Microsoft will change the default way that they handle USB data devices

[williamcll]

15 minutes ago, Dabombinable said:

I completely disabled write caching, aka "optimize for quick removal" is enabled. It started as an effort to find out why I was having Windows (and only Windows) corrupting almost immediately after install on my 240GB SSD. Turns out that its just another bad SSD from OCZ and there were no negatives to actually having write caching off.

So that means like this?

 
or like this? Pardon me for sounding stupid.

[Dabombinable]

2 hours ago, williamcll said:

So that means like this?

 
or like this? Pardon me for sounding stupid.

The first one. I have the "quick removal" option due to hot plug being enabled on 2x of the ports.

[williamcll]

Well, I've ran a test on my 970 EVO plus and this is the result:

I see a performance loss for I/Os and MB/s per second as well as lower average I/O, but a much faster maximum I/O, i'm not sure if it is worth it. I guess this doesn't work with SSDs at all?

[mr moose]

39 minutes ago, williamcll said:

Well, I've ran a test on my 970 EVO plus and this is the result:

I see a performance loss for I/Os and MB/s per second as well as lower average I/O, but a much faster maximum I/O, i'm not sure if it is worth it. I guess this doesn't work with SSDs at all?

I am not too sure you can effectively measure the difference using a program like that,  write caching will vary greatly depending on the task and whatever else is going on at the time.  IO meter tends to just test the specific of each drive and not how the OS routes the data. 

[williamcll]

22 minutes ago, mr moose said:

I am not too sure you can effectively measure the difference using a program like that,  write caching will vary greatly depending on the task and whatever else is going on at the time.  IO meter tends to just test the specific of each drive and not how the OS routes the data. 

What do you suggest then?

[mr moose]

Just now, williamcll said:

What do you suggest then?

I actually don't know of anything that would test it.   Maybe I have missed something in the way it works, but with today's hardware even saving a >1GB file to a USB can be done in the background without noticeable impact on system resources.   So measuring if it is being delayed (through caching options) or copied straight would likely only yield a result detectable by measurement.

[Doobeedoo]

Really I can't even remember when I experienced data corruption after transfer and unplugging the USB drive without clicking to safely remove. I'd wait a tiny bit or after LED stopped flickering. Still for important stuff or larger transfers I'd do the so called safe removal. 

[jiyeon]

3 hours ago, Doobeedoo said:

Really I can't even remember when I experienced data corruption after transfer and unplugging the USB drive without clicking to safely remove. I'd wait a tiny bit or after LED stopped flickering. Still for important stuff or larger transfers I'd do the so called safe removal. 

Forget not remembering, I've proper never seen data corruption after straight unplugging a USB drive without clicking the eject button. Although to be fair, I only ever keep a few hundred megabytes on my USB drives so nothing crucial.

[Stefan Payne]

On 4/6/2019 at 11:48 PM, GoodBytes said:

The default has always been "quick removal" for USB drives.

That's what I also thought and experienced.

Everything else doesn't make too much sense for external USB Drives in general.

To prevent as much Dataloss as possible, you have to work without much caching, if any.

Especially on Flash Memory it makes no sense.

On Harddrives it might make sense for sorting the stuff and reading/writing it in order instead of randomly.