Sign in with
Sign up | Sign in

20 nm Write Endurance: Probably Not Something To Worry About

Intel SSD 335 240 GB Review: Driving Down Prices With 20 nm NAND
By

Endurance: Kind Of A Big Deal

Write endurance is a term often thrown around in discussions of solid-state storage because we all worry about that point when an SSD is no longer able to reliably hold our data (even though very few of us have actually seen such a thing happen).

If you have an SSD in your notebook or mainstream desktop, write endurance isn't something that should keep you up at night. It's highly improbable that you'll ever write enough data per day to exhaust the rated life space of the NAND flash cells composing your drive. Both Micron and Intel estimate that the average desktop user writes between 7-10 GB worth of information per day. Even if you use an exaggerated number, basic math assures us that you almost can't render your SSD useless within its warranty period. If you were to run into a reliability issue, it'd be far more likely to come from some sort of firmware-oriented bug.

IMFT's 25 nm NANDIMFT's 25 nm NAND

How the rated write endurance of NAND changes at each new manufacturing process is interesting to watch, though, as are the ways vendors address the physical realities of NAND wear-out. Because Intel's SSD 335 is the first drive equipped with IMFT's 20 nm flash, we are of course curious about how much, if any, its endurance level differs from the SSD 330. 

Our estimates come from monitoring each drive's media wear indicator (referred to as the MWI), which counts down from 100 to 1. Because the number of program-erase cycles a NAND cell can withstand is finite, the MWI is designed to facilitate a rough estimate of endurance.

In theory, once you reach the end of the counter, all of the memory's rated P/E cycles are exhausted. That's not to say something bad happens when you hit the bottom, but nobody wants to entrust irreplaceable data to a drive living on borrowed time, either. Naturally, enterprises place a lot of importance on the MWI because it represents “a safe zone.”

In Theory, Far Fewer Cycles: But Does That Matter?

Endurance Rating
Sequential Workload, QD=1, 2 MB
Intel SSD 320
Intel SSD 335
NAND Type
Intel 25 nm MLC
Intel 20 nm MLC
RAW NAND Capacity
320 GB
256 GB
IDEMA Capacity (User Accessible)
300 GB
240 GB
Overprovisioning
7%
7%
P/E Cycles Observed (IDEMA)
5460
1037 3117
P/E Cycles Observed (Raw)
51191538 2921
Host Writes per 1% of MWI
16.38 TB
3.60 TB 7.48 TB


Update, November 29, 2012: Intel still doesn’t provide an official endurance rating on its SSD 335, but the company’s recent firmware update (335t) now puts endurance somewhere near 748 TB using incompressible writes. If we work backwards from that result, the math ([Host Writes per 1% of MWI * 100] / Capacity) tells us that IMFT’s 20 nm should be good for somewhere around 3000 P/E cycles. This doubles the result we came away with originally, backing up Intel’s assurance that endurance doesn’t dip as it shifts away from 25 nm lithography.

Now, we’re not huge fans of these updates. We’re reminded that, with a simple firmware, a vendor can change the algorithm used to calculate MWI—an essential SMART field monitored by enterprise customers and power users as a way to quantify the health of their SSDs. Fortunately, if you buy a retail drive, you’re covered for three years no matter what, even if the MWI hits 1. Intel’s OEM SSD 330/335 drives do specify coverage for three years or until the MWI hits 1, though, so our discussion isn’t as purely theoretical as it might sound.

But we weren’t particularly concerned about endurance when we were measuring 1500 P/E-cycle endurance, and we’re still not. Even going by the bugged firmware’s halved rating, assuming 10 GB of writes per day, it’d still take more than 100 years to wear down the drive’s NAND using our workload.

Display all 14 comments.
This thread is closed for comments
Top Comments
  • 13 Hide
    mayankleoboy1 , October 29, 2012 4:07 PM
    Read only the conclusion. Most of these SSD's are "me-too" clones using SF2281 controller. Most have similar performance wins and pitfalls.

    I was super excited about Samsung 840. But these are meh.
Other Comments
  • 13 Hide
    mayankleoboy1 , October 29, 2012 4:07 PM
    Read only the conclusion. Most of these SSD's are "me-too" clones using SF2281 controller. Most have similar performance wins and pitfalls.

    I was super excited about Samsung 840. But these are meh.
  • 3 Hide
    christophermarti , October 29, 2012 5:12 PM
    I would say that your estimates about P/E cycles are incorrect. You also do not mention (on purpose) two modes SSD's (99%) operate in: performance mode (not filled to 90%) and storage mode (filled 90% and more). You also lack to mention that in "middle of" P/E cycle exhaustion SSD's will slow down their speed due to preserve P/E cycles and "survive" to meet warranty agreements.

    http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm/page211

    From test exposed in this forum You can draw conclusion how good MLC used in X-25v 40GB SSD were (more than 35000 P/E). Also that longest "standing" SSD is Samsung 830 256GB, which also do not (as an exception) slow down considerably when it passes 1PB Host Writes mark.

    Although I strongly do agree that seing writes above 10GB per day is rather rare. I'm myself using 80GB X25-M for 4 years and only 4,09 TB and i is possible that it will hold up to 1400 - 3400 TB of writes! That's amazing. What's more, I have it in Dell E6400 on Vista (no Trim, just Intel toolbox).
  • -3 Hide
    christophermarti , October 29, 2012 5:19 PM
    http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm&p=5148307&viewfull=1#post5148307 - 1000 P/E cycles (not even close to 35100 of 50nm old V40 GB).
  • 0 Hide
    alidan , October 29, 2012 5:59 PM
    christophermartihttp://www.xtremesystems.org/forum [...] ost5148307 - 1000 P/E cycles (not even close to 35100 of 50nm old V40 GB).


    keep in mind that is still 240000 gb of data at minimum
  • 2 Hide
    abbadon_34 , October 29, 2012 8:35 PM
    so if it's firmware crippled, can we just flash a the firmware and get a better drive? someone needs to do some testing....
  • 1 Hide
    acku , October 29, 2012 11:42 PM
    christophermartiI would say that your estimates about P/E cycles are incorrect. You also do not mention (on purpose) two modes SSD's (99%) operate in: performance mode (not filled to 90%) and storage mode (filled 90% and more). You also lack to mention that in "middle of" P/E cycle exhaustion SSD's will slow down their speed due to preserve P/E cycles and "survive" to meet warranty agreements. http://www.xtremesystems.org/forum [...] nm/page211From test exposed in this forum You can draw conclusion how good MLC used in X-25v 40GB SSD were (more than 35000 P/E). Also that longest "standing" SSD is Samsung 830 256GB, which also do not (as an exception) slow down considerably when it passes 1PB Host Writes mark.Although I strongly do agree that seing writes above 10GB per day is rather rare. I'm myself using 80GB X25-M for 4 years and only 4,09 TB and i is possible that it will hold up to 1400 - 3400 TB of writes! That's amazing. What's more, I have it in Dell E6400 on Vista (no Trim, just Intel toolbox).


    Our calculations and endurance protocol are not effected by any speed slow down, and in every test, we've confirmed our methodology applies a WA~1%. Thus, are estimates are correct and apply to the NAND itself.

    Second, the speed of a drive has no inherent bearing on endurance. It only affects how fast you can get there. Second, you're referring to a throttling effect, which is a different topic completely. Our analysis was specific to the NAND itself.

    As further verification, another one of my peers (at another site) independently came to results for the SSD 335 similar to ours.
  • 0 Hide
    cumi2k4 , October 30, 2012 2:01 AM
    how come vertex 4 did not get tested? didn't ocz offer 5 years for their vertex 4, and i believe they're offering them at cut-throat pricing?
  • 0 Hide
    mayankleoboy1 , October 30, 2012 2:20 AM
    One thing i notice is that with an Antivirus program(Kaspersky Internet security) running in the background, which 95% of Toms readers would have, the AV program reads and writes data almost continuously. In a typical day, the AV program can write 5-6GB of data.
  • 1 Hide
    jabliese , October 30, 2012 1:28 PM
    Power consumption made me laugh. 1 watt difference between the best and worst. Time for another 1 line standard SSD review comment, "Power usage on SSD's matters little between the best and worst."

    And please keep reiterating the most important thing about an SSD is getting one. Ran into SSD resistance on another forum just a couple weeks back, which, at today's price points, blew me away.
  • 0 Hide
    acku , October 30, 2012 4:05 PM
    cumi2k4how come vertex 4 did not get tested? didn't ocz offer 5 years for their vertex 4, and i believe they're offering them at cut-throat pricing?


    Look again. It is included in our benchmarks.
  • 0 Hide
    acku , October 30, 2012 4:06 PM
    mayankleoboy1One thing i notice is that with an Antivirus program(Kaspersky Internet security) running in the background, which 95% of Toms readers would have, the AV program reads and writes data almost continuously. In a typical day, the AV program can write 5-6GB of data.


    By default, Anti-virus programs (like norton) now implement "smart scanning." One something is scanned, it's not scanned again unless the file signature has changed. Dramatically speeds up a scan and reduces the amount of IO traffic over the course of a day.
  • 0 Hide
    Anonymous , November 1, 2012 5:02 PM
    please provide a link with the OCZ drives that you used casue they are having multiple products with the same name but the speed vary.
  • 0 Hide
    mynith , November 13, 2012 8:16 PM
    I reckon 10 GB per day is perfectly possible. Possibly much more. You fail to take into account things like having swap drives, which may or may not see high throughput. Also, if a laptop is setup to not only suspend to RAM but also to disk upon hibernation (Macbooks suspend to both at the same time by default, and my Kubuntu-installation does as well if the battery runs a bit low), which happens many a time per day, you'd get there quite quickly, I presume. 7 GB per day is a very conservative estimate in my view.
  • 0 Hide
    myardor , December 15, 2012 8:24 PM
    Got one for $139.99 with no taxes and free shipping at egg head. So about 58.4 cents/gb