Whens the best time to get an ssd?

Everyone knows that 32Gbit is 32Gbit, it doesn't change because it is 180nm or 10nm.

However, the natural of NAND would make the SSD vendor to find some way to compensate the lost cycle or else it may not be even able to meet the normal expected life. That can only be done by more OP, effectively reduce the usable size. Even if they don't do it, the buyer should.

As mentioned in another thread, we are not exchanging legal documents so I see no point of writing 15 clauses to just make a point.

 

How can a user compensate for their own over-provisioning? Intel says you can do it by leaving unpartitioned space so it is never written to, but does it work the same for other SSD's? I haven't seen it documented for other manufacturers. I still don't understand why they even bother going to a smaller fab sizes since it greatly reduces life cycles. Makes no sense.

 

I'm still holding out. I would like a 128GB SSD but those are still a little too pricey for me, I might be able to get by with a 64GB SSD, but again, the price hasn't hit the sweet spot yet where I'll just replace my 120GB HDD.

 

OP applies in almost all SSD, no matter which vendor given my understanding of how SSD works. And Intel's recommendation is the way to do it if a user want it. Leaving unused space is the same, if it supports TRIM properly.

You have a point that going forward, smaller gate length will become a real issue and if they don't find a solution, 25nm or 18nm may be the last one.

 

till then, we get memristor storage. problem solved

 

there wasn't even a problem to begin with for the current and coming SSD using 25nm. 3000 p/e cycle would still meet the expected life need(give it a slight increase in OP) other than extremely small size one used as a scratch disk.

 

3000 to 5000 cycles is a 66% increase. That would require a significant amount of OP to have it meet that life, and not sure if it would offset the reduced cost of the smaller fab process.

 

people don't understand lifetime and thus want much more just out of panic. see htwingnut

 

It really depends on usage pattern. My 80G is rated at 7TB(and I am quite confident that it would be more than that given Intel but lets take that number), a 40% reduction would mean 4TB, assuming just replacement with 25nm using the same controller.

I am not a heavy video user but still bang it a lot during work and recorded about 2-3GB per day. That would give it about 1000-1500 days(assuming 365 but in reality it is more like 250), about 4 years. If I kick it up a notch and have the 160GB(should have lower price when 25nm is used), it would be doubled.

If they can improve the controller a bit say further reduce the WA by say 20%, I am seeing 10 years for my usage of a 160G. I would not worry about that. How many people use their computer for over 5 years before upgrade ?

Of course each individual is different but in general, it is not an issue.

 

especially how many ssd freaks like we are in here will use the current ssd in their main system as main ssd for the next 10 years? no they'll get a new, sata3 one in 1-2 years, and then the old one will be used on a much less used system (family, gf, what ever).

it's simply nothing of importance. they life long enough. and they don't suddenly die. which is much more important to me, having had so much hdd crashes in the past rather randomly.

 

Personally, I am not too worry about either of them(afterall, 95% of the computers out there are still using HDD).

But I agree that SSD endurance is an even lesser concern than HDD crash.

SSD controller(firmware) bugs is the real issue and the number one criteria in choosing SSD, IMO.

 

Why not using one just for system? You don't need 250MB/s speed to listen to music, watch a movie etc. (for those use normal, cheap disk).

Right now my full system is 3.8GB, I could fit it into 4GB pendrive... Yeah, so expensive.

 

+1 rep, best said. The recent OCZ fiasco has scared me into other vendors now...

 

I'd wait for the C400, which looks like it'll be relatively "cheap".

 

yea but the c300 is having problems with the latest firmware and they havent fixed that? and they want to put out c400 they at least need to fix the c300 before i put any consideration in crucial.

 

It was an even bigger drop when we went from 50 nm to 34 nm (10,000 to 5,000 cycles), and I don't remember many obvious changes in capacity and overprovisioning. I suspect as things moved into real world testing, they found that their original estimations were overly pessimistic, and that they feel they can get by with things as they are. As well, as stated elsewhere, controller improvements can reduce write amplification, thus squeezing more life out of limited P/E cycles.

As for moving to the smaller process, don't forget for the possibility of refinements that will raise P/E cycles as well. Inital runs of 25 nm NAND had a P/E cycle of 1000(!). It's the later released runs that are getting up to 3000 P/E cycles. With further refinements, it may even be possible to get even more P/E cycles out of it.

 

Well, the difference is 10,000 cycles or 5,000 cycles still overs a reasonable lifespan. I guess larger sizes are better obviously. If you take into account the 160GB Intel that's 160gb x 5000 writes = 800,000 GB assuming equal wear leveling. That equates to 438GB / day over five years. Even if it's half that, that's a pretty substantial amount of writes per day. 3000 writes = 263 GB/day, Go much less than that then you're looking at significantly depreciated life.

 

I think that this whole issue of NAND write-cycle lifespan is way overblown.

(1) The write lifecycles of drives you can buy today are far better than the write lifecycles of previous SSD's, because manufacturers are getting smarter about over-provisioning and wear leveling.

(2) You will not find any examples, anywhere, from anybody, or on any internet forum, of SSD's that no longer write because they have expended all of the write cycles of NAND memory. This is a boogie-man problem like Y2K, and not an actual real-world problem that will actually affect anybody.

(3) Nobody writes 263GB/day, or even 10% of that at 26.3GB/day. Typical consumers will install your OS + apps + games, and then you occasionally write a few GB here and there. The only people who can reliably write several 10GB's or 100GB's per day are enterprise users that run apps like databases... and those users will be using enterprise-level SSD's using SLC-based NAND, and not the consumer-level SSD's using MLC-based NAND.

(4) Assuming that you are a consumer-level user that does write 263GB/day... your drive will last an estimated 5 years? How many of you actually own drives that are 5 years old? Most people will replace drives long before that for speed reasons, capacity reasons, or to replace a dead drive.

(5) Even if an SSD does use all available write cycles, the drive does not "die". It reverts to a read-only state. If, by some miraculous stroke of celestial luck, you are someone who does not fall into category #1,#2,#3,or #4 above, and you do get a drive that is in read-only state...

* You would probably be the first person in the history of SSD's to do this.
* You can buy a replacement drive (which will be faster and have larger capacity), and simply copy over all of the data from your read-only drive to your new SSD.
* You would have easily squeezed every bit of life out of your SSD, and squeezed every penny's worth of value that you spent on that "dead" SSD.

 

Well, nobody said initial p/e cycles were an issue. It's just the 3,000 cycles starts to border on questionable, and sure you can OP the drive, but already it's been shown to have diminished performance with the drive over 60% capacity. So you already should consider a drive with double the capacity you require. What do p/e cycles go down to with the next size down (what is it 18nm?) is it 1,000 cycles? That's ridiculously low.

 

no, because normal users don't do 100GB/s each day 24/7, the whole year. writes, that is. not even close. which according to your calcs would be what you'd get compared to the 3000 cycles.

 

Tomy B just posted a link in the SSD thread:
http://forum.notebookreview.com/har...marks-brands-news-advice-876.html#post7177687

to this Micron statement on 25nm drives:

Micron Innovations Blog How to Build a Solid 25nm SSD (and how not to)

" The 25nm C400 offers the same endurance as the C300 for the 128, 256 and 512 GB models—72 TB TBW. This is equivalent to 40GB per day every day for 5 years, and far exceeds the patterns of any PC user. The 64 GB drive endurance is rated at 36 TB TBW—that’s 20 GB per day over the same time period, which still exceeds typical consumer use patterns."


I'm not sure now even where the 5,000 p/e cycles come from because even the 128GB drive with 40GB per day is equal to only 570 p/e cycles 1:1 ratio. That's only about 10% of the claimed p/e.

 

does it mean the Marvell controller now outshine Intel's given that the 510 series is apparently using that too.

 

whan an advertisement Micron did for themselves ..... stating how their new drive is what all SSD manufacturers should take example of ..... and all this in their own blog, LOL

it's common knowledge that an SSD should hold a constant storage and should improve performance .... LOL

also, they specified the Total Bytes Written ... what a BS. It depends to the highest extend whether or not you write 4,096 files of 1 byte, or you write 1 file of 4,096 bytes ... in which case the TBW will be same.

again, what a BS.

 

What are you talking about ? All SSD are AFAIK function in 512B sector mode so their smallest write unit is always 512B Host write. But in reality(say Windows NTFS), it would always be multiple of 4096.

What do you mean by BS ?

 

We just moved to 25 nm, so I don't know how far they've gotten with even prototyping the next size down. And even then, as I mentioned before, prototypes won't necessarily be a good example of final P/E cycles. The first 25 nm runs only managed 1000 P/E cycles, and the more recent production runs are now 3000 P/E cycles.

The cycles I mention here are what the NAND is rated for; obviously, given write amplification, writing 128 GB to a 64 GB drive may actually result in using 3 P/E cycles (or maybe even more) as opposed to 2 (or less, if using Sandforce style compression).