Anand's last SSD article misses an important point

[Axonn]

I think Anand forgot to mention another reason why the V3 MI is worth it: 32nm NAND means it has more life cycles in it.

In his article, Anand only mentions that for the extra price, you get slightly better performance. But 32nm NAND has more life in it than IMFT NAND.

Right?

 

[RaiderJ]

I think a counter-point to that is that even 25nm NAND will last for so long that it won't be an issue. Forget what the stats are, but apparently the shorter lifespan is still a *long* time.

 

[Axonn]

For the normal user, that is correct. But if you plan to have a computer for 5 or more years, and maybe even carry the SSD more than that, and if you use it intensively 16 hours per day, well, then I think that in about 6 years you could start see decreases in capacity for 25 nm.

Personally, if I'd have to buy a SSD now, I would go for the MAX IOPS exactly because of the 32 NAND. Maybe I'm just a special case though ::- D.

 

[nanaki333]

what good is longer lasting NAND if the controller screws the drive up before you barely make a dent in write cycles?

 

[Axonn]

With SandForce, your post actually makes sense *grin*. Still, only a small percentage get there. I'll wait 3-6 months before I buy SandForce's new generation.

 

[Voo]

For the normal user, that is correct. But if you plan to have a computer for 5 or more years, and maybe even carry the SSD more than that, and if you use it intensively 16 hours per day, well, then I think that in about 6 years you could start see decreases in capacity for 25 nm.

So you're saying people will be using 80gb large OS drives in 6 years? Say how many 100gb drives (which I think is what we had around 05) are you using in your PC today? Yeah me neither.

Why worry about what you'll do in 6years when in that time you'll be able to get something faster, 10times larger for less of the price you paid now? Even the largest SSDs will be uselessly small by then anyhow - heck the average 80/120gb SSD is small even NOW.

 

[Axonn]

In my case, you're wrong ::- D. I bought dual 35 GB Raptors 5 years ago and yes, I am still using them in RAID 0, and yes, they're more than enough ::- ). I don't need a gazillion GB for OS and Visual Studio and a few games. And I'm a power user as well, but not freaked out by upgrading my machine very year.

70 GB is super-enough for system, and I got an 8 TB NAS in my house ::- D. So... storage never was and never will be an issue, at least for good years to come.

So yeah, 32nm NAND will probably make sense for me, especially because I don't blow up a computer when I upgrade it ::- P. I usually give it to someone close because my systems last a very long time and it's a pity to throw away all the work & feelings that goes into them.

 

[jonny13]

Actually, 500GB was released in 2005 and 750GB was released in 2006. I'm still running quite a few of those drives in my house and about 5 other computers that I've built for family members. So, I could definitely see someone still running a 120 or 240GB drive 5-6 years from now for their main OS.

 

[VirtualLarry]

I could see keeping an 80GB or larger SSD for 5 years, easily.

I still have some 320GB SATA HDs, that I bought when they were the largest or nearly largest available, in use. (Well, one of them at least.)

 

[Axonn]

Exactly. The MAX IOPS is 120 GB which is more than enough for a damn good time. Hell, even my Dual Raptors (70 GB together) could still accommodate Windows 7 with ease + games & programs near it.

 

[aceO07]

I have a 300GB Intel 320. In the first 3 months, I've written about 500GB. There are people testing SSDs, and they're at 140TB writes and ongoing with the 40GB Intel 320.

Even IF theirs died at 140TB and assuming my rate of writes was 500GB/3months, it's still take over 7 years to kill it.

 

[groberts101]

yep.. all the lifespan nuts will eventually figure it out and laugh at the time spent moving all the stuff off the SSD when they look back in 5 years.

SLC is even going to be replaced with eMLC shortly even in enterprise environments if that's any clue to go by.

All this worry about wanting to keep a drive that'll be the size of a USB stick in 6 years is actually quite amusing. They'll be delagated to scratch disk duty, USB enclosures, and lowly netbooks by then(if the netbook doesn't already have a faster drive built in).

People that keep a system and it's components for that long are in the extreme minority and could even be considered "tight" or "sentimental". lol I let my kids use a XP/P4 machine but that's only because it was free and is all that's required for their light usage so far. Once they hit their teens and/or start PC gaming I will be forced into upgrading or handing down one of my other machines by then. Sure wouldn't want to work or do anything time sensitive on it though.

 

[RaiderJ]

In my case, you're wrong ::- D. I bought dual 35 GB Raptors 5 years ago and yes, I am still using them in RAID 0, and yes, they're more than enough ::- ). I don't need a gazillion GB for OS and Visual Studio and a few games. And I'm a power user as well, but not freaked out by upgrading my machine very year.

70 GB is super-enough for system, and I got an 8 TB NAS in my house ::- D. So... storage never was and never will be an issue, at least for good years to come.

So yeah, 32nm NAND will probably make sense for me, especially because I don't blow up a computer when I upgrade it ::- P. I usually give it to someone close because my systems last a very long time and it's a pity to throw away all the work & feelings that goes into them.

The timeframe makes sense, I pretty much have the same setup, a dedicated NAS for all my bulk storage. Still, do you see yourself doing that much activity on an SSD that you'll actually wear it out before you replace it?

With SSD reliability as it is, you might have to replace a bad controller long before the actual NAND wears out.

 

[taltamir]

For the normal user, that is correct. But if you plan to have a computer for 5 or more years, and maybe even carry the SSD more than that, and if you use it intensively 16 hours per day, well, then I think that in about 6 years you could start see decreases in capacity for 25 nm.

Personally, if I'd have to buy a SSD now, I would go for the MAX IOPS exactly because of the 32 NAND. Maybe I'm just a special case though ::- D.

1. it doesn't decrease in capacity, it turns into a read only device.
2. people did the math and it comes out to well over a hundred years, not 5 years.
3. It doesn't matter how many hours a day your computer is on, it is a question of how many erase cycles you perform. What is that "intensive use for 16 hours a day"? Video games? because unless you are running a server or some such you don't just generate data nonstop all day

 

[VirtualLarry]

1. it doesn't decrease in capacity, it turns into a read only device.

No, it decreases in capacity first, as bad blocks get remapped out. When all of those are gone, it goes into read-only mode. At least, that's how I understand it.

 

[Seferio]

1. it doesn't decrease in capacity, it turns into a read only device.
2. people did the math and it comes out to well over a hundred years, not 5 years.
3. It doesn't matter how many hours a day your computer is on, it is a question of how many erase cycles you perform. What is that "intensive use for 16 hours a day"? Video games? because unless you are running a server or some such you don't just generate data nonstop all day

Point 2 is the most important. If you're using your SSD as the OS and program drive it will last for decades. You would have to constantly benchmark the drive and use it as a storage drive to really cause it to go into read only mode.

 

[DaveSimmons]

I have a collection of old 160 - 250 GB SATA drives that still work, and even a couple of 120 -250 IDE drives, but I wouldn't bother to put them into a new build instead of using a $50 1 TB brand new drive.

So I really don't care if in 6 years a 120 GB SSD might be dead and I have to buy a new 240 GB drive that runs 4 times as fast for $80 instead.

 

[JSt0rm]

I have a collection of old 160 - 250 GB SATA drives that still work, and even a couple of 120 -250 IDE drives, but I wouldn't bother to put them into a new build instead of using a $50 1 TB brand new drive.

So I really don't care if in 6 years a 120 GB SSD might be dead and I have to buy a new 240 GB drive that runs 4 times as fast for $80 instead.

Are you saying I shouldnt buy 1000 510 120gb intel drives as a long term investment? D: 😀

 

[DaveSimmons]

Are you saying I shouldnt buy 1000 510 120gb intel drives as a long term investment? D: 😀

Not even if you fill them up with bitcoins 🙂

In 6 years the cost / GB will probably be somewhere between 1/8 - 1/16 the current cost, based on the history of both RAM and platters.

I'd love to have a 480 GB SSD for my music server (which uses lossless FLAC, currently over 300 GB) but I'm going to have to wait 3-4 years before that will make sense 🙁

 

[Axonn]

RaiderJ & taltamir: it's true that the lifetime of the drive will exceed by FAR my normal use. But I'm on the "better safe than sorry" guy. And anyway, I won't buy the MAX IOPS for that 32nm but because it's the first 120 GB drive with performance similar to the 240 GB (although I didn't understand why it did so bad in random read iometer tests - practically was the last).

aceO07, groberts101: I don't know why all the aggro only because I'm a "lifespan nut". I was merely pointing out that MAX IOPS has more value than JUST performance. It doesn't matter if it will translate or not. I think it will, because since there are no mechanical failures involved, a SSD can last 20 years. And in 20 years it can trickle down to a kid's system.

And if you think keeping a computer for more than 5 years is "sentimental", then I say that keeping a computer UNDER 3 years is wasteful. That's my opinion, of course.

Also, SOFTWARE started to lag behind. I can still make do pretty well with my 5 years old Core 2 Duo @ 2.4. And it was very good for at least 2 years. It is only now, 5 years later, that I really look for an upgrade, and I aim to keep it running for another good 4-5 years.

But hey, if you got money to waste and/or are an enthusiast who likes the game of upgrades be my guest ::- ). Just don't say that us guys keeping our systems for longer or caring about the quality of components are "life span nuts" or "sentimental". It's a matter of how you use your machine ::- ). I use it for software development and gaming @ 1920, so nothing very hardcore there.

 

[DirkGently1]

With my daily usage the 25nm nand on my Intel 320 will last ~120 years. Is the OP sure that this lifespan will not be enough for him? :hmm:

 

[Axonn]

1. *Your* daily usage.

2. All mechanical drives reach their MTBF before dying?

3. So far, nobody had a SSD for more than 5 years since they're too new. But I think that a lot of the cycles stuff is inflated marketing bullcrap. We'll talk again in 2014 when they'll start dying like flies.

 

[a123456]

So far, nobody had a SSD for more than 5 years since they're too new. But I think that a lot of the cycles stuff is inflated marketing bullcrap. We'll talk again in 2014 when they'll start dying like flies.

If you mean you think SSDs will drop like flies from hardware failure in 5 years, why buy super long NAND if it's just going to die?

The newer NAND have been tested. 40GB drives have enough write cycles to last anywhere from ~150-300+ TBs in real-life, write-to-SSD-until-it-dies testing. That goes even higher if you're talking about 120GB drives. I'm sure you can extrapolate out your personal usage and figure out how many years a given drive is going to last.

 

[Phynaz]

1. *Your* daily usage.

2. All mechanical drives reach their MTBF before dying?

3. So far, nobody had a SSD for more than 5 years since they're too new. But I think that a lot of the cycles stuff is inflated marketing bullcrap. We'll talk again in 2014 when they'll start dying like flies.

Seems you are under the impression the drives fail when they run out of write cycles. They don't outright fail, they become read only.

You also seem to be under the impression the manufacturing node of the flash memory is the only thing that determines drive lifespan. There are many other things have as much, if not greater, of an effect on drive lifespan.

 

[Cerb]

2. All mechanical drives reach their MTBF before dying?

IIRC, MBTF for platter drives is more for failure rate assumptions than a single drive life (IE, w/ a 10yr MBTF, you should average one failure per 10 drives per year).