Question about how robust is a SSD RAID 1 on an AMD motherboard ?

[Jandor]

I used the raid system implemented for the chipset on several systems with SSD raid1.
I found sata to be more convenient and eating less PCIe lines than Nvme.

I always used one SSD Samsung 860 (TLC) and one SSD Crucial MX 500 (in case of some BIOS or chipset failure, it won't be the same reason at the same time).
I also left around 15% of the space free at the end of each SSD before creating the RAID 1 enclosure (software but chipset supported by AMD).
The raid 1 system is supporting the OS (aka Windows 7).
Both SSD are treated in Raid mode at the BIOS level of the motherboard.

Question 1 : does someone know if the free space on top of every SSD is used for wear-leveling ? I suppose so but I am not sure. best would be to be treated as over-provisioning, but some people say that free space (in AHCI mode) is treated like over-provisioning on modern SSDs avec wear-leveling is so good that trim isn't necessary... But some people say that treating the SSD in RAID mode annihilated any own feature of the chipset of any SSD and I should have used AHCI. Those people say that in Raid mode the OS/chipset writes on the SSD as if it was a simple flash drive and could rewrite on the same cell without any management. But AHCI and Raid are incompatible (the raid software puts back AHCI mode in Raid mode when configuring). I believe Trim doesn't work even if some Raid modes are able to activate the feature (mostly RAID 0 and supposedly some recent Intel chipset for RAID 1 with most recent driver).
Question 2 : Did any one experienced wear levelling by using RAID 1 or 10 or 5 on a SSD system, or even some other kind of failure ? How did it happen and how did he managed to solve the problem (without reinstalling the system) ?

 

[bluestang]

You're using 2 different SSDs in RAID1...that is highly not recommended. BIOS and chipset failure?...makes no sense in why the 2 different drives. I think you mean the drive Firmware most likely.

Not sure if AMD ever fixed the fact that they didn't support TRIM in RAID mode. But the drives themselves should ha e garbage collection built into their FW so the drives can do it themselves when left idle for a while.

Yes, free space left if not partitioning the whole drive is Over -Provisioning. The drives FW should take care of using that appropriately for wear leveling and such, but unless you are doing a high amount of writes to the drive or filling the drive almost completely then there is not much reason to.

 

[daglesj]

I would have thought different spec SSDs in RAID1 would be less critical than say in RAID0. Essentially as its just writing the same old data to two different drives. It's not like two identical HDDs or SSDs performed exactly the same either.

As long as they are the same size. Sure for neatness its nice to have the same...

 

[ZodaEX]

If both drives did fail at the same time, would you be able to restore all of your data from a backup somewhere?

 

[Jandor]

If both drives did fail at the same time, would you be able to restore all of your data from a backup somewhere?

I'm doing backups, even system backups (Windows 7 way). Yes. Not sure how to restore on a Raid system if both fail. But if one fails it should be okay.
I was taking about BIOS and assuming is was the SSD BIOS or firmware.
SSDs have mostly those kind of failure is they are not very much used. So having 2 brands, 2 models (same size of course) is better than having the same, because Raid 1 is made for safety.
Now hard drives quite never fail because of a bug in the firmware, but did have mechanical failure that could happen at any time. So having 2 hard drives of the same kind was the better option to avoid loosing too much performance that one doesn't care on SSDs because unnoticeable especially over SATA.

 

[lopoetve]

I was taking about BIOS and assuming is was the SSD BIOS or firmware.
SSDs have mostly those kind of failure is they are not very much used. So having 2 brands, 2 models (same size of course) is better than having the same, because Raid 1 is made for safety.
Now hard drives quite never fail because of a bug in the firmware, but did have mechanical failure that could happen at any time. So having 2 hard drives of the same kind was the better option to avoid loosing too much performance that one doesn't care on SSDs because unnoticeable especially over SATA.

This... is not really accurate. None of these are really issues anymore on the SSD side. You're overthinking it.

Also - NVMe drives don't "eat" PCIE lanes. The lanes to those are dedicated - you can't "get them back" by not using them, unless it's one of the later M.2 slots that shares with SATA ports/etc.