Overprovisioning on a Multi Boot SSD

Does each OS partition need it's own or will one do for all?

I also want to mess around with Linux. What caveats are there to that since it is not NTFS?

Thanks,

FT

 

One 'unallocated' space will be enough.

With Linux and other O/S's (which are not Win7/Win8.1) you may have an issue with TRIM, but the basic GC built into the SSD's firmware should be enough to minimize any issues (along with the 'unallocated' partition - which I recommend to be ~30%, of course).


Make sure you're not running an EVO or the original TLC Samsung 840 in that setup either...


See:
http://forum.notebookreview.com/sol...ds-old-files-evo-fix-september-19-2014-a.html

 

Part B... Why 30%? That's a chunk.

And is it truly an unallocated space? I thought it was NTFS with Windows, but don't have any idea really.

 

http://forum.notebookreview.com/linux-compatibility-software/753861-trim-ssd-ubuntu.html This may clear up any questions about trim, If you are going with a different distro you could post in the Linux forum here or in a forum specifically for the distro

 

30% is tiller's personal recommendation because that level has worked best with his systems and workflow. There are varying recommendations for the percentage of the SSD to leave empty for overprovisioning, depending on the source. For instance, Samsung recommends 10% overprovisioning on their drives, and Anandtech has found performance improvements from 25% overprovisioning (compared to a full drive). Feel free to experiment for yourself.

Overprovisioned space is unformatted and unavailable for storage. It shows up as unformatted unused capacity when using partition managers, but does not show up in system drives or folders. For instance, in a multi-boot scenario, a 256GB SSD could have 100MB for MBR or GPT, perhaps the same for GRUB, 100GB for Windows, 4GB for swap, 96GB for Linux, and 38GB (approx. 16%) unallocated for overprovisioning. This is just an example, and all these partition sizes could be adjusted. But that gives you an idea of what it could look like.

 

All the partitions share the same logical block address pool (actually the SSD itself does not understand the concept of "partitioning"), so you only need to worry about the total amount.

Do keep in mind that you need to TRIM the OP'ed block range or secure erase whole SSD first, or the blocks you want to free will not be marked free inside the SSD and you'll go back to standard 7% built-in OP only.

 

Some drives may already have some over-provisioned space, so take that into account in your calculations. Also, if you think you'll need the space go for 10%, if you think you don't, you may as well go for 25% and you can always extend the volume afterwards if needed.

 

Due to the difference between 1000^3 and 1024^3, SSDs usually have at least 7% built-in OP to play with. Most consumer drives just use this percentage. Industrial drives would go much higher.

OP by leaving empty space in partitioning is not the same as built-in OP in firmware. With today's controllers they should be similar, but the same can't be said for old drives.

 

Are you sure about this?

Let's say that you have a 256 GB drive, the drive will be 256 000 000 Bytes or 256x1000 ³ while Windows will report the amount in GB to be 256x1000 ³/1024 ³. The end result being that Windows reports less capacity, so there is a loss of 7% space, not a gain of 7% space for OP. It was my understanding that NAND chips used to make SSDs weren't like DRAM chips and that they weren't made in capacities that are in multiples of 1024 ³. If you can prove me wrong, I'll be pretty happy about it though.

 

I can't find any clear documentation atm (will try to Google) but as far as I can remember, no discrete storage device is made with capacity being power of 10. Actually if you do cut the address space at a base 10 boundary on each die you would have nasty gaps in your base 2 address space when combined.

The concept of "I brought a 256GB drive, I should have exactly 256 000 000 Bytes of physical storage" only applies to magnetic HDDs or tapes, which writes discrete data on an analogy physical media and therefore easy to scale. Discrete NAND doesn't scale like that. The base 10 counting of space is only used for labeling or P​​​R. The entire storage pipeline is strictly binary. (No X*1000 cluster size in file systems to worry about.) In the case of consumer SSDs, the controller simply cuts 7% on the FTL table so you can't request to occupy everything.


In short, when you buy a HDD you lose 7% of physical capacity compared to 1024 ³ x advertised GB count. When you buy a consumer SSD you get exactly 1024 ³ x advertised GB count of physical capacity, however logically you still lose 7% due to the FTL restricting your access.


BTW, TLC SSDs have more physical capacity than advertised due to the fast failing NAND needing more backup. At least with Samsung drives you can adjust the accessible FTL table length to make built-in OP even more aggressive if necessary.

 

Counting the 'included' or 'implied' OP'ing of an SSD has no real world benefits from my experience. At 25% OP level of actual capacity, I see the performance gains and consistency improvements from any SSD I've used become 'real'. Another 5% (my recommended 30% total) effectively puts that consistency in the high 99.9999% range for most workflows. In my desktop workstations, I typically use 50% OP'ing of actual capacity of that specific SSD and have found that 65% OP'ing was also beneficial for some older setups.

This is not something you can cheat at (like getting extra change back from McD's than you should...) - just OP over and above whatever the SSD ships with - the benefits are real and lasting.

 

Thanks for using that term, In all my Windows 7 (as primary) OS SSD's I have at one time created ~20 GB partitions for Linux. At this point I am not running Linux at all so I have ~20 GB unallocated on my systems. the 20 GB is contiguous with the OS partitions (I use only 1 partition at this point per physical drive). Is there a difference AFA performance between unallocated vrs free space?
Thanks for any thoughts and HAPPY Holidays to all
NEVERMIND, I should have searched first TilleroftheEarth has already schooled us on this issue
http://forum.notebookreview.com/sol...707276-unallocated-space-ssd.html#post9068560

Thanks Mr. Koala + 1 as soon as I "spread it around"

 

No. Your SSD is a block device, it doesn't understand any file system.