SanDisk Announces 4TB SSD, Hopes for 8TB Next Year

CommanderFrank

CommanderFrank

Cat Can't Scratch It
Joined
May 9, 2000
Messages
75,399
This week SanDisk announced that it will be releasing its 4TB enterprise-class SSD. Included in the announcement was the aim of SanDisk to produce 6 and 8TB units by next year. SanDisk officials also expect that the capacity of their SSDs will double every year or so.
 
You'll want two, 8TB seems like a scary amount of data to entrust to a single device. Or am I just being paranoid?
 
As pointed out by Aesma in the previous thread, "SAS drives are rarely used alone, these will be bought in quantity when bought."
 
Out of curiosity, what is the typical data life expectancy on an SSD? After a flash cell is written to, how long will the data on it be readable?
 
These well never replace a spinning hdd completely. But I would like to think of a day that it can.
 
Kueller said:
Out of curiosity, what is the typical data life expectancy on an SSD? After a flash cell is written to, how long will the data on it be readable?
Click to expand...

Any decent SSD these days if properly maintained in an average user's system will far outlast a mechanical drive in almost every case. The SanDisks I've tested, I rack up 6-7GB a day as a heavy user, and their tested lifespans based on write count would have it lasting in the 3-digit year (100) range. By properly maintained I mean you should leave 25-50% of the space free and not damage it with excessive testing, no defragging, avoid anything that does unnecessary writes, etc.
 
To elaborate on your question (can't edit), I forgot to say that the reliability of data is proportionate to the number of writes done on those cells, but from what I've seen on AnandTech and other places, SSD's theoretically will hold data for ten years or so.
 
soulman901 said:
Bye bye Spin drives don't let the door hit you on the way out....
Click to expand...

Um, no.

1.These are SAS drives, not SATA drives.

2.You can expect to pay $6000-$8000+ a pop for them if the price of their 2tb SSD is any indication. That is on top of the cost of the SAS controller you would need to use them.
 
Someone did a test with Samsung Evo SSD and had written up to 700TB of data before it gave out. I'd cite a reference but I am on the phone at the moment typing this. Google SSD life expectancy and you will find it. I think SSD are at a good point now that the cost and reliability are worth it.
 
damicatz said:
Um, no.

1.These are SAS drives, not SATA drives.

2.You can expect to pay $6000-$8000+ a pop for them if the price of their 2tb SSD is any indication. That is on top of the cost of the SAS controller you would need to use them.
Click to expand...

You read too much into my post. The point was about the death of spin drives as bigger and better SSD come out. Regardless of the controller they are supported on. Yes I know these will be expensive at first but will drop in price as other manufacturers catch on.
 
Maxx said:
To elaborate on your question (can't edit), I forgot to say that the reliability of data is proportionate to the number of writes done on those cells, but from what I've seen on AnandTech and other places, SSD's theoretically will hold data for ten years or so.
Click to expand...

Does that mean that after 10 years, basic low-level data (partition table, recovery partitions, OS files) on SSDs that have had no reason to be modified since being written will start to become unreadable? Or does the firmware (particularly on first generation non-wear leveling drives) periodically re-flash old blocks to refresh them?
 
Methos said:
You'll want two, 8TB seems like a scary amount of data to entrust to a single device. Or am I just being paranoid?
Click to expand...

Yeah for any kind of mass storage I'd definitely want to do raid, even with SSDs.

In fact I'd want to swap out SSDs at random intervals to ensure they're not getting equal wear or they're bound to all fail at once. Could have a hot spare and then just force a rebuild once in a while.

A bunch of these in raid 10 would be pretty awesome. :D
 
Kueller said:
Does that mean that after 10 years, basic low-level data (partition table, recovery partitions, OS files) on SSDs that have had no reason to be modified since being written will start to become unreadable? Or does the firmware (particularly on first generation non-wear leveling drives) periodically re-flash old blocks to refresh them?
Click to expand...

SSD controllers will automatically rewrite static data, at least on enterprise drives. For consumer drives I really suggest people backup and fully rewrite or format once a year if they have concerns about it, it's not a significant hit on lifespan.
 
DPI said:
I'm not sure you're really understanding this announcement here..
Click to expand...

I understand it perfectly fine. Seems like you missed the point of my post. Larger SSD will hasten the death of spin drives. Spin Drives are hitting the limits that SSD do not have. That is why I think with what San Disk announced will spurn an SSD race benefiting business and conumers.
 
That SanDisk announced is typical progression, this isn't some revolution they started. The other manufacturers will catch on, but these drives won't lower in price, not until there is a consumer demand, which isn't very high for big SSDs right now.

The big push will be when a company releases these bulk SSDs to the consumer at an affordable price.
 
What I found interesting is that SSD has caught up with HDD in terms of capacity, and if SanDisk's prediction is true, HDD will never be able to keep up in terms of capacity while the cost of a SSD of the same capacity may catch up with HDD in the near future.
 
Maxx said:
Any decent SSD these days if properly maintained in an average user's system will far outlast a mechanical drive in almost every case. The SanDisks I've tested, I rack up 6-7GB a day as a heavy user, and their tested lifespans based on write count would have it lasting in the 3-digit year (100) range. By properly maintained I mean you should leave 25-50% of the space free and not damage it with excessive testing, no defragging, avoid anything that does unnecessary writes, etc.
Click to expand...

For individual use, I'd agree. Based on the write numbers from the Samsung EVO drive I installed in my home system a few months ago, it should easily last over 100 years.

However, these are SAS drives are meant for server usage which has a completely different usage pattern. I doubt they will last anywhere near the same.

SSD's have now surpassed mechanical drives in capacity, but until the prices get a lot closer, there will still be a lot of mechanical drives being sold.
 
soulman901 said:
Someone did a test with Samsung Evo SSD and had written up to 700TB of data before it gave out. I'd cite a reference but I am on the phone at the moment typing this. Google SSD life expectancy and you will find it. I think SSD are at a good point now that the cost and reliability are worth it.
Click to expand...

The problems with those type of tests, is that they don't give you any long term retention numbers on the SSD. These test only measure when the drive start fails the short term data retention.

Even a new SSD will eventually loose the data stored in the cells. Most are rated at 10+ years when new, but the more you write to the cells, the less time they will retain their charge. Eventually the time will drop enough and you will not be able to read the data. If they had stopped testing that EVO drive at 350TB and put it on the shelf for a few months, there's a good chance they would not have been able to read all the data on it.

The data life on a magnetic drive is closer to 50+ years. You are more likely to have a mechanical failure than to have the magnetic data fade away.
 
Meh, it's still going to be a long while until you can justify storing media (movies, music, etc.) on SSDs.
 
Hornet said:
What I found interesting is that SSD has caught up with HDD in terms of capacity, and if SanDisk's prediction is true, HDD will never be able to keep up in terms of capacity while the cost of a SSD of the same capacity may catch up with HDD in the near future.
Click to expand...

Yeah I just hope mechanical drives don't actually die off. I find SSDs have their place, as OS drives or sacrificial high performance arrays for high I/O stuff like VMs, but for true mass data storage where raid level performance is enough, spindle drives are better as you don't have to worry about them wearing out.
 
Maxx said:
To elaborate on your question (can't edit), I forgot to say that the reliability of data is proportionate to the number of writes done on those cells, but from what I've seen on AnandTech and other places, SSD's theoretically will hold data for ten years or so.
Click to expand...

10 years for a new SSD maybe. As it gets worn out it won't retain data for as long when unpowered. At end of life expediency SSDs should hold data for 1 year when unpowered for consumer drives at some temperature, less than that for enterprise.
 
Red Squirrel said:
Yeah I just hope mechanical drives don't actually die off. I find SSDs have their place, as OS drives or sacrificial high performance arrays for high I/O stuff like VMs, but for true mass data storage where raid level performance is enough, spindle drives are better as you don't have to worry about them wearing out.
Click to expand...

Punch cards were better than HDDs because they weren't vulnerable to magnetic fields and EMP bursts, but that didn't stop them from dying off, did it?
 
Red Squirrel said:
Yeah I just hope mechanical drives don't actually die off. I find SSDs have their place, as OS drives or sacrificial high performance arrays for high I/O stuff like VMs, but for true mass data storage where raid level performance is enough, spindle drives are better as you don't have to worry about them wearing out.
Click to expand...

:confused: You absolutely have to worry about them wearing out.
 
Tankerton said:
:confused: You absolutely have to worry about them wearing out.
Click to expand...

Not based on usage though. HDDs can randomly die, but SSDs *WILL* die after a more or less set amount of usage.

SSDs in a mass storage system would basically be consumables like printer ink. Especially high I/O stuff like VMs or large databases. Especially stuff that involves lots of small writes.
 
Either way I hope they come up with good recycling programs for SSDs, or come up with a tech where it does not actually wear out.
 
Maxx said:
Any decent SSD these days if properly maintained in an average user's system will far outlast a mechanical drive in almost every case. The SanDisks I've tested, I rack up 6-7GB a day as a heavy user, and their tested lifespans based on write count would have it lasting in the 3-digit year (100) range. By properly maintained I mean you should leave 25-50% of the space free and not damage it with excessive testing, no defragging, avoid anything that does unnecessary writes, etc.
Click to expand...

If it's lasting up to 100 years then why bother even turning off anything? lol. Defrag all day long son!
 
Methos said:
You'll want two, 8TB seems like a scary amount of data to entrust to a single device. Or am I just being paranoid?
Click to expand...

SSD's aren't nearly as bad as long as the controller doesn't fail. The actual storage will just refuse to let you write again. Spinning platter drives are far worse when they fail.
 
I don' t think this question has been answered: What kind of NAND flash chip are they using in that 2.5-inch SSD?

Sandisk back in 2012, made the world's smallest 128Gbit (16GB) NAND flash, but that would require approximately 683 chips to come up with 4096TBytes (4TB).

The next thing they could use is something like Samsung's first mass produced 3D NAND flash-- V-NAND. However, that capacity is, again, 128Gbit.

It's likely they might be using 1024Gbit NAND chips developed with Toshiba. You'd only need 32 of those to make up 4TB. That would not fit entirely in a 2.5-inch SSD chassis, however, unless they were stacked against each other while still maintaining a chassis height of 9.5mm or less.

There are no 2048Gbit NAND I can find in Google, which would only require 16 chips to reach 4TB. The largest and highest density chips I can find are from HK Hynix and Samsung-- 8Gbit memory chips but these are LPDDR4 (Low-power DDR4), not NAND. They're meant for laptops and mobile devices.
 
soulman901 said:
I understand it perfectly fine. Seems like you missed the point of my post. Larger SSD will hasten the death of spin drives. Spin Drives are hitting the limits that SSD do not have. That is why I think with what San Disk announced will spurn an SSD race benefiting business and conumers.
Click to expand...

You're clueless, no offense.

NAND can't scale infinitely - that's something you should read up on.

Every time there's some larger SSD announced, people that aren't in the storage market go "yay spinning disks are dead". Incredibly naive.
 
octoberasian said:
I don' t think this question has been answered: What kind of NAND flash chip are they using in that 2.5-inch SSD?

Sandisk back in 2012, made the world's smallest 128Gbit (16GB) NAND flash, but that would require approximately 683 chips to come up with 4096TBytes (4TB).

The next thing they could use is something like Samsung's first mass produced 3D NAND flash-- V-NAND. However, that capacity is, again, 128Gbit.

It's likely they might be using 1024Gbit NAND chips developed with Toshiba. You'd only need 32 of those to make up 4TB. That would not fit entirely in a 2.5-inch SSD chassis, however, unless they were stacked against each other while still maintaining a chassis height of 9.5mm or less.

There are no 2048Gbit NAND I can find in Google, which would only require 16 chips to reach 4TB. The largest and highest density chips I can find are from HK Hynix and Samsung-- 8Gbit memory chips but these are LPDDR4 (Low-power DDR4), not NAND. They're meant for laptops and mobile devices.
Click to expand...

If it's using 32 chips, I wonder what the speed will be like with these. I hope the controller got a serious upgrade.
 
You must log in or register to reply here.
Back
Top