limited some of those plans for super cheap SSDs.
I suppose SSDs will eventually take over completely, maybe for your grandchildren. I'll most likely be dead by then.
Process shrink limited us for a while until they started making 3D NAND. Now Moore's Law is alive and well again, at least for another 5 years. 5 years is an eternity in this business. It's easily possible XPoint or MRAM or some other technology will be ready to continue density increases past that point. I'd be surprised if HDDs were made for anything but enterprise applications in another decade. I'll be surprised if they're made at all in two decades. The problem is they're a 2D technology. As you make the bits smaller reliability inherently decreases. It may well be that factor which kills them before anything else. I personally won't buy HDDs any more for exactly that reason. I'm not thrilled by the reliability of today's drives.

Almost all consumer vehicles may indeed include hybrid power in the 2020s, but only a limited number will be EVs for reasons already discussed ad nauseum.
Three things have prevented mass adoption of EVs-range, recharge time, and cost. The first two are kind of interchangeable. If you have enough range, then recharge time starts to matter less and less. Once you get to the maximum number of miles an average person might drive in a day, say 600 or 700, then overnight recharging is fine. We're about halfway there. Or you could have today's range, but with rapid recharge times. Either way works. I think we're down to 30 minutes for quick recharge. Get that down to 5 and range isn't an issue. I'd say we're 5 years away from solving both these issues. It's also worth noting for the vast majority who almost never drive further than 100 miles in one shot EVs are already good enough. Chevy noted that with the Volt. The 40 mile range covers 90+% of trips.

Cost will take care of itself if you made EVs in the same quantities as ICEs. In fact, they would likely cost less because the drivetrain is much less complex. It's a chicken or egg thing. The cost won't drop until we make enough but people won't buy them until the cost drops. The solution is captive markets, namely large cities passing ZEV requirements, particularly for fleet vehicles. That looks almost certain to happen in the next decade. I'm trying to get NYC to do exactly that right now. Once automakers know they have a captive market for many EVs they'll make them in the quantities needed to match ICE prices.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#6
One more thing worth noting since I mentioned reliability. Eventually, both HDDs and SSDs will reach a density plateau. Right now the fact both are increasing in density means lifespan isn't super critical. If an HDD lasts 3 or 4 years that's good enough since by then you'll likely want to swap it out for a larger model. What happens when there is no larger model a few years down the road? That's when lifespan matters. Studies of HDDs by companies like BackBlaze show mechanical wear issues resulting in increased failure rates starting at about 4 years. SSDs have their own wearout mechanisms but for most uses it will take many decades to reach that point. As far as I know, other than the limited number of writes SSDs should show similar reliability to any other electronic device. In other words, properly designed they'll function for many decades. Once sizes stop increasing regularly this will be a major point in their favor. A mechanical hard disk will wear out sooner or later. It's just the nature of the beast. I don't think we can design in more than 5 to 7 years of reliability without making them prohibitively expensive. SSDs on the other hand have the potential to last much, much longer. The situation is analogous to LEDs. They're taking over even when the installation cost is more simply by virtue of the fact that they outlast other light sources by a large margin.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#7
I have hard drives 7 years old that are fine, although they are rarely inuse. I thought SSDs were supposed to hold data for ten years.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#8
I have hard drives 7 years old that are fine, although they are rarely inuse. I thought SSDs were supposed to hold data for ten years.
I have HDDs over a decade old which work just fine. However, remember those have much lower density than today's HDDs. As bit density goes up reliability goes down. That applies to both HDDs and SSDs. That's why the move to 3D NAND was a smart one. We actually moved back to larger cells, hence more reliable data storage.

I'm not sure how long SSDs will hold data with the power off. I think the very first SSDs touted ten year data retention but those used much larger cells. I know the 840 series was having some issues of slow reading because cells which weren't written to in a while were starting to lose their charge. Supposedly the latest firmware upgrade fixes this by periodically refreshing the cells. 3D NAND with much larger cells shouldn't face this issue. The bottom line is I probably wouldn't use an SSD for archiving anything important unless it was in a machine which powered up at least a few times a year. For that matter I wouldn't use one of today's high density HDDs for archiving either. The magnetic bits would likely degrade after a few years. I would probably stick to tried and true DVD-RW with archival media. However, I also keep anything important to me in multiple places (on a few hard disks, on USB drives, sometimes on optical media). Highly unlikely all these things will fail at once.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#9
Chevy noted that with the Volt. The 40 mile range covers 90+% of trips.

Cost will take care of itself if you made EVs in the same quantities as ICEs. In fact, they would likely cost less because the drivetrain is much less complex. It's a chicken or egg thing. The cost won't drop until we make enough but people won't buy them until the cost drops. The solution is captive markets, namely large cities passing ZEV requirements, particularly for fleet vehicles. That looks almost certain to happen in the next decade. I'm trying to get NYC to do exactly that right now. Once automakers know they have a captive market for many EVs they'll make them in the quantities needed to match ICE prices.
Large cities are rather a dying breed. NYC is less than 3% of the population. Even if you add Chicago or SF, etc. I don't see that as representative of the country as a whole.

40 miles per charge is insufficient. 90% of total round trips may be less than 40 miles, but the number of people that drive no more than 20 miles each way to any destination in a year must be much smaller, perhaps under 50%. The range really needs to be >100 miles in winter or summer and there need to be fast charging stations in every gas station or other accessible locations to make it practical. Millions of people don't have access to direct charging, for example those living in some apartments/condos or parking on the street, living in areas where upgrading the infrastructure is impractical, etc.

In the spring we drove 2700 miles in eight days all over Utah, Arizona, Colorado and maybe New Mexico. I passed a number of signs with warnings of no fuel for the next 40, 50, 80miles, etc. Maybe there will be a full-sized, electric 4WD SUV that does that trip in ten years, but maybe not. I just don't see it being economical to set up charging stations upeverywhere.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#10
Large cities are rather a dying breed. NYC is less than 3% of the population. Even if you add Chicago or SF, etc. I don't see that as representative of the country as a whole.

40 miles per charge is insufficient. 90% of total round trips may be less than 40 miles, but the number of people that drive no more than 20 miles each way to any destination in a year must be much smaller, perhaps under 50%. The range really needs to be >100 miles in winter or summer and there need to be fast charging stations in every gas station or other accessible locations to make it practical. Millions of people don't have access to direct charging, for example those living in some apartments/condos or parking on the street, living in areas where upgrading the infrastructure is impractical, etc.

In the spring we drove 2700 miles in eight days all over Utah, Arizona, Colorado and maybe New Mexico. I passed a number of signs with warnings of no fuel for the next 40, 50, 80miles, etc. Maybe there will be a full-sized, electric 4WD SUV that does that trip in ten years, but maybe not. I just don't see it being economical to set up charging stations upeverywhere.
Gas stations have to have electric, don't they? That makes setting up charging stations just about trivial if there's enough demand for it (and the station owner can make money doing so). When gas cars first came out there weren't gas stations everywhere. And yet somehow people managed. Here's a map which most likely proves you wrong. Zoom in just about anywhere and you see charging stations all over the place. I was honestly surprised myself how many there are. In truth, I had expected maybe a few dozen in major metro areas and just about nothing in between.

Range already is well past 100 miles for pure electrics. If they had forgone the engine and put in a bigger battery, the Volt could have had at least 150 mile range.

80.7% of the US population resides in urban areas according to the 2010 census. Don't know where you get your figures from. NY metro area is over 6% of the US population. Urban development hardly ends at NYC's borders. Right across the Hudson River is starting to resemble Manhattan, for example. All the way between Washington and Boston you have large cities. Something like 1/3 of the US population is on the east coast. It's perfectly feasible if need be to set up charging stations so one can drive all the way from Boston to Miami if the mood suits them. In fact, that may already be the case.

Not a whole lot of people will drive 2700 miles on vacation these days. Most will take a plane and if need be rent a car when they get there. Yeah, gas cars might be needed a while longer in the truly rural parts of the US, but that that accounts for less than 20% of the population. When Tesla or someone else comes out with a vehicle with 750 mile range they won't even be needed there. Given all the heavy development of battery tech for other uses besides cars, it's only a matter of time before batteries increase in capacity substantially.
 
Last edited:

Handruin

Administrator
Joined
Jan 13, 2002
Messages
12,784
Location
USA
#11
Unless all your data will be planned to be used as a sustained stream you're going to have a bad day getting any kind of random data out of a 100TB mechanical disk spinning at 7200 RPMs.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#12
Unless all your data will be planned to be used as a sustained stream you're going to have a bad day getting any kind of random data out of a 100TB mechanical disk spinning at 7200 RPMs.
Sure, but it would be nice to have some of those 100TB drives RAIDed for the bulk storage and backups, like the RAW files after processing is done. :)
Even the 15.3 TB SSDs in RAID will crap out occasionally. How are they backed up?

For now (as a working stiff) I'll keep using a few SSDs for the hot data and 7200 drives for other data.
 

Handruin

Administrator
Joined
Jan 13, 2002
Messages
12,784
Location
USA
#13
Sure, but it would be nice to have some of those 100TB drives RAIDed for the bulk storage and backups, like the RAW files after processing is done. :)
Even the 15.3 TB SSDs in RAID will crap out occasionally. How are they backed up?

For now (as a working stiff) I'll keep using a few SSDs for the hot data and 7200 drives for other data.
Sure but at what data rate will you as a consumer/user be satisfied with getting your data onto one of these drives? Many HDDs today seem to be in the 125 - 150MB/s range of sustained transfer performance. Some may be higher but no where near what an SSD at 16TB is capable of. At 100TB it would take you close to 8 days or more of nonstop read/writing to get your data in and out of that drive assuming 150MB/sec. I'd hope the performance number would go up quite a bit given the GB per sq. inch but realistically what will it be with all the CRC and error-checking that will need to be done?

If you're going to put a 100TB drive into RAID, you'd want some 4-5 parity drives to ensure a non-recoverable read error doesn't occur during a rebuild. I don't know what they'll be rated at so I'm making a huge guess with those numbers.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#14
STR is more or less proportional to the square root of bit density. If we managed to make 100TB hard disks, probably the STR would be in the 500 to 750 MB/s range. Still, your point remains. At those speeds it would take 1.5 to over 2 days to get the data in or out of the drive. In fact, I think this just puts yet another nail in the coffin of HDDs. SSDs are capable of massive parallelism if designed for it. Getting data on or off a huge drive shouldn't be overly time consuming, perhaps an hour or two at most.

It takes about 4.5 hours to do a surface scan on my 2TB HDD. It takes 20 minutes to do the same on my EVO 840 500GB SSD. By extrapolation a 2TB disk would take about an hour, 20 minutes. Little doubt if PCs had standard interfaces which were faster a drive could be designed to do the same thing in 10 or 15 minutes.
 

snowhiker

Wannabe Storage Freak
Joined
Jul 5, 2007
Messages
1,424
#15
Well maybe in 10 years most urban dwellers will have bi-directional 1 to 10 Gb internet access. And if everybody had 2-4 of the 100 TB monsters you could do backups to your friends drive and vice versa. Eventually everybody will have a copy of everybody's data.

I think no matter how large storage devices become, there will be a significant number of people that will want/need 100-5000x that much storage.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,645
Location
Flushing, New York
#16
Not sure about that. Back when I started with PCs in the late 1990s I remember drive sizes being somewhat limiting. My first machine had a 1.2GB HDD. Eventually I was at the point where I would have started needing to make decisions about what to keep and what to get rid of. Then I bought an 8.4GB HDD and those worries went away. Of course, this was a DOS 5.0/Windows 3.1 system. When I went to Windows '98 on my next machine I started off with 18 GB. That was enough for a while. By the time things started getting cramped, I moved on to the next machine which I started with 40GB. Eventually I added 80GB and 100GB drives for extra storage. On to the system after that. I put in 200GB and that was enough for a long time. I bought a 2TB HDD to supplement that eventually. Finally, on the machine I'm using now I started off with a 240GB SSD and the aforementioned 2TB drive. I added a 500GB SSD, mostly so I could dual boot to my old XP system which was on the 200GB drive, but now with extra storage space. I'm getting a little tight with space on the 240GB SSD but I can move things to other drives if need be. The 2TB drive is only half full. Mostly that's backups from the other drives. The bottom line is for me anyway, probably for lots of others, storage capacity reached the point where a new drive will take many years to fill. I can certainly see myself with a 10TB SSD a few years down the road. If some killer app comes along needing lots of storage I may even be able to fill a 100TB SSD but I highly doubt it. There may be a minority who always need more storage no matter how large storage devices become but I tend to think that minority is shrinking. Sure, there is data and program bloat, but fortunately drive capacity has outpaced it by a huge margin. I tend to think the vast majority of people will be set for many years with a few TB. Quite a few can get by with a few hundred GB.

On another note, if Samsung is saying they'll be able to make a 128TB SSD in 3 or 4 years, I shudder to think what we'll have in 2025 when the supposed 100TB HAMR drives will exist, assuming they're even made. It's possible we'll have multipetabyte SSDs by then, although I can't see any scenario where I would fill up even one petabyte, never mind several.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#18
Well maybe in 10 years most urban dwellers will have bi-directional 1 to 10 Gb internet access. And if everybody had 2-4 of the 100 TB monsters you could do backups to your friends drive and vice versa. Eventually everybody will have a copy of everybody's data.

I think no matter how large storage devices become, there will be a significant number of people that will want/need 100-5000x that much storage.
I have very few friends and they use Apples or laptops, etc. with a portable backup drive or Time Machines. Most people don't have 10TB of drives much less 100TB. Other than some members here, I don't know anyone else that has over 100TB of personal drives like me.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#21
Video on the subject. Thermal cycling can't be good for longevity.
Presumably there has been extensive development and testing. It's not clear to me if that heating requires zone recording like the f'ing drives with the shingles.

I much preferred the dancing PMR video than this boron speaking to the LCD with an advert at the end.
 

Tea

Storage? I am Storage!
Joined
Jan 15, 2002
Messages
3,716
Location
27a No Fixed Address, Oz.
Website
www.redhill.net.au
#22
But what past of ahard drive weasrs out? Spindle bearings esentially last forever because there are no moving parts in contact. The key, I shoild tbink is bearings for tbe actuator arms. I can't think of an easy way to wear-proof those.
 
Joined
Feb 4, 2002
Messages
19,277
Location
Monterey, CA
#23
With tiny parts of the platter being randomly heated to 500C I would fear for the integrity of the platter itself. Motors, bearings, and arms are things that I feel are pretty solid tech at the moment.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#25
If they have a warranty they'll stand behind it's alright with me. Where did that 500C number come from? The laser is tiny.
The patters are glass/ceramic. Anyway, if the laser were strong enough to damage them, the magnetic layers would be toast.
 

Tea

Storage? I am Storage!
Joined
Jan 15, 2002
Messages
3,716
Location
27a No Fixed Address, Oz.
Website
www.redhill.net.au
#26
There are certainly fatigue effects associated with temperature cycling of any solid material. (At least any that I can think of.) Generally these become significant where the material is a load-bearing structure, such as a gas turbine blade or a bridge span. I can't see it being significant in this context.
 

Tea

Storage? I am Storage!
Joined
Jan 15, 2002
Messages
3,716
Location
27a No Fixed Address, Oz.
Website
www.redhill.net.au
#28
Not much of one, Dave. Not when you consider the soolid, contiguous nature of the disc. On something constructed like a bicycle wheel, there is a lot of centrifigual force on the spokes because they are so narrow, but a solid disc has so much metal/glass in it that the force is distributed away to almost nothing.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,693
Location
USA
#29
The Seagate HAm-R is finally in limited production. Perhaps the yields truly suck or something so they won't be commercial for a while yet.
Maybe that MAm-R will be out by 2020 also.
 
Top