Meanwhile most enthusiasts have an SSD in their PC and since manufacturers are successfully conquering the mass maket with SSDs, these new, quick drives are nowadays widely spread. In the past there were quite a few drives that died an all sudden death. Even today customers are not sure for how long their SSD will last and especially the P/E-cylce count (program/erase cycles) is something that worries quite a few well informed users. In this diary we're going to check what it takes to use up all P/E-cycles an OCZ Vector SSD has to offer, that features 256 Gigabyte capacity. So what is your guess, how long is it going to take until the drive refuses to work.
Reliability of SSDs appears to be a topic on which useres would
like to get additional information. In order to get a feeling for how long SSDs
are actually going to survive heavy torture testing, we're trying to write on an
OCZ Vector SSD with 256 Gigabyte capacity until the NAND flash memory starts to
go over into the read-only mode. Read-only mode is acutally a good keyword. Most
users associate dying NAND flash memory with a dying SSD. These are actually two
different things. If you're worried about using up all P/E-cycles of the NAND on
the SSD you bought, this doesn't mean, that, even if you manage to do so, your
SSD isn't responding anymore. When a NAND flash cell is out of P/E-cycles then
it goes into a read only state, meaning you'll still be able to retrieve data in
order to make a backup. Users usually associate using up all P/E-cycles with the
absolute worst case, which would be a drive that doesn't respond to anything
anymore.
[pagebreak]
Test procedure and test environment
In this test we want to get to the point where we've used up as many P/E-cycles as possible to then show what can still be done with such an SSD.
In an effort to do so, we're using Iometer where we've setup workers that do
constant, sequential writes, which is basically the one way to get rid of as
many P/E-cycles in as little time as possible. As you can see from the
screenshot below, the test drive we're running these tests on has already been
used and a total of 500 Gigabyte has been written on the NAND before we started
with the acutal endurance test. In the lower part of the screenshot you find all
relevant smart data and should you be a little bit familiar with S.M.A.R.T.
data, then you'll notice this is a perfectly healthy drive.
Diary
We're going to update this article on a daily basis. Each day we
will add a new page to this article where we post a screenshot of
CrystalDiskInfo. If there is anything suspicious going on with the drive we will
point it out and explain why this is happening.
Test Setup
Mainboard |
ASUS P9X79-E WS |
CPU |
Intel Core i7-3930K |
Memory |
G.Skill Ares 2133MHz 16GB |
Video |
ASUS GTX 650 |
Software |
Windows 7 x64 |
PSU |
Seasonic X-Series 1250 Watt |
SSD |
OCZ Vector 256 GB
Endurance: 20GB host write for 5
years ~ 36.5 TBW |
[pagebreak]
Day 1
Today there was nothing
suspicious going on. The drive was working the way it is expected to be working.
So far we've managed to write 5500 Gigabyte. If we check "Total Host Writes" and
compare them to "Total NAND Writes" we can make a different assumption. I looks
like this makes the actual "Write Amplification Factor" of the drive visible.
Since the host writes stand for the actual amount of usable data that has been
written, the "Total NAND Writes" show how much data acutally had to be written
in order to store that amount of data. Let's dived "Total NAND Writes" by "Total
Host Writes" and we might have a good approximation of the actual write
amplification factor of this drive. In this case there appears to be a WAF of
1.46. Since this value is somewhere in the between 1 and 2 it is really good.
[pagebreak]
Day 2
On
the second day there is no noticeable difference apart from the fact, that the
"Remaining Life" value, that comes from the S.M.A.R.T. data, went down by two
percent to 96 percent. Assuming the drive keeps degrading at this rate, then we
will hit 0 percent in about 48 days. Keeping in mind we're averaging about 10
Terabytes written per day we will have to write roughly 500 Terabytes to reach
the point where the NAND flash is out of P/E-cylces.
[pagebreak]
Day 7
It is been seven
days now we're trying to run out of NAND flash cycles and still the drive is
reporting 90 percent regarding health status. If things keep on going like this,
then we will have to write ten times as much data then what we've done so far.
The screenshot above is showing you, we've been writing 45 Terabyte to the OCZ
Vector with 256 Gigabyte capacity. If we now keep in mind, that OCZ is
specifying endurance for this drive with 20 Gigabyte written per day, translated
into total Terabytes written it means 35 Terabyte. In other words we've
surpassed that value by 10 Terabyte already and remaining life is still at 90
percent. It is starting to look like, OCZ in this case, was engineering this
drive with a rather large factor in mind regarding endurance. So far it appears
this drive is going to survive at least ten times the total NAND writes, which
OCZ advertises in the specs.
[pagebreak]
Day 22
As you can see
from the picture above, we've meanwhile managed to write a little bit over 200
Terabyte on the OCZ Vector 256 Gigabyte. This is in fact about six times as much
as OCZ advertises this drive for. A quick look at the "Remaining Life" value in
the S.M.A.R.T. data reveals there is still 54 percent of life left, meaning we
might have to write even more than 400 Terabyte until some magic starts to
happen. So far it also doesn't matter what we do with the drive it simply does
what it's supposed to do in the way you expect things to be done. Also
performance is still on a resonable level. In order to show you that peformance
is not degrading, we attached an Atto screenshot, which indicates that you still
get rather reasonable performance from this drive.
[pagebreak]
Day 39
In the meantime we've been writing more than 400 Terabyte on the OCZ Vector 256 Gigabyte Drive. After this toruture the "Remaining Life" Value is still at 3 percent and the drive runs perfectly flawless. This means it starts up without any issues at all and also performance is still what it should be like. Overall we're curious to see what's going to happen, when "Remaining Life" drops to 0 percent.
[pagebreak]
Day 142
This has been our first endurance test of an SSD. At the
beginning of this diary we've been trying estimate how long it will take until
the drive is going to die. It turns out that calculating with 5'000 P/E-cycles
was a reasonable thing to, since in the end the drive achieved 5'129 P/E-cycles.
We were guessing the drive would withstand our torture testing of continuous
writing for 125 days and that turned out to be close to the actual value too,
whereas the final result was 142 days.
If we start to do the maths, based on the values we gathered we're going to see
some rather surprising endurance values. OCZ specifies this drive to be capable
of enduring 20 Gigabyte written per day over a time period of five years. This
is what the manufacturer of this drive guarantees. As you can see from the table
below, based on our results, when writing 20 Gigabyte per day, you could use
this drive for 121 years. This is over 24 times more than actually advertised
and it gives you a feeling on how conservative manufacturers are when it comes
to endurance ratings of their SSDs. Even if you were writing 50 Gigabyte per day
the drive would still last 48 years and at 100 Gigabyte per day this Vector 256
GB would run for 24 years.
Lately another endurance rating used by manufacturers started to appear. Now the
vendors are talking about drive writes per day. Some enterprise grade SSDs for
example are capable of delivering ten drive writes per day over five years. Pair
this up with a drive capacity of 1 Terabyte and you will get some seriously high
numbers. Calculating drive writes per day (DPWD) for the Vector 256 GB we've
been testing here, when factoring in a period of five years, we end up with a
value of 1.89. In the end this means you could be writing the capacity of the
drive over 1.8 times per day for five years.
Since the Vector 256 GB is a consumer drive and consumers will be the ones
buying and using such a drive, a suitable use case should be applied. In case of
an average user it's highly likely that no more than 20 Gigabyte will be written
on the drive per day, which is over 24 times less then what the Vector 256 GB
could actually do.
Data written per day |
Host Writes |
NAND Writes |
WAF |
Cycles |
Endurance |
10 GB |
884 TB |
1'313 TB |
1.48 |
5'129 |
242 years |
20 GB |
884 TB |
1'313 TB |
1.48 |
5'129 |
121 years |
50 GB |
884 TB |
1'313 TB |
1.48 |
5'129 |
48 years |
100 GB |
884 TB |
1'313 TB |
1.48 |
5'129 |
24 years |