SSD hall of fame needs new entries
vit9696 opened this issue · 108 comments
Guide Anti-Hackintosh Buyers Guide (link)
After our discovery of a severe bug in the TRIM implementation of practically all Samsung SSDs we spent time investigating which SSDs are affected by all kinds of issues, and so far came with several names worth mentioning.
Working with TRIM broken (can be used with TRIM disabled, at slower boot times, or as a data storage):
- Samsung 950 Pro
- Samsung 960 Evo/Pro
- Samsung 970 Evo/Pro
Working fine with TRIM:
- Western Digital Blue SN550
- Western Digital Black SN700
- Western Digital Black SN720
- Western Digital Black SN750 (aka SanDisk Extreme PRO)
- Western Digital Black SN850
- Intel 760p (including OEM models, e.g. SSDPEMKF512G8)
- Crucial P1 1TB NVME (SM2263EN) (need more tests)
Working fine with TRIM (SATA):
- SATA PLEXTOR M5Pro
- SATA Samsung 850 PRO (need more tests)
- SATA Samsung 870 EVO (need more tests)
Working fine with TRIM (Unbranded SSDs):
- KingDian S280
- Kingchuxing 512GB
Incompatible with IONVMeFamily (die under heavy load):
- GIGABYTE 512 GB M.2 PCIe SSD (e.g. GP-GSM2NE8512GNTD) (need more tests)
- ADATA Swordfish 2 TB M.2-2280
- SK Hynix HFS001TD9TNG-L5B0B
- SK Hynix P31
- Samsung PM981 models
- Micron 2200V MTFDHBA512TCK
- Asgard AN3+ (STAR1000P)
- Netac NVME SSD 480
There are several very good comparison charts containing various SSDs and their controllers: one, two. Since SSD compatibility usually is controller-based, picking up an SSD with a known to be compatible controller has a high compatibility chance as well.
In addition to that, I would like more people to run tests on their platforms to determine whether their SSD TRIM implementation is broken or not.
Preconditions:
- Tested SSD is the boot SSD (i.e. macOS is installed on it).
- Tested SSD is APFS-formatted.
- Tested SSD must be actively used with macOS installation being at least a month old.
- OpenCore 0.6.7 or newer is used.
The idea is to measure the boot time either between motherboard logo and macOS login screen or between Apple logo and macOS login screen. The former is strongly recommended due to possible GOP issues. All the measurements needs to be done with different SetApfsTrimTimeout values set in OpenCore. Each measurement must be made at least 2 or 3 three times to ensure no sporadic results. The values to test are as follows:
0(means TRIM is disabled)-1(standard timeout, equals roughly 10 seconds, means TRIM is enabled and runs up 10 seconds during boot)4294967295(maximum timeout, TRIM is enabled and runs as long as needed)
- An SSD with a decent TRIM implementation should behave as follows. Value 1 will be negligibly the fastest. Value 2 might be slightly slower than 2 (way less than by 10 seconds, maybe 2-5). Value 3 should be equal to value 2.
- An SSD with a broken TRIM implementation, like Samsung, will have value 2 slower than value 1 by 10 seconds, and value 3 will be noticeably slower than value 2, usually by 30-60 seconds.
Please write include the following information in the report:
- SSD type (NVMe or SATA), model, and storage size
- Motherboard and CPU used
- macOS version used
- Test results grouped in 3 sections (a total of 6 or 9 values).
Note: If you use FileVault 2, you can use sudo fdesetup authrestart command to skip UEFI login. AuthRestart must of course be enabled. I would also advise to disable the OpenCore Picker (ShowPicker = NO).
Gigabyte Z490i Aorus Ultra + 10900k
11.3 Beta (20E5229a)
Samsung 970 Evo NVME:
- 18, 17, 18.
- 29, 30, 30.
- 97, 85, 86
Crucial BX500 120GB SATA 2.5-inch
GA-H61M-S2PV + i5-2500K
11.3.1 (20E241)
- 31 30 30
- 39 44 40
- 46 42 41
(sudo trimforce enable)
From the motherboard logo to macOS system, pressing Enter immediately at OpenCore menu.
WD Blue SN550 1TB NVMe
Z370 AORUS Gaming 3 + i5-9600K
11.3.1 (20E241)
- 20 20 20
- 34 33 32
- 36 37 37
From Apple logo to macOS system.
Crucial MX500 500GB SATA 2.5-inch
MacBookPro9,2 + i5-3210M
11.3.1 (20E241)
- 39 37 39
- 41 42 45
- 45 46 46
(sudo trimforce enable)
From Apple logo to macOS system.
Samsung SM951 256GB NVME (MZVPV256HDGL-00000)
Gigabyte GA-Z97X-UD5H, +i7-4790K, iMac 15,1
10.15.7 (19H1030)
- 22.90, 22.82, 23.29
- 27.46, 26.77, 26.43
- 27.10, 26.62, 26.33
From gigabyte splash screen to macOS Login screen (used stopwatch on iPhone to time each run)
Pressing Enter immediately at OC GUI.
Samsung 830 256GB SATA SSD
Asus P9X79WS + Xeon E5-2667v2 + 64GB ECC + MacPro6,1 SMBIOS
macOS 11.3.1 + OpenCore 0.6.9
Time in seconds from pressing enter in OC-Picker to login screen:
sudo trimforce disable:
- 23, 21, 22
sudo trimforce enable (SetApfsTrimTimeout={999, -1, 4294967295}):
- 32, 31, 32
- 42, 42, 41
- 45, 44, 44
I expected 'trimforce disable' and 'trimforce enable' with 'SetApfsTrimTimeout:999' to be approximately equal. Why isn't that the case? 🤔
HideAuxiliary Enabled, ShowPicker, and Timeout Disabled. (I always use it this way.)
ADATA XPG SX8200 Pro 1TB NVMe
ASUS TUF Z390-Plus Gaming + Intel 9900K
11.4 Beta (20F5055c) iMac 19,1
- (43, 45, 46)
- (46, 47, 48)
- (53, 54, 55)
It is always slow in any case. There is only one OS in this Build.
Samsung 870 EVO 1TB SATA
Gigabyte Z390 Aorus Pro / i7-9700K
11.3.1 / iMac 19,1
Times taken between Apple Logo and Login Screen
- 15, 15, 15
- 18, 19, 19
- 18, 18, 18
Corsair MP600 1TB NVMe
Asus ROG Strix X570-I / Ryzen 9 3950X
11.4 Beta / MacPro7,1
- 23 24 24
- 28 27 27
- 27 27 28
I don't know if you'll ever consider the following, but apparently many models of Phison E12 (like Sabrent Rocket NVMe 3.0 TLC 64 layers NAND) have a broken trim support and are going to failure of the disk.
In less than 1 year, I wrote just 5.2 TB of data for my 512GB model and the drive estimated life was 94%.
According to Sabrent TBW warranty, the 500GB 3.0 TLC model should have 800TBW of life which is a joke.
Please avoid buying NVMes with Phison E12 controller.
I opted for a WD Black SN750 and it works flawlessly both with and without TRIM ^^
Crucial P2 CT500P2SSD8
Lenovo L390 Yoga with i5-8265U / Big Sur 11.4 / MacBookPro 15,2
- 22 22,3 21,5
- 30,9 31 31,3
- 32,1 31,9 32,20
About 10 seconds more with trim enabled
DELL XPS 7590 FHD | i7 9750-H | HP SSD EX920 512GB
macOS BigSur 11.4 | MacBookPro16,1 | OpenCore 0.6.9
- 24 25
- 27 27
- 27 27
Testing done from POST logo to macOS login screen
Intel 660p
SSD:
Model: INTEL SSDPEKNW010T8
type: NVMe
size: 1TB
motherboard: LattePanda Alpha 800s
CPU: m3-8100Y
macOS: macOS 11.4 (20F71)| SetApfsTrimTimeout | time (s) |
|---|---|
| 999 | 44.84 |
| 999 | 49.88 |
| 999 | 47.36 |
| -1 | 53.05 |
| -1 | 52.27 |
| 4294967295 | 56.26 |
| 4294967295 | 56.47 |
Samsung 860 EVO
SSD:
Model: Samsung SSD 860 EVO 500GB
type: SATA
size: 500GB
motherboard: GA–Z77–HD4
CPU: i7–3770K
macOS: macOS 11.4 (20F71)| SetApfsTrimTimeout | time (s) |
|---|---|
| 999 | 35.62 |
| 999 | 35.38 |
| -1 | 39.75 |
| -1 | 39.81 |
| 4294967295 | 39.61 |
| 4294967295 | 39.63 |
Samsung 970 EVO Plus
SSD:
Model: Samsung SSD 970 EVO Plus 1TB
type: NVMe
size: 1TB
motherboard: GA–X299–UD4
CPU: i9 7900X
macOS: macOS 10.15.7 (19H1217)| SetApfsTrimTimeout | time (s) |
|---|---|
| 999 | 41.30 |
| 999 | 42.39 |
| -1 | 51.43 |
| -1 | 51.62 |
| 4294967295 | 1:46.87 |
| 4294967295 | 1:45.62 |
Samsung NVME 970 Pro 512GB
Big Sur 11.4
Z490 Vision D + 5700 XT
OC 0.7.1 iMac20,2
NVMEFixup.kext enabled
VALUE | POST>OC | OC>OSX | TOTAL BOOT
999 | 11 | 25 | 36
-1 | 11 | 36 | 47
4294967295 | 11 | 90 | 100
Confirms your findings about Samsung
Hi! I’m using a Samsung 970 Evo Plus 500GB, but I’m on Clover. Is there anyway to test or implement this without OC? Thanks!
Samsung SSD 960 PRO 512GB
SSD:
Model: Samsung SSD 960 PRO 512GB
type: NVMe
size: 512GB
motherboard: GA–X299–UD4
CPU: i9 7900X
macOS: macOS 11.4 (20F71)| SetApfsTrimTimeout | time (s) |
|---|---|
| 999 | 34.32 |
| 999 | 33.34 |
| -1 | 42.25 |
| -1 | 42.83 |
| 4294967295 | 1:27.75 |
| 4294967295 | 1:28.21 |
c.f. #192 (comment) for similar reports.
Western Digital Black SN850 500gb (in PCIe Gen3 Environment):
motherboard: Asus x299 TUF Mark1
CPU: i7 7800X
macOS: macOS 11.5.2 (20G95)
999: 25.57 | 25.10 | 22.80
-1: 23.54 | 25.19 | 24.38
4294967295: 25.39 | 24.65 | 24.74
From Apple logo to macOS system.
Mine is Gigabite 240 GB M.2 PCIe SSD (GP-GSM2NE3256GNTD) and it hangs randomly.
FWIW, I’ve tried Plextor M10P with Innogrit controller, and it does not work with macOS. Read operations cause disk I/O freeze, resulting in loss of the device from the macOS device tree (!) until a system reboot.
Writes, however, work fine.
Windows works fine.
Just wanted to warn about this case.
WD Black SN 750 NVMe 500 GB
Z390 Aorus Elite + i7-9700
Monterey beta 10
MacPro7,1
Since motherboard beep to desktop active.
999: 20 | 22 | 22
-1: 19 | 18 | 18
4294967295: 19 | 22 | 20
WD Black SN 750 NVMe 500 GB Z390 Aorus Elite + i7-9700 Monterey beta 10 MacPro7,1
Since motherboard beep to desktop active.
999: 20 | 22 | 22
-1: 19 | 18 | 18
4294967295: 19 | 22 | 20
Hi - May I ask you what settings you've implemented for your SN750? I can't seem to get away from NVMe kernel panics happening eventually with my SN750 2TB -- thank you!
WD Blue SN550 1TB NVMe Z370 AORUS Gaming 3 + i5-9600K 11.3.1 (20E241)
- 20 20 20
- 34 33 32
- 36 37 37
From Apple logo to macOS system.
In Monterey ‘4294967295’ is 8 seconds faster.
Samsung 970 Pro, Aorus Z390 Pro
No matter what timeout I set I get 3:20+ long boot with Monterey, where before upgrading the boot takes less than one minute. Massively annoying.
Okay, I ran a test per the instructions. Results on my computer at left:
NVMe, Samsung 970 Pro 1 TB
Gigabyte Z390 I AORUS PRO WIFI and Intel i9 9900K CPU
Mac OS Big Sur 11.6.1 (Build 20G224)
SetAPFSTrimTimeout at:
999 -- 22, 22, 21 secs
-1 -- 34, 32, 32 secs
4294967295 -- 59, 57, 58 secs
I decided to check my Samsung 970 PRO's boot times in my "Mini-ITX 3" computer to see if the same TRIM boot problem occurs running High Sierra 10.13.6 using OpenCore 0.7.5. Result is below:
SSD: NVMe, Samsung 970 PRO 1 TB
Mobo: Gigabyte Z370N-WIFI with i7 8700 CPU
MacOS: High Sierra 10.13.6
"4294967295": 22, 18, 20 secs from beginning AMI display to desktop
So the "broken TRIM implementation" on the Samsung 970 PRO is not apparent running under MacOS 10.13.6, at least in "Mini-ITX 3." Bad in MacOS 11.6.1, good in MacOS 10.13.6. FWIW.
Samsung SSD 970 EVO 500GB
motherboard: Asus PRIME Z490M-P
CPU: i7 10700K
macOS 12.0.1 (21A559) (updated from Big Sur)
I can't get any of the values to change the timeout settings, I don't think I need to be reseting NVRAM or anything else, seems that others can just change the value in config.plist. I started taking some manual stopwatch times from the opencore picker to the macos desktop when I realized the log times were not changing. I have included those if they help at all.
Checking the time with
"log show --predicate "processID == 0" --start $(date "+%Y-%m-%d") --debug | grep ", trims took"
999: 74.656852 s | 74.733054 | 72.474862 s (1:54 stopwatch)
-1: 73.301794 s (2:01 stopwatch) | 72.573530 s (1:56 stopwatch)
4294967295: 73.471943 s | 74.791656 s | 73.852858 s (1:55 stopwatch)
@shadegits From what I've gathered the SetApfsTrimTimeout setting does not work on Monterey anymore
From what I've gathered the
SetApfsTrimTimeoutsetting does not work on Monterey anymore
@aluveitie That would explain it then... Thanks!
For me Nitro 5 (i5 7300HQ)
With SSD: Kingston 500NV1
999: 109,73s
-1: 108,44s
4294967295: 108.89s
GA-Z370N WiFi - i8700k
Big Sur 11.6.1
Samsung SSD 970 EVO 250G M.2 NVME
- 14.43 15.10 15.20
- 24.53 25.12 25.10
- 29.35 30.22. 28.80
After our discovery of a severe bug in the TRIM implementation of practically all Samsung SSDs
I would like more people to run tests on their platforms to determine whether their SSD TRIM implementation is broken or not.
I haven't understood in what way the TRIM implementation is broken in a SSD controller if the described behavior is following from the (weird) APFS driver implementation?
The APFS driver explicitly ignores previously unmapped areas and repeatedly trims them on boot.
Is it possible to patch the APFS driver to not ignore unmapped areas but perform TRIM for them on the go?
I can confirm that Samsung NVMe drives are not so friendly with MacOS during boot (controller trim issue). Monterey has the worst boot time.
Just updated Samsung EVO 970 Plus 1TB > Western Digital Black SN750 1TB
Cloned by: Clonezilla (raw disk copy)
Boot time: -30sec
Anyone tested the Samsung 980 (not Pro) 1TB NVMe with Monterey (12.x)?
Anyone tested the Samsung 980 (not Pro) 1TB NVMe with Monterey (12.x)?
I have one . I will let you know tonight or tomorrow.
Gigabyte Z170 gaming K3 / i7 9700K
Samsung 980 1TB
OS: Monterey 12.1 Beta 3
999: 24.07s
-1 : 25.40s
4294967295: 23.87s
The only issue I have are the write/read speeds much lower than on Windows. I am getting 2200MBps for Write and 1900MBps Reads, instead of 3500MBps advertised. On windows these are 3250/2950 respectively.
Thanks for testing, but I already went for the 1 TB WD Black SN750 to replace my troublesome 500GB Samsung 970 Evo.
There are several very good comparison charts containing various SSDs and their controllers: one, two.
@vit9696 The second link was taken down unfortunately, so unless somebody has a backup of recent date (2021) all we've got now is rather outdated snapshot (14 March, 2019) from archive.org here:
Working fine here.
I've got Western Digital Black SN850 500 GB GEN 4. The opening post states: "need more tests".
Can I do additional tests to help out?
Replaced my problematic EVO 970 500 GB, this WD seems to do fine (reduces boot time from minutes to seconds).
Mac OS Monterey 12.0.1
Times (from pushing powerbutton to login sceen):
999: 28 | 28 | 28
-1: 28 | 27 | 27
4294967295: 28 | 27 | 27
system:
Intel Core i7-7700K
Gigabyte GA-Z270N-WIFI
With a stopwatch, so indicative. At least no significant fluctuation.
@Klaas-Kramer please try some older version of macOS (e.g. Catalina or Big Sur) since SetApfsTrimTimeout may not be working, hence you always get around 27-28 secs of boot time regardless of the value you set
Decided to re-test the NVME on Big Sur 11.6
Gigabyte Z170 gaming K3 / i7 9700K
Samsung 980 1TB
OS: Big Sur 11.6
999: 14.70 14.88
-1 : 21.70 21.55
4294967295: 22.40 22.47
The results are quite different from the test I run on Monterey 12.1 Beta 3 ( in the post above) .
Hello. I have seen that there has recently been a firmware update of the NVME 970 EVO Plus to version 4B2QEXM7. 2 days ago I changed my 970 EVO Plus for a WD 550NM, so right now I cannot test if this firmware corrects the TRIM and returns the boot time to normal time. Can anyone test it? Thanks.
@atomizasser I have 970 EVO Plus and just tried to upgrade. However non of tools provided by them allow me to do so. I have tried both the Grub installer and Magician, both refuse to work.
Additionally I found this thread so it seems like I am not the only one that's having issue with it.
EDIT: After a bit more digging it seems that there are more than one revision of 970 EVO Plus with different controller. I don't know if that it why causing this issue.
UPDATE: According to a new reply from the thread:
Hey everyone,
Wanted to give an update with my experience. I emailed Samsung and their support informed me that the new update 4B2QEXM7 only applies to certain 970 Evo Plus drives (there is a S/N whitelist). If the update isn't showing up for you in either Magician or the USB updater then it looks like the update isn't applicable to your drive. Would be helpful if they made this clear on the firmware page.
XPG Gammix S5 256 GB (AGAMMIXS5-256GT-C)
MacOS Monterey 12.1
Stopwatch started at MB Boot logo, no buttons pressed at the picker:
SetApfsTrimTimeout |
Run 1, s | Run 2, s | Run 3, s |
|---|---|---|---|
999 |
30 | 31 | 32 |
-1 |
28 | 29 | 30 |
4294967295 |
30 | 31 | 32 |
The drive was new, but died after several hours of usage, most likely it was a firmware problem. I wasn't able to use it long enough for a proper test, but at least OS can be installed and started, unlike with Samsung PM981.
I have the 970 Evo Plus and although the update from Catalina to Big Sur indeed made the boot process ~20 seconds slower, it was something I eventually got used to.
Yesterday I updated to Monterey and now it takes a whooping 10 minutes to boot. Is this also due to this nvme issue, or could something else be wrong?
@spergotron9999 please use some other OS other than Monterey for your test, your results are too similar to be meaningful about TRIM behaviour of your SSD
Big Sur (11.6.2) is fine for the test
Since I have a Samsung 950 pro. Do you guys think it is causing this problem?? Is this issue related to my issue at all. My boot time seems fine. Atleast for now. Boot is under 30s, but is it causing this side effect??
I have the 970 Evo Plus and although the update from Catalina to Big Sur indeed made the boot process ~20 seconds slower, it was something I eventually got used to.
Yesterday I updated to Monterey and now it takes a whooping 10 minutes to boot. Is this also due to this nvme issue, or could something else be wrong?
Read comment above.
I had same problem Samsung SSD long boot time after updating to Monterey. I have replaced MacOS Samsung to WD.
Samsung EVO 970 Plus 1TB > Western Digital Black SN750 1TB
Cloned by: Clonezilla (raw disk copy)
Boot time: about 30sec (Monterey)
I have the 970 Evo Plus and although the update from Catalina to Big Sur indeed made the boot process ~20 seconds slower, it was something I eventually got used to.
Yesterday I updated to Monterey and now it takes a whooping 10 minutes to boot. Is this also due to this nvme issue, or could something else be wrong?Read comment above. I had same problem Samsung SSD long boot time after updating to Monterey. I have replaced MacOS Samsung to WD. Samsung EVO 970 Plus 1TB > Western Digital Black SN750 1TB Cloned by: Clonezilla (raw disk copy) Boot time: about 30sec (Monterey)
I can confirm that switching to a new Kingston KC2500 made the whole long boot issue go away; I am now booting Monterey in ~30 seconds.
Been using 970 EVO PRO 1TB x 2 (one for macOS, one for Win10) for over a year and after this Monterey I pretty wondered whether I should use 999 or 4294967295 - may I ask a stupid question that if I simply put SetApfsTrimTimeout to something big enough, say 21000000 does it mean pretty much the same thing as 4294967295? The number is big enough for any timeout anyway, or is 4294967295 a magic value for macOS?
The reason why I asked this is because some plist editors, say OpenCore Configurator, once touched the value will change the 4294967295 into -1, which is the 2's-complement of 32-bit signed int.
After upgrading to Monterey, my 970EVO+ 1T old version was having this slow boot issue and I found this thread. I have a 980 Pro 500G and I saw someone OK with it so I replaced my 970 with it. It was good. I did not measure the boot time but is way faster than it was. But things seemed to change after I upgrade the firmware of my 980 Pro to the newest version 4B2QGXA7 using Magician software. Then it is 5 mins boot time after I entered the password for FileVault lock screen..(By the way, I am using 6600XT, it seems also to boot slower than RX580) I suspected it is the trim issue so I tested the trim timeout configs in this post. All the same, 5 mins. Might not be the trim issue? I thought. I gave a try by having a fresh Monterey install on my 660P 2T. It boots so fast with and without FileVault. Then I clone the Monterey system from my 980 Pro to the 660p. And 660P can still do a fast boot. So maybe 980Pro is also having some issues. Not sure if it is the trim issue because the trim timeout tests have no effect on its boot time. I will do a fresh install on 980Pro later. Currently better avoid SAMSUNG SSDs for Monterey.
Adding commentary to this list so that maybe @vit9696 can update it.
The Hynix P31 actually works fine with TRIM. The 2TB works out of the box as it is newer and comes with newer firwmare, The 500GB and 1TB versions need a firmware upgrade.
I discovered also the all my SMI SM2262ENG drives of various brands work but cause a 40s delay at boot. My one SMI SM2262G (Adata SX8200 pro) also works fine but causes a ~15s additional delay during boot. None are as bad as the Samsungs with Monterey.
Test results:
HP Elitedesk 800 G5 i9-9900
Monterey 12.2
Hynix P31 2TB FW 31060C20
From BIOS logo to login screen.
999: 20.5s
-1: 19.8s
4294967295: 20.6s
Likely variance are due to my finger reactions speed..
Was fine after a clean install of Monterey with my 970 Evo, but now takes 66s. I guess it gets longer over time… Hopefully there will be a fix one day. 🙂
Was fine after a clean install of Monterey with my 970 Evo, but now takes 66s. I guess it gets longer over time… Hopefully there will be a fix one day. 🙂
Try this: Latest OC (0.7.9 build) , Kernel/Quirks/ SetApfsTrimTimeout set it to 0. Also Kernel/Block/Strategy set to Disable. See if it solves the issue.
@aniuks33 won't this actually skip the trim process completely? If so, that could seriously damage the drive, no?
I'm thinking of buying Patriot Viper VPN100 2TB (or VPN110 but I think the VPN100 is better) and leave my Evo as a secondary drive. Does anyone have any experience with Patriot Viper VPN*?
@panosru Yes, it will skip the TRIM process. That itself should not have an impact on wear leveling if that’s what you were asking. With TRIM, the os does instruct the SSD to clean up free blocks for later reuse. If TRIM is not performed, the SSD has to do that before writing new data. That will slow down write performance once the SSD is running out of empty blocks.
One workaround you can do is leaving TRIM off, and occasionally boot with TRIM enabled to clean the SSD up to prevent write performance degradation. Long term, a different SSD would probably be the preferable option.
@aluveitie thank you for the clarification, I'm asking because I saw many posts related to SSD's being completely damaged, resulting in data loss... I think I will "play it safe" and I will get a new NVMe to boot from. I haven't decided yet which to buy, VPN100 or 110, I'm leaning on the 100 because even though the 110 is newer, the 100 has better specs.
Tested Samsung X5 1TB External Thunderbolt on a XPS 15 9570 Coffee Lake:
No matter what I've configured the Boot-Time was always around 64 Seconds (boot log shows more than 40 seconds for trim). After I've opened the X5 and replaced the NVMe with a Western Digital Black SN750 2TB the Boot-Time is around 25 Seconds. The speed test shows on MacOS the WD has the same speed during reading (2300), but is a little bit slower during writing (750). But this is hard to compare since the Samsung Speed on writing is not constant and slows down on longer writing processes.
Hey guys, I have a Samsung 970 EVO Plus 500g that used to be my boot drive, after updating my system to Monterey it was booting in more than 110 seconds from bios beep to Login Screen, so I replaced it with an old ADATA XPG SX6000NP 120g now booting in 52 seconds in all three trim settings 999, -1 and 4294967295 from bios beep to login screen so I guess its working well, with trimforce enable and trim support: yes in system report...
but it is too small for my needs and causing the PC to freeze on bios splash when I try to multi-boot Linux when paired with the 970 EVO Plus (windows 11 working fine), so the question is
can someone please state which controllers should be avoided and which is working properly and can be future proof?
I'll be grateful if anyone can confirm the following drives are supported, as most of the recommended working drives in the above list are not available or very expensive where I live:
Crucial P2
HP EX900 or EX900 PRO
Kingston NV1
Kingston A2000
PNY CS1030
Silicon Power P34A60 OR P34A80
Thank you in advance
After updating OpenCore from v0.7.6 to v0.7.9 and setting SetApfsTrimTimeout to 0, my Samsung 970 EVO Plus 500GB (older Version) now boots in ~9 seconds (from Apple logo to macOS login screen). It was more than ~140 seconds before the OpenCore update, no matter what SetApfsTrimTimeout was set to. Using macOS Monterey v12.1. Very happy!
Just remember to occasionally boot with it set to -1 to TRIM the disk.
Just remember to occasionally boot with it set to -1 to TRIM the disk.
no need. SetApfsTrimTimeout set to 0 just effect boot time, the system trim is still on.
@stiffyfoto The disk still supports trim, but by setting the timeout to 0 you prevent macOS from actually running it.
Working fine with TRIM:
- Western Digital Blue SN550
- Western Digital Black SN700
- Western Digital Black SN720
- Western Digital Black SN750 (aka SanDisk Extreme PRO)
- Western Digital Black SN850 (need more tests)
- Intel 760p (including OEM models, e.g. SSDPEMKF512G8)
- Crucial P1 1TB NVME (SM2263EN) (need more tests)
Hi, I would like to know if there is any news?
I have problems with the Samsung SSD 970 EVO and I would like to solve them definitively by switching to Western Digital Black better SN750, SN750 SE or SN850?
They are the same price, which one would be the most compatible?
My samsung now boots in 3 minutes and gives me some problems.
I also have a sabrent that boots up fast but I think it has some problems with waking up.
Thanks
OK tests completed. Timed from Gigabyte splash screen to macOS Login screen, using iPhone stopwatch.
System as follows:
- Samsung SM951 2560GB NVMe (MZVPV256HDGL)
- Gigabyte Z97X-UD5H
- i7-4790K (Stock)
- Big Sur 11.6.4
999 - 41.44, 41.41, 42.11
-1 - 45.75, 45.70, 46.02
4294967295 - 45.41, 45.72, 46.25
Assume Trim is working fine on this old NVME drive as difference between the three tests only 4-5 seconds. With test 2 (-1) equal to time for boot for test 3 (4294967295).
28/04/2022 Just to clarify, I have owned this NVMe and another identical drive since 2015, so losing just 5% of the life is more than acceptable.
I don't know if you'll ever consider the following, but apparently many models of Phison E12 (like Sabrent Rocket NVMe 3.0 TLC 64 layers NAND) have a broken trim support and are going to failure of the disk. In less than 1 year, I wrote just 5.2 TB of data for my 512GB model and the drive estimated life was 94%. According to Sabrent TBW warranty, the 500GB 3.0 TLC model should have 800TBW of life which is a joke.
Please avoid buying NVMes with Phison E12 controller. I opted for a WD Black SN750 and it works flawlessly both with and without TRIM ^^
I have Phison E13 controller. And must be avoided also. Experimenting same brand of NVMe (Phison E13 and Maxio Chipset). Maxio work better than Phison E13. After that, i swap those NVMe to Kingston A2000. Work near native and flawless.
Samsung 860 EVO 500GB SATA
Z490-A Prime + i7-10700K
macOS 11.6.5 Big Sur
Skipped UEFI login and disabled OC Picker. Timed from motherboard logo to macOS login screen.
TRIM appears to be working correctly: -1 takes about 5 seconds longer than 999, and -1 and 4294967295 take roughly the same amount of time.
999: 22.89, 22.73, 22.43-1: 27.79, 28.25, 27.564294967295: 27.46, 26.86, 27.71
Hey @vit9696 , can we get an update on the supported NVMe drives? Does SN850 still needs more tests?
What would be your model recommendation in terms of macOS support?
Thank you immensely for your + others work! I deeply appreciate it.
I would appreciate if somebody compiles a list of suggested changes. SN850 works great in macOS indeed, but I must admit it is fairly hot.
Samsung 970 pro 1Tb Nvme
Asus Z390-E, i9-9900K
Sapphire Radeon Pulse RX 580 8GB
MacOS Monterey 12.4
Running for about 30 days in OC v0.7.2
Running for about 30 days in OC v0.8.0
- 38.95, 38.13, 37.51
- 38.95, 37.53, 37.93
- 37.66, 38.31, 37.28
Edited:
OK I just now read that this test may no longer be working in Monterey.
My question is because I don’t know so much about trim is if I have an SSD and the NmMe M.2 drive and I wipe one and clone it back to the other every few weeks would that help any as far as the drive not degrading overtime?
I think I might be relating this to the old spindle drives fragmentation defragmentation method and it has nothing to do with this at all or will it help stop degradation of the drive?
say 21000000 does it mean pretty much the same thing as 4294967295?
4294967295 does not have some magical meaning here IMO; it's just the biggest value possible for the option and something programmers etch into their minds. But 21000000 is, well, actually not that long since it's "just" 21 seconds whereas Samsung TRIM can take an extra minute or so if you look at the numbers others have posted. Add a zero to the end for something "reasonable".
Just remember to occasionally boot with it set to -1 to TRIM the disk.
Doesn't that revert to the 9.999999 sec timeout, which won't be enough for the slow TRIM drives? I think disabling the timeout with a value of 0 would be better here.
While we are at it, the testing directions should be updated for macOS 12 by replacing the 4294967295 with 0.
@Artoria2e5 0 disables TRIM.
Either specifying timeout, or completely disabling trim with 0
On MacOS 12 you have to set -1 for TRIM to run completely, on Big Sur 4294967295
Gigabyte Z490 VISION D i9 10850K
Sapphire Radeon Pulse RX 580 4GB
MacOS Monterey 12.4
OC 0.8.3
Samsung 970 pro 2Tb Nvme
999=58-58-58
-1=58-58-58
4294967295=58-58-58
Samsung SSD 850 PRO 512GB
999=24-24-24
-1=25-24-24
4294967295=23-24-24
@vit9696 i want to buy wd blue sn570 nvme pcie ssd which is avalaible for my country
Is it working or not? please reply me i m waiting for your answer
Yes, the WD SN570 works just fine in macOS.
Can confirm SN570, SN750 SE, and SN850 works just fine. I’ve used them in my hacks as boot drive.
Guys @i0ntempest @Edhawk64 please report your results as mentioned in the OP, measuring boot time as described with the values SetApfsTrimTimeout set on -1 999 and 2^32 - 1
I think I have reported the time some time ago in this thread, can’t remember which exact disk though. Will do when I have the time.
@1alessandro1 @i0ntempest thanks guyz for confirming
I bought wd sn570 and tested work fine but read&write speed is not actual
Bitween windows&macos but boot speed very fast about 15 second and less
This screen of read&write speed under windows
MacOS Monterey shows
Write 1204 read 2845
M i missing something for my configration or due to power managment
I m using nvmefix.kext and without kext no increase read&write
Please reply me
Samsung SSD 860 EVO 500GB SATA
motherboard: GA - Z390 M Gaming
CPU: i7–8700K
macOS: Monterey 12.5
Opencore 0.82
From Apple logo to macOS system :
1 : 15 15 15
2 : 15 15 15
3: 15 15 15
No difference at all with the three values (Why ? Maybe because I have a few other hard drives plugged in the computer and it's the minimum time to detect them during the boot ?)
I also did the test with the Samsung SSD 850 EVO 500GB SATA,
It was the same values but after a few boot, it increases a lot (about 47 seconds).
And also the Western Digital - WD Blue SSD - SSD interne 2To M.2 SATA
is not compatible with Trim (About 47 seconds for booting)
Thanks a lot for discovering the problem !!
Samsung SSD 860 EVO MZ-76E250BW 250GB (GPT) SATA
Dell Inspiron 3558 (UEFI) - Intel Core i5 (5th Gen) 5200U /2.20GHz - 8GB RAM DDR3L - Intel(R) HD Graphics 5500/ nVIDIA GeForce 920M - 1366x768
macOS: Monterey 12.6
Opencore 0.85
From Apple logo to macOS system :
1 : 35 35
2 : 35 36
3: 35 36
No difference much at all with the three values. But it take long time 35s to boot completely. Any suggestions? Thank
Hello,
finally, after two months, with the ssd saamsung 860, it came back to a very long boot, 45 seconds, and I don't know why... I tried everything (trimfore disable and enable again, clean Nvram in opencore....
The short boot last 1 month and half and now the problem came back.
So, maybe the 860 is not compatible with trim finally ???
Hello,
finally, after two months, with the ssd saamsung 860, it came back to a very long boot, 45 seconds, and I don't know why... I tried everything (trimfore disable and enable again, clean Nvram in opencore....
The short boot last 1 month and half and now the problem came back.
So, maybe the 860 is not compatible with trim finally ???
mine is 35s boot Monterey even postinstall
How is Kingston KC3000? It has Phison E18. I can get it pretty cheap but want to be sure that it will not be effected.
Thank you.
thank you wd sn550 works fine
威刚S70避坑，不能用
I went on kind of a shopping spree for new NVMe this holiday and have some data points. (Sorry I didn't follow the template)
My machine is a Intel NUC i7-8809G running macOS 12.6.1 with FileVault enabled.
- Samsung 860 EVO M.2 SATA 1TB This was the drive I originally had and it worked fine for years. TRIM was enabled.
- HP EX920 M.2 NVME 1TB Two year ago, I bought this with the intention of replacing the SATA drive but I had random panics. Probably due to high loads. Didn't do too much testing so I didn't know what the issue was or what the TRIM status was.
- Mushkin Enhanced Helix-L M.2 NVME 1TB I bought this a year ago again with the intension of replacing the SATA drive. Also had random panics and actually managed to kill the drive probably do to extreme IO loads. Returned it a week later. Do not ever buy this brand.
- Samsung 980 PRO w/ Heatsink M.2 NVME 2TB I bought this before reading the first post. Had slow boot times that were NOT solved by SetApfsTrimTimeout. In fact, I was not able to boot my primary partition (FileVault enabled) at all. It just got stuck at the loading screen for maybe half an hour before rebooting with a panic. I can boot into my second install (unencrypted) but it would quickly panic when under load.
- Kingston NV2 M.2 NVME 2TB I bought this last week because there was a TechPowerUp review that gave a really good score and said the NV2 had a Phison controller. Turns out, after reading some Amazon reviews, either Kingston sent reviewers a different sample or the 1TB model uses a different controller. The 2TB version has a Silicon Motion controller and panics on high loads.
- Western Digital Black SN850X M.2 NVME 2TB This worked perfectly (two weeks so far without any panics). This was expected from the first post, but I can confirm the X model didn't change anything drastically.
Some other thoughts:
I think having FileVault enabled really exacerbates the issue. With any NVME that panics with a heavy IO load, it will also instantly panic when booting into my FileVault enabled installation.
I don't see enough people talk about HMB and DRAMless SSDs which are the majority of "cheap" NVMEs. macOS does not support HMB and so will get worse random I/O throughput. The performance on sequential reads shouldn't be affected and that's what everyone seems to run on benchmarks. I found that "low throughput" seems to be one of the factors that cause macOS to panic. For example, here is a benchmark run on the Kingston NV2 on macOS:
As you can see the write speeds are terrible and way under the rated speed of 2500MB/s. I'm not sure why the numbers are this bad, but I tried to also run a benchmark on my Windows install with the same drive put into a TB3 enclosure in order to simulate the lack of HMB and you see similar numbers for random writes.
Compare that with the same drive installed internally (note that TB3 overhead probably contributes to the difference in peak read speeds)
Moral of the story: go with WD if you are building a hackintosh.
Haven't taken exact measures because I'm running debug verbose mode and could be misleading but I can confirm that my Samsung 980 (non Pro) loads in -1 as fast as in 4294967295. Probably around 20 to 30 seconds tops.
I'm running this hackintosh since August in a 256gb partition out of 1 TB of space. I'm using the other partition in Windows as an expansion disk (not OS) for games. Health level remains at 100%, no wear detected (I purchased the disk around May / June). No slowdowns or slow boot times.
Has anyone tested the WD SN770? I'm currently running a Samsung 970 Pro and considering the SN770 for replacement.
Has anyone tested the WD SN770? I'm currently running a Samsung 970 Pro and considering the SN770 for replacement.
Hi, i have this one, 2Tb, no issues so far.
I bought also the SN850 2Tb, it works fine too.
is there shorter 2260, 2242 smaller 128/250gb ok that works without this panic/nvme problem? thx
WD SN850x will work? I've an Asus Z390E with i9900k. SN750 is out of market and high price too.
@Kaisar870 read a couple posts up: #192 (comment)
@Kaisar870 read a couple posts up: #192 (comment)
Thank you 😊 my concern is now does it work with z390 ? I mean PCI gen 3x4
Yes, I also had PCI gen 3x4. However, after months of usage, I did notice that there are some times (maybe 1-2 times a week) where the storage would cause macOS to freeze up for 10-15 seconds. I don't know if it has to do with this specific NVMe or if it is an issue with every NVMe. I only used SATA before and never had that issue.
Yes, I also had PCI gen 3x4. However, after months of usage, I did notice that there are some times (maybe 1-2 times a week) where the storage would cause macOS to freeze up for 10-15 seconds. I don't know if it has to do with this specific NVMe or if it is an issue with every NVMe. I only used SATA before and never had that issue.
Yes, I also had PCI gen 3x4. However, after months of usage, I did notice that there are some times (maybe 1-2 times a week) where the storage would cause macOS to freeze up for 10-15 seconds. I don't know if it has to do with this specific NVMe or if it is an issue with every NVMe. I only used SATA before and never had that issue.
Thank you for your patience 😊 do you think this issues cause by SN850X? I've already 970 Evo plus and boot time 1 minute 30 sec. Some apps Open take long time specific Ms office. Also glitch the screen and freeze Evey week even multiple times a day. I've previously installed 850 sata SSD and this was incredibly fast both boot and opening apps.
My advice: stick with the 850 SATA. I had a 860 SATA before and had 0 issues for years. You won't notice any speed increase in day to day use, especially since it's PCIe 3x4.
Samsung SSD 980 500GB NVME (Non-Pro)
motherboard: GA - Z490 Aorus Elite
CPU: i3-10100F
macOS: Ventura 13.4.1
Opencore 0.9.3
HideAuxiliary Enabled, Showpicker and Timeout to 0.
From Apple logo to Login Screen:
1 : 15 15 15
2 : 42 40 40
3: 40 40 40
I can only assume the Trim is either working slowly or failing, I am unsure if I should leave it on -1 or 0 for day to day use until I can buy a more compatible SSD, any tips are welcome.
How is Kingston KC3000? It has Phison E18. I can get it pretty cheap but want to be sure that it will not be effected. Thank you.
Maybe not that great? https://www.tonymacx86.com/threads/general-nvme-drive-problems-fatal.316546/post-2324146
Is there a way to manually trigger the trim or some tool to do that? Because I rather trigger manually when I'm idle than waiting an infinite amount of time at boot.