Fusion-io Still Sees a Future For Hard Disk Drives

[Guest]

Fusion-io, one of the most visible makers of enterprise flash drives, does not believe that solid state disk drives will be able to replace the traditional hard drive anytime soon.

Fusion-io Still Sees a Future For Hard Disk Drives : Read more

 

[drwho1]

Fusion-io Still Sees a Future For Hard Disk Drives....

Duh!

 

[freggo]

Stating the obvious... must be a slow news day

 

[mavroxur]

As long as hard drives are more affordable than SSDs for bulk storage, of course they'll still have a future.


In other news, water is wet.

 

[victorious 3930k]

[citation][nom]drwho1[/nom]Fusion-io Still Sees a Future For Hard Disk Drives....Duh![/citation]
Fusion-io is an SSD company...

In other news, I wasn't aware Woz worked there.

 

[ricdiculus]

Personally I store all of my data on Water. Its the latest thing...

 

[razor512]

SSD's may never catch up to hard drives in storage because storage is constantly being improved for spindle based drives.

By the time we get 2-3TB standard SSD's, we will see probably 12+ TB HDD's

 

[alidan]

[citation][nom]Razor512[/nom]SSD's may never catch up to hard drives in storage because storage is constantly being improved for spindle based drives.By the time we get 2-3TB standard SSD's, we will see probably 12+ TB HDD's[/citation]

for both ssd and hdd, there is a limit to how much you can store... the difference is that some day ssd will be able to be 3d, while hdds will still only be able to be surface only. if the process is cheap enough to go 3d, you could see ssds far out doing hdds in terms of storage.

 

[luciferano]

[citation][nom]alidan[/nom]for both ssd and hdd, there is a limit to how much you can store... the difference is that some day ssd will be able to be 3d, while hdds will still only be able to be surface only. if the process is cheap enough to go 3d, you could see ssds far out doing hdds in terms of storage.[/citation]

Maybe, but we already have the tech to make like 30TB drives if we count up several advances that have yet to be commercially implemented yet, so SSDs have a long way to go in catching up in capacity per price. I'm not aware of any SSDs that breach 1GB per 50 cents except maybe with occasional promotions, but hard drives can have a small fraction of that cost per GB.

 

[alidan]

[citation][nom]luciferano[/nom]Maybe, but we already have the tech to make like 30TB drives if we count up several advances that have yet to be commercially implemented yet, so SSDs have a long way to go in catching up in capacity per price. I'm not aware of any SSDs that breach 1GB per 50 cents except maybe with occasional promotions, but hard drives can have a small fraction of that cost per GB.[/citation]

im thinking of it this way, there is only so far you can scale a hdd down, well... with a hdd, its just on or off right now right? its not different levels of on and off?

i completely forgot about that possibility... that in the future we are able to accurately set it to levels beyond 1 or 0 and have it accurately tell the difference, hell just bringing 8 levels to it would add about 3 times the space... ok... i didnt think that through all the way... but im not sure if i would trust a hdd with more than on or off over a ssd...

 

[luciferano]

[citation][nom]alidan[/nom]im thinking of it this way, there is only so far you can scale a hdd down, well... with a hdd, its just on or off right now right? its not different levels of on and off? i completely forgot about that possibility... that in the future we are able to accurately set it to levels beyond 1 or 0 and have it accurately tell the difference, hell just bringing 8 levels to it would add about 3 times the space... ok... i didnt think that through all the way... but im not sure if i would trust a hdd with more than on or off over a ssd...[/citation]

Laser-guided magnetic R/W heads have been proven to be much more accurate than non-guided R/W heads and might be able to be used for what you talk about here.

However, I was actually talking about stuff such as these:
http://www.tomshardware.co.uk/table-salt-sodium-chloride-18tb-hdd-10-nm,news-36755.html
http://www.tomshardware.com/news/Helium-Filled-Hard_Disk-hdd-western-digital,17573.html

 

[Cespenar]

SSD's have their place and use, but when millions of dollars are riding on electronic storage 24/7 at 365 days every year, the platters are going to stay for decades.
SSD's have to get their long term quality up to scratch before they can replace platters.
The best security is non volatile memory and SSD's are totally volatile. It is their nature.
I may eventually get an SSD for games when they become better and cheaper, but for now I wont use them.

 

[alidan]

[citation][nom]luciferano[/nom]Laser-guided magnetic R/W heads have been proven to be much more accurate than non-guided R/W heads and might be able to be used for what you talk about here.However, I was actually talking about stuff such as these:http://www.tomshardware.co.uk/tabl [...] 36755.htmlhttp://www.tomshardware.com/news/H [...] 17573.html[/citation]
the helium one, that doesn't increase the density, but increases the amount of platters, and with ssds, they are currently close to if not already at the same density as hdds, per size, and both have an upper limit as to what they can get to, ssd being about 6nm, and i'm assuming that hdds can get close to that too.

 

[luciferano]

[citation][nom]alidan[/nom]the helium one, that doesn't increase the density, but increases the amount of platters, and with ssds, they are currently close to if not already at the same density as hdds, per size, and both have an upper limit as to what they can get to, ssd being about 6nm, and i'm assuming that hdds can get close to that too.[/citation]

The helium method increases capacity by supporting more platters. It doesn't matter if there is a density per platter improvement in that because having two more platters improves theoretical capacity by 40%. The other one does increase density and it does so greatly, but I didn't mention density anyway, so although this is relevant, it doesn't prove nor disprove me.

SSDs can get as high capacity as they want to, but they can't do so as cheaply as a hard drive can. With just the density improvement link's methods we could more than quadruple maximum capacity without increasing manufacturing costs by much and the helium method probably isn't too expensive either and has a good capacity bonus too. Also, together, they'd be over 25TB, greater than the highest capacity SSD that I've ever heard of, OCZ's incredibe PCIe 16TB SSD.

SSDs are limited to where their cells can function properly. Hard drives don't need electronic cells, they just need little magnetic shards. They can get smaller than a NAND flash cell can ever hope to get with silicon.

 

[alyoshka]

Isn't this the Apple guy..... no wonder he comes up with such stuff..... and especially the price of the Fusion -Io..... that's something which never made sense to me earlier.... but now... with an Apple in the equation... I get it all.

 

[alidan]

[citation][nom]luciferano[/nom]The helium method increases capacity by supporting more platters. It doesn't matter if there is a density per platter improvement in that because having two more platters improves theoretical capacity by 40%. The other one does increase density and it does so greatly, but I didn't mention density anyway, so although this is relevant, it doesn't prove nor disprove me.SSDs can get as high capacity as they want to, but they can't do so as cheaply as a hard drive can. With just the density improvement link's methods we could more than quadruple maximum capacity without increasing manufacturing costs by much and the helium method probably isn't too expensive either and has a good capacity bonus too. Also, together, they'd be over 25TB, greater than the highest capacity SSD that I've ever heard of, OCZ's incredibe PCIe 16TB SSD.SSDs are limited to where their cells can function properly. Hard drives don't need electronic cells, they just need little magnetic shards. They can get smaller than a NAND flash cell can ever hope to get with silicon.[/citation]

actually, unless current hdds get a major kick in the right direction, ssds could come down to about the same cost/gb ratio soon

i mean within 5 years. and lets not talk about adding infinite platters, lets go with 2-4 being the norm still.

i mean if 1tb costs 100$ and a 1tb ssd costs 130$ which one are you going to get? i dont care about your budget at that point, you are going ssd even if its for storage.

actually, i just did the math, at current process, if ssds went to the 7nm process today, they would easily beat out hdds on cost per gb. the next major jump in hdd will hear assisted will be when? and im having a hard time thinking that will be cheap, or affordable when it comes out, unless they get their crap together and bring the price down to pre flood, they are really screwing themselves, because i will play at most a 50% premium on an ssd over a hdd at the same size just for the no moving parts portion of it, meaning that in meaning that in about 3 years for me ssd will be more viable an option for storage than hdd.

 

[luciferano]

[citation][nom]alidan[/nom]actually, unless current hdds get a major kick in the right direction, ssds could come down to about the same cost/gb ratio sooni mean within 5 years. and lets not talk about adding infinite platters, lets go with 2-4 being the norm still.i mean if 1tb costs 100$ and a 1tb ssd costs 130$ which one are you going to get? i dont care about your budget at that point, you are going ssd even if its for storage. actually, i just did the math, at current process, if ssds went to the 7nm process today, they would easily beat out hdds on cost per gb. the next major jump in hdd will hear assisted will be when? and im having a hard time thinking that will be cheap, or affordable when it comes out, unless they get their crap together and bring the price down to pre flood, they are really screwing themselves, because i will play at most a 50% premium on an ssd over a hdd at the same size just for the no moving parts portion of it, meaning that in meaning that in about 3 years for me ssd will be more viable an option for storage than hdd.[/citation]

HDDs have the technology to increase in size several times over right now, but SSDs, regardless of how cheaply they can be made, would not just drop in price like that. The companies would undoubtedly just milk the profits and continue dropping prices slowly. HDD companies don't need some sort of kick, they just need to choose if they want to utilize tech that allows for higher capacity and if SSDs get too close in price per capacity, then chances are HDD companies would start utilizing the improved tech to get higher capacity for about the same BOM. Beyond that, I didn't say anything whatsoever about infinite platters.

Until NAND flash is replaced, unless the HDD companies really screw up, SSDs shouldn't get anywhere near a mere 30-50% more expensive price premium. Moving to smaller processes constantly hurts reliability/endurance of the flash while also increasing latencies (although I suspect that the latter is caused by not improving designs well enough to optimize for die shrinks, but I digress), so simply shrinking the die size is not enough, especially when we start talking about something like NAND flash built on a 7nm process.

 

[alidan]

[citation][nom]luciferano[/nom]HDDs have the technology to increase in size several times over right now, but SSDs, regardless of how cheaply they can be made, would not just drop in price like that. The companies would undoubtedly just milk the profits and continue dropping prices slowly. HDD companies don't need some sort of kick, they just need to choose if they want to utilize tech that allows for higher capacity and if SSDs get too close in price per capacity, then chances are HDD companies would start utilizing the improved tech to get higher capacity for about the same BOM. Beyond that, I didn't say anything whatsoever about infinite platters.Until NAND flash is replaced, unless the HDD companies really screw up, SSDs shouldn't get anywhere near a mere 30-50% more expensive price premium. Moving to smaller processes constantly hurts reliability/endurance of the flash while also increasing latencies (although I suspect that the latter is caused by not improving designs well enough to optimize for die shrinks, but I digress), so simply shrinking the die size is not enough, especially when we start talking about something like NAND flash built on a 7nm process.[/citation]

i can not for the life of me find out how big the chips are in ssds, the silicon not the black finished product, but i am able to read the process size, and figure out pricing from that.

if you want to talk about reliability, in the amount of writes, a 256-512 gb drive would need to over written with 25tb or 51tb just to rewrite the whole drive 100 times, and current drives have a ware level of 2500-5000 writes, my point being is that even if the drives had crap reliability in that regard it would still take forever to get that carp to the point the drive is useless.

 

[luciferano]

[citation][nom]alidan[/nom]i can not for the life of me find out how big the chips are in ssds, the silicon not the black finished product, but i am able to read the process size, and figure out pricing from that. if you want to talk about reliability, in the amount of writes, a 256-512 gb drive would need to over written with 25tb or 51tb just to rewrite the whole drive 100 times, and current drives have a ware level of 2500-5000 writes, my point being is that even if the drives had crap reliability in that regard it would still take forever to get that carp to the point the drive is useless.[/citation]

That's with current flash. Shrinking the process down to 7nm would greatly hinder write endurance. Heck, just about two or three years ago, MLC NAND flash could have orders of magnitude higher endurance, so it would seem that endurance goes down more than density goes up when you shrink the process.

 

[Cespenar]

I have read about reliability problems with SSD's but don't have first hand experience.
When it comes to HDD's all my ten or so HDD's are still running bar one that was out of alignment and lasted 24 hours.
All these HDD's are over 6 years old and some are over 10 years old. Except for the one I bought this year.

 

[luciferano]

[citation][nom]Cespenar[/nom]I have read about reliability problems with SSD's but don't have first hand experience.When it comes to HDD's all my ten or so HDD's are still running bar one that was out of alignment and lasted 24 hours.All these HDD's are over 6 years old and some are over 10 years old. Except for the one I bought this year.[/citation]

So long as unless you buy Intel, you don't buy a SandForce-based SSD nor an OCZ Octane and buy a modern, SATA 6Gb/s SSD using a different controller, you're probably not going to run into any reliability issues. Really, anything that either has a Samsung or a Marvell controller or is made by Intel generally doesn't fail. SSDs with Marvell controllers would include Crucial M4, Plextor MS5, OCZ Vertex 4, Corsair Performance Pro, and several other SSD families. There are other, lesser known SSD controller companies that are also good and such, but this list has a lot of the best performers at the best prices without sacrificing reliability. There are other great options, but this is what I can think of right now.

 

[alidan]

[citation][nom]luciferano[/nom]That's with current flash. Shrinking the process down to 7nm would greatly hinder write endurance. Heck, just about two or three years ago, MLC NAND flash could have orders of magnitude higher endurance, so it would seem that endurance goes down more than density goes up when you shrink the process.[/citation]

yea... current ssd flash gets 1000-5000 writes, i think the average is 2500, and with one type up to 1000000

thats why i did the math with 100 writes, a level i seriously doubt they will ever sink to, and am looking at 256 and 512gb drives as they will be at around the 50-100$ point by the time of 7nm, if not even less. for the average person, they will still be something you never have die because of the amount of writes unless you are moving whole harddrives every day worth of data.

 

[luciferano]

[citation][nom]alidan[/nom]yea... current ssd flash gets 1000-5000 writes, i think the average is 2500, and with one type up to 1000000thats why i did the math with 100 writes, a level i seriously doubt they will ever sink to, and am looking at 256 and 512gb drives as they will be at around the 50-100$ point by the time of 7nm, if not even less. for the average person, they will still be something you never have die because of the amount of writes unless you are moving whole harddrives every day worth of data.[/citation]

Going from 34nm to 25nm is what this signifies. 25nm is 7 times less enduring, but probably only doubled density. That's not a good ratio of density improvement to endurance drop. I'd honestly be worried about 7nm flash if it doesn't bring some sort of other improvements, such as using SOHOS or SHiNOS instead of NAND flash.

 

[Cespenar]

Dont need to write a whole HDD to a chip to ruin it.
Only when you can disable the little bit that is broken will a chip be as reliable as a platter.
We don't need chip style yet. But it is a new tech and geeks love new tech.