• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

SSDs in RAID - disk volume size

Trajan

Member
I just got a pair of the Samsung EVOs and am setting them up to run in RAID 0. I understand that for optimal performance, I should leave 15-25% of the space unformatted, but I don't 100% understand why and therefore I'm not sure the best way to do it, or if the method even matters:

* When I connect the drives and set them to RAID in the BIOS, is this the point that I should select the 15-25% reduced volume size?

or

* Should I select maximum volume size in the RAID step, and then choose a smaller volume size when I'm actually formatting the drive and creating a partition in Windows?

Thanks for any advice. I am a little bit worried, if I do this the wrong way, that the SSD controller/driver somehow won't recognize that it has the extra space and I'll end up shortening my usable volume without getting any of the benefits!
 
From what I understand, the reason that the space is left is to leave adequate room for wear leveling. To keep certain cells from being written to over and over again (wearing them out) while other cells lie dormant (for example, the OS system files may sit for years in the same spot), the drive will periodically move stuff around (without you or the OS knowing) and remap the cells so that it's transparent to the host operating system. Once the drive starts reaching near capacity, it has little room to move stuff around. By over provisioning, you are giving it unused space (permanently since it's not partitioned) to used for the wear leveling routine.

As how to do it, simply set your partition size in Windows (or whatever you partition it in) to 15% to 25% smaller than the total available.

Keep in mind that most drives come with some space already provisioned for this (i.e. your 250GB drive may actually be 256GB with 6GB given to the provisioning for wear leveling).

In my opinion, I'm not a fan unless you're going to be filling up the drive at which point, I would simply get something bigger. Oh, and the 15 to 25% sounds quite high. That's a lot of drive to by purchasing to just give away to the wear leveling routine.
 
Last edited:
Keep in mind that most drives come with some space already provisioned for this (i.e. your 250GB drive may actually be 256GB with 6GB given to the provisioning for wear leveling).
It would be 256GiB of flash, with about 23.17GiB set aside, of which around 14.17GiB should be usable for wear-leveling, and around 9GiB being used for the write buffer, I'm pretty sure.

IMO, if you're worried about endurance, you just should spend just a wee bit more per SSD, and go with an MLC drive. It likely won't be a problem either way, today, but that would take care of it.

Since IRST support TRIM in RAID 0, now, though, just leave plenty of free space on the file system most of the time, and you should be fine (so long as you have a 7- or 8-series chipset, or are willing to hack IRST for the 6-series).
 
Last edited:
It would be 256GiB of flash, with about 23.17GiB set aside, of which around 14.17GiB should be usable for wear-leveling, and around 9GiB being used for the write buffer, I'm pretty sure.

Is the 250GB number (size of the drive as listed) in GB or GiB? I'm assuming you are correct in that the drive is listed as 250GB (250,000,000,000 useable bytes when it really contains 256GiB of storage space).

(note: I always forget about the size sometimes is in 1,000,000,000 bytes per GB and sometimes it's 1,073,741,824 bytes per GB)

Also, someone remind me what GiB stands for? 😛

(Edit: GibiByte)

Another option is to buy a SLC drive for endurance and speed, but at a cost.
 
Last edited:
Just partition the entire thing and try to leave about 10-15% unused. Are you anticipating entirely filling the drive? If so spend a bit more and either get a larger one or another one to spread out the data.

You shouldn't have any problems regardless. You will want about 5% minimum free just for the OS if on the boot drive.

Also why are you putting them in RAID 0? Are the base speeds not fast enough? Those drives are ridiculously fast from the factory. Just curious if you have a specific need for that speed.
 
Just partition the entire thing and try to leave about 10-15% unused. Are you anticipating entirely filling the drive? If so spend a bit more and either get a larger one or another one to spread out the data.

You shouldn't have any problems regardless. You will want about 5% minimum free just for the OS if on the boot drive.

Also why are you putting them in RAID 0? Are the base speeds not fast enough? Those drives are ridiculously fast from the factory. Just curious if you have a specific need for that speed.

This is good advice ^^
just partition them 100% and let the controller do it's job.
 
Another option is to buy a SLC drive for endurance and speed, but at a cost.
No need, unless you know you'll be really really hard on it, at which point RAID 0 would not be your option of choice, anyway. Even if you are, a little OP will do the trick for any 1st/2nd-party MLC, these days (Samsung, Crucial, Intel, Toshiba, and probably Sandisk). Some makers, like Samsung, have even transitioned their enterprise drives off of SLC.

TLC is generally good enough, and the SLC-mode caching should keep really bad random cases at bay (even considering the added space used for pseudo-SLC, the SLC-mode space should be able to handle several times the writes of the TLC space, assuming Samsung's are similar to Toshiba's--I haven't even tried to check out Samsung's specs), but MLC is good enough to just not worry about, as long as you leave some space, even if you do get stuck with higher than average WAs. Yes, that's one sentence 🙂.
 
Back
Top