Performance gains from secure-erasing and SSD?

[Coldblackice]

Are there performance gains to be had by secure-erasing an SSD? (the secure-erase function, not securely zero'ing out data)

I'm using an Intel 320 SSD 160GB as my main OS drive. It's been a couple years since I've reformatted and started fresh. I'm wondering if there'd be any performance gains from imaging the drive, secure-erasing it, and then laying the image back down (rather than reformatting from scratch)?

I seem to recall seeing a suggestion somewhere that it was good measure to secure-erase an SSD every now and then, even if it's TRIM-capable, because there's some manner of continual performance degradation until a secure-erase is done and the drive is factory-fresh. It might've just been SSD snake-oil, however.

 

[Ghost26]

I personally always secure erase my SSDs before reinstalling the OS because, according to some tests, there's actually a performance improvement when secure erasing the drive.

 

[bigdogchris]

The write performance will temporarily increase.

 

[Coldblackice]

I personally always secure erase my SSDs before reinstalling the OS because, according to some tests, there's actually a performance improvement when secure erasing the drive.

I always secure-erase before a reformat, of course. I'm just wondering if there are actual performance gains from doing such -- and if so -- then perhaps I should image my drive, secure-erase it, and then restore the image... essentially like a reformat (on a low-level).

The write performance will temporarily increase.

What's "temporary"? Minutes/hours/days/weeks/months? (...even further compounded by one's disk use)

 

[faugusztin]

You will lose one write cycle on your all flash drive blocks, the performance gain is from all blocks being cleared and marked as free, so there is no need to clear it when actual data is being written. And that is all the performance increase you gain. Once you write the amount of data equal to your SSD capacity (=write to every flash drive block at least once), the performnace gain is gone.

 

[ob1]

And honestly, the performance increase, (if any), will only be shown on benchmark results, not in real world performance.

 

[Ghost26]

And honestly, the performance increase, (if any), will only be shown on benchmark results, not in real world performance.

True.

I do it because I have time to do this operation, because the "gain" is not noticeable at the usage, only in benchmarks, and as faugusztin said, only for a period of time.