SSD Testing

[Evan-WMS]

Considering that excessive writes shorten a SSD drives life. What would be an acceptable strategy to test an SSD, but not significantly impact it's predicted life. We typically do a 4 hour burn on a new test setup, I'm looking for suggestions on how to include testing of the SSD device in this testing.

[David (PassMark)]

It isn't really an issue.
The chips used in new SSD's typically have a life span of 100,000 to 1,000,000 cycles.

So yes you are using up some of the life of the SSD (maybe < 0.001% ), but the amount is tiny.

Even if it was 3 orders of magnitude worse, a 1% loss of life would probably be acceptable to be sure the drive actually worked. (But if you could really use 1% of the disk's life in 4 hours, then the entire life span of the disk would be just a month, and this is not the case. Lifespans are typically 5 to 10 years, better than traditional HDD in some cases).

[Evan-WMS]

Thanks much for responding so quickly. Your answer is what I was looking for!

Evan

It isn't really an issue.
The chips used in new SSD's typically have a life span of 100,000 to 1,000,000 cycles.

So yes you are using up some of the life of the SSD (maybe < 0.001% ), but the amount is tiny.

Even if it 3 orders of magnitude worse, a 1% loss of life would probably be acceptable to be sure the drive actually worked. (But if you could really use 1% of the disk's life in 4 hours, then the entire life span of the disk would be just a month, and this is not the case. Lifespans are typically 5 to 10 years, better than traditional HDD in some cases).

[FFJ]

SSHD wearout

We had a similar question:

We are wondering if the BurnIn test significantly reduces a SSHD drive’s life and how the selected duty cycle impacts the wear.
Referring to this thread the impact is negligible.
However, our testing strategy is more challenging for the system. We do:

• 2 x 12 hours BurnIn tests
• with simultaneous CPU, GPGPU, 2D & 3D graphics, RAM and hard drive testing
• all duty cycles are set to 90%

The system becomes very slow and can hardly be operated while the test is running. According to the task manager the hard drive’s workload is constantly at 100%. Reducing the duty cycle to 50% reduces the workload to about 50% and the system reacts much quicker to user input.
Can you give an estimate how many writes to the SSHD are approximately performed during a 12hrs BurnIn with a 90% duty cycle and what a reduction in the duty cycle means in terms of write cycles?
Our hard disk is a Hybrid drive with a 8GB SSD partition. The predicted life span is 100,000 cycles per cell. Do we have to fear that we significantly reduce our hard drive’s life by our excessive testing? And would you recommend to reduce the duty cycle for the hard disk to 50% though?

The answer we received from the Passmark-Team:

We cannot give you estimate for your drive, as this is highly dependent on the drive and system specs and the setup configuration within BurnInTest. The duty cycle determines the delay between operations. A high duty cycle will cause operations sent one after another with a smaller delay. A low duty cycle will insert a larger delay between operations. As you witness with the load on the drive when you lowered the duty cycle from 90% to 50%.
If you want to determine the number of operations completed at the various duty cycle for this particular drive, the easiest way is to run you current configuration setup with the disk test at 90% and 50%. You can also probably run it for less than the full 12 hours (e.g. 1 to 2 hours) and extrapolate what it would be for 12 hours.

Hybrid drives only cache the most frequently use accessed data to the SSD portion of the drive. The disk test will write to the entire drive, while some data from the drive test may be cached, it's not likely that a significant portion contents from the drive test will be cached as the data is continually written,read and deleted.

Thanks a lot for the quick response. We will do some further testing and compare the different duty cycle settings.
If you have any further recommendations we would be pleeased if you share them with us here.

Cheers,
Felix

[David (PassMark)]

Update: The reply above was written 6 years ago, but we continue to get questions.
Since 2012 SSDs have got even larger and better from an endurance point of view.

As a practical example, the Samsung 970 M2 1TB has 1200 TBW and a 5 year warranty.
TBW = Terabytes written.
Note: The average life of these drives would typically be much longer than the warranty period. I don't know how Samsung did their calculation, but I would expect they calculate than 95%+ of drives last beyond the warranty period.

So if you do 1 cycle of this drive (maybe 2 hours testing to read / write 2TB in total) then you have used up 0.08% of the warranty period and maybe 0.01% of it's life. Which we think is a good trade off for knowing if the drive fully works or not.