Are there any rules stating you can not share your serial numbers on MacRumors?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Are there any rules stating you can not share your serial numbers on MacRumors?
No, not at all. It was meant as friendly advice, and has nothing to do with rules. If you want to freely give it away to Hackintosh clone makers, it’s your choice.
UPD: Slow write speed corrected by SSD recondition operation: I destroyed the raid-0, 5 times overwritten all SSD`s with zeros, re-created raid-0. Everything became ok.
One shouldn't do this - ever. It doesn't clean the SSD, it just makes it 100% full of zeroes -- and causes needless wear to the SSD.
If you created filesystems on the SSDs, and created files full of zeroes, and deleted those files with TRIM enabled - then you would have TRIM'd the files when you deleted them. That would give better performance at the beginning - but unless your RAID setup supports TRIM it will gradually slow down over time as the free space becomes dirty.
If your RAID-0 setup doesn't support TRIM, the way to improve write performance over time is to:
- From a standalone boot environment, boot a utility that can perform a "secure erase" on the SSD. (This is very fast - it doesn't overwrite the drive, it just sets all the flash to "unused").
- Create the RAID-0 volume
- When you create the partition/filesystem on the volume, leave 20% to 25% unpartitioned (search for "SSD over-provisioning")
Last edited:
Some SM951 firmware versions have problems with trimforce. Since it's almost impossible to update the firmware with SM951, you have to workaround this. One thing that works is to boot single user mode and run e2fsck -fy every time the write speeds become slower.
Your problem with write speeds have nothing to do with microcodes. You were looking on the wrong place.
That assumes that the RAID driver/controller supports TRIM. Few of them do.
And by definition TRIM can't be used on RAID-5 - over-provisioning is your only option.
@jsts have a Amfeltec Squid 2.0 with 4 SM951, AppleRAID supports TRIM and he can even run e2fsck blade by blade.
Help me understand this.
The metadata is striped across all of the member drives - no single blade has a complete view of the filesystem, just scattered fragments of meta-data and file data. (...and there's no reason to expect that if the meta-data for a file is on one blade, that the file-data will be on the same blade. In fact, for all but small files the data would be scattered across multiple drives. "Small" is typically in the 64 KiB to 256 KiB range.)
Are you saying the fsck will look at the complete filesystem across all blades, and issue TRIM commands for the free fragments on a particular blade?
Last edited:
After the FS sets that a particular sector is ready for deletion, the firmware of the blade have to automatically do the TRIM operation, cleaning the cells, this don’t change if it’s part of a AppleRAID volume or not.
The problem is that some SM951 firmware versions don’t automatically do this when TRIM is enabled and you have to kickstart it via e2fsck.
My question is about "blade by blade".
I interpret that to mean that you can say "TRIM free space on the second blade only of this four blade RAID-0 volume".
When I had 2 SM951-AHCI 256GB into a stripe AppleRAID, the easiest way to restore write velocity was to run ef2ck -fy into all blades of the AppleRAID.
Early on, I started doing the e2fsck into the AppleRAID volume, but sometimes this didn't restore the write speed, so I started doing it directly, into each blade.
TRIM implementation of some versions of SM951 is just borked. I'll never buy an OEM blade that can't have easy firmware updates again.
My mac pro 2010 have 3 ssd liteon 240 3dnand raid card apple gtx680 rom apple and have installed Mojave Raid strip 0 speed 750 MB/s trim active in APFS full work.
Install mojave in other HD or SSD, boot the Mojave by usb and make the Raid manual by terminal, reboot trimforce enable, reboot to other HD or SSD and clone all CCC to Raid strip 0 APFS and full work...
And you have to do it again when Apple releases an update since updates don't work…
Yes need to make all from 0, the installer Apple not recognize Raid for install Mojave or update.
The problem is the installer not the OS.
Good Morning,
I built a hackintosh system with a Z390 in i9 9900k.
It had two SSDs in Raid 0, had a performance of 1000mb / s and put 3 more sds, totaling 5 ssd in Raid 0.
I use the softraid, apparently everything is working normally, but the transfer rate that I thought was going to exceed 2500mb / s is 1500mb / s.
What could be causing this loss?
From what I've been reading this z390 could support up to 3500mb / s.
This raid 0 that I mounted is not the starting disc, for starting disc I have a 970 M2 Pro 512gb.
Any suggestion ?
I built a hackintosh system with a Z390 in i9 9900k.
It had two SSDs in Raid 0, had a performance of 1000mb / s and put 3 more sds, totaling 5 ssd in Raid 0.
I use the softraid, apparently everything is working normally, but the transfer rate that I thought was going to exceed 2500mb / s is 1500mb / s.
What could be causing this loss?
From what I've been reading this z390 could support up to 3500mb / s.
This raid 0 that I mounted is not the starting disc, for starting disc I have a 970 M2 Pro 512gb.
Any suggestion ?
How to target an individual blade with the e2fsck command? Needs the right parameter I assume.
I can't run e2fsck in single user mode as I do not have an EFI enabled GPU. Can I boot from another HD and then run e2fsck on the SM951 blades from the fully booted Mojave system?
Only really works for the root drive, when formatted as HFS+ and via single user mode.
Thanks.
Any recommendations how to recondition an APFS boot drive on a SM951 blade that has degraded write performance?
The only app that claims to fix this is DigLloyd's disk tester but it has not been updated since the introduction of APFS.
Would just erasing the APFS volume and recreate it help?