Crucial M5 or Samsung EVO??

Im going to buy a laptop

and was wondering which SSD will last longer
I never had any experience with SSD and have read that they die out faster than HDD's

I was looking at the 120 and 250 gb variant

 

In what expected day to day workflow?

 

some gaming, web surfing, a bit of coding, watching movies
research and word document

 

They will last u pretty much forever.

 

For your uses, both will survive for a long, long time (as in, at least a decade or so). Hypothetically, MLC NAND can last for well over a century, though practically, there isn't much difference for you.

Go for whichever one is cheaper for the given capacity you want.

 

Crucial it is then


In a hypothetical scenario:
in what estimated work flow
will it then give up say in 4 years?

 

You have to run either of them extremely hard in order to kill it in four years. Say, several dozen GB of data writes per day, every day, which just isn't going to happen with your workflow by accident.

 

I do a lot of downloading
although I will just probably store them directly in the 700 gb hdd

Thanks for the replies
I really wanted to get my hands on a SSD as performance significantly increases (boot times)

 

If you're worried about downloading, you can configure Windows to automatically direct downloads to the HDD instead of the OS drive.

 

to kill it in 4 years. Maybe over 500gb of writes per day on a 250gb ssd?

 

Thanks for the info

now I feel confident going for a SSD as my OS

 

Over 6 months I have written 1.2TB on my SSD.
I do the usual stuff: Watch movies, game, torrent, code etc. I use mine every day and its the only drive I have in my notebook.

Techreport have made 500TB writes on their Samsung 840 240GB with TLC and its still going strong.
With that in mind, my Samsung 840 EVO should last me 210 years years. Even more since I have 500GB and the wear on the NANDs will be smaller since the OS scatter usage of the cells in the drive.
There are other considerations here, but it will easily last me 50++ years if nothing other in the SSD breaks before the TLC.

You be much better off with a 250GB EVO than a 240GB M500.
Its a massive performance difference between those two.
540/520MB/s - 97K/66K IOPS
vs
500/250MB/s - 72K/60K IOPS

 

^agreed with Cloudfire, the (much) slower write speeds on the M500 is not just worth it for the $30 you'll be saving.

And just to throw some real numbers out there, I've written nearly 2.7TB to my 840 Pro (512GB) over 4 months (~23 GB/day), and it's only gone through 8 P/E cycles thus far. The 22nm MLC in 840 Pro is rated for 3000 P/E cycles I believe, so at the current rate, the MLC NAND will last well over a century (3000/8 x 4 = 1500 months = 125 years). I'm almost certain something else (controller, cache etc) will fail long before the MLC NAND wears itself out (not that I want anything to fail soon *knocks on wood furiously*).

To present a grossly simplified theoretical calculation, 512 GB x 3000 P/E cycles / 23GB per day / 365 days = 183 years. So theoretically, a 512 GB SSD with 3000 P/E cycles should last 183 years based on an average of 23 GB of writes per day.

The reason why this is much higher than the 125 years calculated above based on real life usage is most likely due to write amplification. If we calculate the write amplification for my 840 Pro as thus: ( Number of erase cycles x Physical capacity in GB ) / Amount of data written from the host in GB = 8 x 512 GB / 2.7 TB = 1.517

Now if we divide the 189 years from theoretical calculation by the write amplification factor of 1.517 we get 124.6 years, which is remarkably close to the 125 years I calculated above.

=======================================================================================================================

With all that out of the way, lets take a look at the 250GB 840 Evo. It has 19nm TLC that is rated for 1000 P/E cycles. The theoretical endurance, assuming a write of 23 GB per day, is 250 GB x 1000 P/E cycles / 23 GB per day / 365 days = 30 years.

Even assuming a write amplification of 3x, the drive will still last you 10 years with 23 GB of writes to it EVERY SINGLE DAY (atypical for average consumer workload). So unless you plan on keeping your computer around for 10+ years, write endurance is not something you'd have to worry about.

 

Sounds like you use your computer more than I do

Yeah its pretty good knowing that SSDs are built to last. All you have to do is make sure you get a decent one with a good controller with a working firmware I say. Looking at you OCZ...
I hope Toshiba will whip them up to shape with their recent acquisition.

I have owned Intel X25-M G2, Crucial C300, Intel 510, Samsung 840 Pro and 840 EVO and I have never had any trouble with any of my drives.

 

Based on my calculations above, and assuming a write amplification of 1.5x, 1000 P/E cycles for TLC and 3000 P/E cycles of MLC:

For a 250GB SSD

TLC: 250 GB x 1000 P/E cycles / 1.5x write amplification / 4 x 365 days = 114.16 GB per day
MLC: 250 GB x 3000 P/E cycles / 1.5x write amplification / 4 x 365 days = 342.46 GB per day

 

While the above is useful on a theoretical level, the lifespan of an SSD doesn't depend on only how much is written to it; but also how.

Sequential writes (which is what most assumptions above are) can be very, very high (decades, if not centuries as noted above) and also very, very misleading if they don't match your workflow.

Random 4K writes can kill almost any MLC/TLC SSD in a few weeks. There is a difference...


That is why OP'ing (reduces WA), in addition to leaving free space (for the O/S, programs and Scratch disk use) and making sure you buy an SSD that has a controller that is fully optimized with optimally interleaved nand chips and fully populated channels.

That is why some SSD's are not recommended for smaller than 480/512GB capacities (M500 Series) while others are only 'best' at the 240/256GB capacity point (Intel 520 Series, SanDisk Extreme I's).

When considering the capacity you can live with; you need to consider OP'ing (I recommend 30% while Anand recommends 25% still), free space and the above controller optimizations/nand interleaving methods manufacturers can decide upon depending on the die size of the nand chip they use and the controllers used to drive those chips.


If you're considering replacing this system (and SSD) within ~18 months; buy whatever is cheapest.

If you're considering keeping this system as long as possible (including the SSD); then I would recommend a SanDisk Extreme II 480GB SSD instead. Your cost of storage subsystem over the expected lifetime will be substantially lower and you'll enjoy the fastest performance too at the same time.


Here is a great deal on the recommended SSD ($289.99) at this time;

See;
Sandisk Extreme II Solid State Drive, 480GB at Memory Express


But keep in mind that they go on sale like that regularly too, if you can't pick up one right now.


Hope this helps.

 

Somehow I knew the sequential vs random write would be brought up lol

Yes, you're absolutely right, heavy random 4K writes would kill any non-enterprise grade SLC SSDs quickly. But here's my question: how relevant is the "heavy random 4K writes" scenario in the context of an average workflow for the average consumer?

The reason I put together all the calculations in my post above is to show that given an average consumer with an average workflow, an MLC SSD will last well over a century, and even TLC will last at least 10 years. I dare say my workflow is

a) heavier than your average consumer, given that I write 20+ GB to my SSD everyday (I'm referring to the general population, not just the members here on NBR)
b) representative of the workflow of at least 80% of SSD users -- I browse, game, livestream, download, do productivity work, benchmark, and engage in the occassional computation heavy stuff -- and this seems to match the expected workflow of the OP

Yes workflow can and will change with time, but if yours is such that you can kill an MLC drive in weeks, you probably shouldn't be looking at consumer-grade MLC/TLC drives.

Anyway, don't want to digress much further, but my point is, if you don't plan on keeping your system for more than a few years, and have an average workflow (ie nothing out of the ordinary), write endurance is not something you should be concerned with.

Also the 840 Evo 500GB is now on sale at Amazon for $251.99

 

Don't hold your breath waiting for a coherent answer.



Sent from my iPhone using Tapatalk

 

Those are absolute bare minimums. Pretty much all real world testing has resulted in at least three times that amount. Why people are so enamored with the failure of an SSD is beyond me. In three years you will be wanting the drives that are twice the capacity at half the cost (per GB) with twice the performance. If you're really concerned about the life of your drive and it's that important, pony up for an enterprise class SSD. Otherwise, it will be perfectly fine.

Read these good articles on the bases of SSD's:
http://storageeffect.media.seagate.com/files/2010/08/Storage_Device_Longevity_FINAL.pdf
http://www.kingston.com/us/ssd/enterprise/best_practices/enterprise_versus_client_ssd

 

I wote cheaper option. If you are not moving serious amounts of data (20+gh) on a daily basis and download stuff directly onto your second hdd there is no point in buying marginaly faster drive even for 10$. I just wejt from samsung 830 128gb connected to sata3 to crucial m500 240gb connected to sata2 and i can not seem to see any difference in speed it takes to boot or to load various programs. Yes benchmarks say my old drive was faster but i just fant notice it in my day to day work (and gaming). I would rather spend extra money on a 6 pack and pizza than on a marginaly faster drive which i wont notice.

Sent from my C1905 using Tapatalk

 

Passing down the SSD to your grandchildren is more important than enjoying better performance while you live

Best regards,
tilleroftheearth

 

Well, let's see...SandForce? Intel 8MB bug? It's not like SSDs of the yesteryear haven't instilled fear in a lot of people...

I've been somewhat of an early adopter when it comes to this technology and it hasn't exactly been a painless journey.

 

Two issues which had nothing to do with failure of the NAND, which I think is what HT was implying with his post.

 

Correct on the technical level.

However, the fact that NAND itself is unlikely to fail means nothing to a an average consumer who's afraid of the drive failing altogether for whatever reason, and we have seen a fair bit of that in the past...

My $0.02 only...

 

it seems to me that he's more concerned about the long term performance as the drive is used than its lifespan.

In any case, it's nice to know some of this stuff; each individual can figure out whether it's relevant to them or not.

 

I see we have all failed economics 101: cheaper and better is better than cheaper and cheaper. lol...

Always.

 

Answer the question as it was presented to you.

 

That is how it was answered.

So, I failed to provide an incoherent answer to satisfy your wishes... so instead you fail to understand the answer?

Life lesson: don't buy cheap twice, buy quality once. Especially if quality comes at a lower price.


Btw; the price quotes I used were for my immediate area; ymmv. Try to read for comprehension of the topic at hand - not the topic of 'disagree with 'tiller no matter what he says').

If you are into buying inferior products (which we have agreed TLC is in this thread... re-read it if you need to...) that is your call. But recommending for others to buy them with their own funds is not kosher with me. Paying even 10x for a quality item is cheaper in the long term than saving half, ime.

Live for today; spend for tomorrow.

 

You're really trying to cram your philosophy down everyone's throats. That's sad. There's some solid advice in this thread, and even yours had a bit of value, INITIALLY. Now, it's more trolling than anything. You kinda remind me of the egocentric bald guy in The Princess Bride.

 

Let's just say that different people have different priorities when spending money.

I've bought several SSDs in my time and each has kept running until superseded although the only one that I was anxious to replace as quickly as possible was the 128GB Sandisk U100 that Samsung bundled with a lot of Series 9 notebooks. If Sandisk are good at designing SSDs then their design objective with the U100 was to bottleneck the notebook's performance. I've known no other storage device that could cause a computer to freeze for several seconds while loading a big file (and needing to dump some RAM to virtual memory at the same time).

I'm now using a 1TB EVO mSATA primarily because it gives me twice the storage capacity of any other mSATA SSD. It also appears to be frugal on power when on idle which is of advantage for those occasions when I want maximum run time away from a power socket.

John

 

Sandisk makes cheap SSDs too, Asynchronous NAND Cheap sells.

 

What is sad is people like you that don't offer anything to the discussion at hand.

I was asked for my opinion (twice).

 

No man is a prophet in his own land.

 

I'm gonna defend TotE, he's just putting the info out there and people like me are very grateful for it because I haven't been keeping up with computer hardware at all so have no idea what to look for when buying SSDs. I specifically look out for what he has to say because he seems very clued up about SSDs and it's easier than trawling through articles on Anandtech trying to make sense of all the different charts. It's not like he's asserting his points with no evidence or reasoning to back it up.

 

To add to that, it's not like anyone has to listen to anyone's advise. People can simply ignore him (or others) if wanted, though more information is better than less and while I don't really agree with tiller most of the time on his SSD recommendations, it's still nice to have a second opinion.

 

Indeed. I've seen one too many people just follow the "I want the cheapest possible" trail, and many of them lived to bear the consequences.

 

True, though sometimes it pays off (not just in computer tech). You just have to shop smartly and make sure that the product meets your expectations.

 

I'm all for being frugal. Never for being cheap.

Buying garbage - in *any* area of life - because it's supposedly cheap is just the wrong way to go about things in my book.

There's a reason why the (ancient as well as modded) ThinkPad that I'm typing this on sports an Intel SLC SSD as a boot drive...

 

It seems like deja vu because I could've sworn there was a similar thread to this not long ago. Anyway, to understand the choice between the M5 and the EVO, you must understand the underlying engineering philosophy behind both drives.

basically what I wrote for the previous post:

How do electronics get cheaper/faster with time? Moores Law,

However, NAND simply doesnt scale well with Moore's Law. Endurance plummets and reprogram latencies increase (unlike CPUs, GPUs, RAM etc) on smaller lithographies.

With this dilemma:
The question has always been TLC at a mature process node or MLC at the cutting edge process node? Furthermore, is it better to go TLC for higher density or MLC with huge pages (aka bigger die area)?
Performance, capacity, endurance and cost will always be engineering tradeoffs.

Simply, TLC costs less because it gives significant advantages during manufacturing (higher density, smaller die area) provided it is manufactured on a mature process node (TLC actually costs a lot more to make than MLC on cutting edge nodes due to poor yields), obviously the performance tradeoffs include lower endurance and high latencies. In a consumer product, TLC is thus advantageous for Samsung as 90% of consumers use their SSDs in burst workloads with lots of long term data retention. As I understand, the EVO uses a portion of the NAND die which is programmed as SLC to act as a write buffer. The buffer pretty much equalizes the TLC's slow write disadvantage vs MLC drives in consumer workloads while the long term data retention habit of most consumers renders the high endurance MLC NAND moot. Thus, in theory, the average person can enjoy an SSD with performance which matches their usage style with good capacity at a lower cost.

The alternative approach to TLC is to use large page MLC drives on cutting edge nodes (mostly championed by Micron). While large pages give the MLC similar densities as TLC but the larger die sizes must be offset by the newer process node to be price competitive (I.e. MLC lithography must always be full or at least a half node ahead of the TLC method). Combined with generally better yields on MLC vs TLC (NAND is usually binned according to endurance) this approach produces cost effective, high density MLC NAND which is slightly more expensive than Samsung TLC. However, due to the cutting edge nature of the lithography, it is difficult to immediately match the volume of usable die output of the TLC approach. While large page MLC retains the traditional advantages of higher endurance, the performance still suffers somewhat (though reprogram latency is still superior to TLC) compared to small page MLC which are purely shrunk by lithography. From a consumer point of view, the real advantage of this kind of SSD is better endurance (than TLC), high capacity but slightly worse uncached write performance compared to traditional MLC. I daresay however, this approach is a harder sell than the 840 EVO simply because it doesn't align with the majority of consumer workloads as well (due to the lack of SLC cache). However, for the prosumer, enthusiast and professional, the endurance and sustained performance is a much more attractive proposition.

Therefore, the decision to go with the EVO or the M5 is completely dependent on your workload. While the EVO uses slower and less durable TLC NAND, the combination of a fast, proven controller and SLC write cache means that it achieves a level of performance which surpasses most MLC drives in its price range provided you have a mostly burst oriented usage style with long term data retention (i.e. low rate of data turnover) or if you are after a large capacity OS drive (and cannot utilize a secondary drive).
The M5 is slower in burst workloads but the high density MLC cells are far more durable than TLC (by a factor of 3 at least). This drive is more appropriate if you are using it as a storage device with a high rate of data turnover or if you need good sustained write performance (e.g. if you are using the drive as a disk cache).

For example, I use a Samsung 840 500gb drive as my Steam Games folder as my workload is primarily Sequential Read oriented with limited data turnover (games usually installed once and is retained).

 

^what if you wanted the best of both worlds ie high burst and sustained read/write performance?

 

I'm not answering for Marksman30k, as I value his opinion too.

Sandisk Extreme II 480GB It has nCache, Sandisk's SLC layer and MLC It's sustained performance (consistency) is top notch. It's "worst case" performance is also top notch. Provided it's on sale, it can be reasonably close in price.

 

Lol, then you get pretty much any flagship 2013-2014 model Tier 1 SSD that utilizes MLC that is optimized for performance (instead of high capacity/price). Basically with these high performance drives, you cant go higher than 512gb at the moment and you will be paying $1 per GB or more.
Consider:
Samsung 840pro: this drive is probably the most optimized for burst performance at the moment with acceptable sustained performance, however performance consistency is awful unless a decent level of over provisioning is given (i.e. you have to set aside at least 25% of the SSD manually).

OCZ Vector: this used to be the king of write performance but it suffers from weaker random read performance and......OCZ's reputation

Sandisk Extreme II: interesting drive because it is a really good balance of burst, sustained and performance consistency. Plus its reasonably priced too. Probably the most well balanced implementation of the Marvell 88SS9187 controller + Performance MLC NAND + SLC Cache i've seen to date.

Intel S3500: this is the king of sustained performance and write consistency, obviously has weaker burst performance (due to the MLC-HET pedigree of the NAND) but it is almost unrivaled when it comes to heavy sustained mixed I/O at low latency. Powerloss protection is a nice perk but definitely the priciest .

EDIT: after some more reports, the Intel 730 seems to be worse than the Intel S3500 from which it was derived in favour of very slightly better burst performance. Cost difference between the two actually favours the Intel S3500 which is also surprising. As such, I cant actually recommend the Intel 730 over the Intel S3500 considering the latter is actually cheaper in some cases.

 

Have the 840 Pro, haven't run into issues regarding sustained performance (not part of my normal workflow anyway). Consistency issues haven't crept up either since it's at 25% OP, but I can honestly say that between the 10% OP I used for the first 3 months and the 25% OP after that, there was no perceivable difference or any sluggishness to me. If anything upgrading the ram from 1600 to 1866 made the system snappier.

OCZ has gone bankrupt, so I'll just leave it at that.

SanDisk Extreme II, now that's one I may actually try sometime if/when the price is right, and I'm allowed to return it within 30 days with no strings attached.

Intel not going to consider since burst performance is much more important than sustained performance for me.

 

OCZ is now under Toshiba. Likely, the same NAND as Sandisk/Toshiba, from Vertex 460 forward. The lower pricing shows being part of a fab. More importantly for OCZ's rep,, not using the consumer to beta test in a rush to market.