APFS (30sec) TO HFS+ (9sec) boot times

[fontanero]

Hi, I have a Samsung 960 Evo NvME. Everything works fine in my hack but boot times were long (30 sec), I tried to change the APFS format to HFS + and the boot is less than half time. (9 sec)
I do not know why ... but in my case the format change has been important.

 

[P1LGRIM]

Hi, I have a Samsung 960 Evo NvME. Everything works fine in my hack but boot times were long (30 sec), I tried to change the APFS format to HFS + and the boot is less than half time. (9 sec)
I do not know why ... but in my case the format change has been important.

If you have enabled trim then try with it disabled.
APFS and trim are not the best of friends.

 

[fontanero]

Hi, i don't use trim with APFS, always disabled.
Now is enabled with HFS+

 

[vulgo]

Hi, i don't use trim with APFS, always disabled.

NVME, TRIM is always enabled.

 

[fontanero]

When I enable TRIM on 960NVME, speed it's a little slow.
I try with Trim, without TRIM on APFS and boot is slow compare with HFS+ (30sec to 9sec)
Enable TRIM on HFS+ without difference

 

[vulgo]

I try with Trim, without TRIM on APFS and boot is slow compare with HFS+ (30sec to 9sec)

If your SSD is connected by AHCI then this is common, booting from third-party SSD is slow due to a 'find unused blocks' operation before TRIM during mount. It seems to get worse over time. With TRIM disabled the check/TRIM before mount does not take place.

 

[Thomazella]

If your SSD is connected by AHCI then this is common, booting from third-party SSD is slow due to a 'find unused blocks' operation before TRIM during mount. It seems to get worse over time. With TRIM disabled the check/TRIM before mount does not take place.

do you know if XHCI suffers from the same issue?

 

[vulgo]

do you know if XHCI suffers from the same issue?

No idea. If the protocol over USB and macOS driver end up doing the same thing it may well be slow, possibly it is the implementation on the device itself that is slowing things down. Some NVME users report a similar delay for example, and TRIM cannot be disabled on those drives. Quick search results suggests TRIM is not supported on USB storage so any slow down may be due to something else.

 

[asheenlevrai]

If your SSD is connected by AHCI then this is common, booting from third-party SSD is slow due to a 'find unused blocks' operation before TRIM during mount. It seems to get worse over time. With TRIM disabled the check/TRIM before mount does not take place.

Is it OK or not to use 3rd party sATA AHCI SSDs formated with APFS and having TRIM disabled? Or is there any drawbacks? Is it going to lead to issues in the long term? Shorter life span? Potential loss of data? Decreased performances?

Tx
-a-

 

[Celeri]

Here is a quick feedback on my hackintosh.

I installed a 2TB Intel 660p about one year ago, wanting to leverage NVMe performance over SATA. The performance was alerady not-so-great at the beginning, but I even noticed a progressive drop in global performance over time, especially on things like boot (TRIM operation taking more and more time), displaying custom icons in Finder, or loading screens in games.

After reading some articles here about strange problems with APFS on NVMe SSDs a few weeks ago, I CCC'ed my system on an external disk, wiped the 660p in HFS+ and restored my data on it. And since then, I just found back a whole new hackintosh : 20s of boot time, snappy Finder and games... and benchmarks that talk for themselves !

AmorphousDiskMark test on APFS :

Same test in HFS+ :

I sure hope this APFS problem will be solved by the time I will have to uptate to Catalina... Good thing this time has not come yet, Mojave being very solid on this build, and still having a few 32-bit apps.