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.
Any news on TRIM ssd implementation for OS X.
- Thread starter applecultvictim
- Start date
- Sort by reaction score
thanks!, some of the Sand force based SSDs are consuming more watts compared to the Intel SSD.
Also not so sure about the garbage collection (looks like Kingston SSD does that automatically)
since OS X does not support TRIM, kinda of confused which one to choose, intel (Intel Controller) or Kingston (toshiba controller) or OCZ/OWC (Sand Force controller)
thanks!, some of the Sand force based SSDs are consuming more watts compared to the Intel SSD.
Also not so sure about the garbage collection (looks like Kingston SSD does that automatically)
since OS X does not support TRIM, kinda of confused which one to choose, intel (Intel Controller) or Kingston (toshiba controller) or OCZ/OWC (Sand Force controller)
The Sandforce based controllers are currently the only devices that effectively mimic the functionality of TRIM in the SSD firmware. For that reason alone, a Sandforce drive would be my choice.
I've just installed 10.6.5 upgrade.
Seems to be aware of TRIM functionality in the system info :
CRUCIAL_CT256M225 :
Capacité : 256,06 Go (256*060*514*304 octets)
Modèle : CRUCIAL_CT256M225
Révision : 1916
Numéro de série : P609410-MIBY-202A002
NCQ (Native Command Queuing) : Oui
Profondeur de la file dattente : 32
Support amovible : Non
Disque amovible : Non
Nom BSD : disk0
Type de support : SSD
Gestion du TRIM : Non
Type de carte de partition : GPT (Tableau de partition GUID)
État S.M.A.R.T. : Vérifié
Volumes :
Capacité : 209,7 Mo (209*715*200 octets)
Inscriptible : Oui
Nom BSD : disk0s1
SSD :
Capacité : 255,72 Go (255*716*540*416 octets)
Disponible : 87,27 Go (87*273*480*192 octets)
Inscriptible : Oui
Système de fichiers : HFS+ journalisé
Nom BSD : disk0s2
Point de montage : /
Seems to be aware of TRIM functionality in the system info :
CRUCIAL_CT256M225 :
Capacité : 256,06 Go (256*060*514*304 octets)
Modèle : CRUCIAL_CT256M225
Révision : 1916
Numéro de série : P609410-MIBY-202A002
NCQ (Native Command Queuing) : Oui
Profondeur de la file dattente : 32
Support amovible : Non
Disque amovible : Non
Nom BSD : disk0
Type de support : SSD
Gestion du TRIM : Non
Type de carte de partition : GPT (Tableau de partition GUID)
État S.M.A.R.T. : Vérifié
Volumes :
Capacité : 209,7 Mo (209*715*200 octets)
Inscriptible : Oui
Nom BSD : disk0s1
SSD :
Capacité : 255,72 Go (255*716*540*416 octets)
Disponible : 87,27 Go (87*273*480*192 octets)
Inscriptible : Oui
Système de fichiers : HFS+ journalisé
Nom BSD : disk0s2
Point de montage : /
No. It only means what it says: system profiler shows the option if it's an ssd. It doesn't do much more than that.
Hello,
hdparm is a disk utility for linux, which can also trim sector ranges. The hdparm package contains wiper.sh, a script to trim filesystems. I made a patch to trim unmounted hfsplus (and ntfs) filesystems for wiper.sh. After a success report from a snowleopard user, i informed the hdparm developer and after further tests, the patch is now in the hdparm-9.36 package.
i have no mac, the patch was made to trim my ntfs partition from linux. The extension to hfsplus was made some month ago, because hfsplus has a similar bitmap allocation file.
macusers with linux knowledge could use/test wiper.sh, but make a backup before. Eg. using the latest Ubuntu 10.10 live cd to trim a unmounted hfsplus filesystem:
boot the Ubuntu 10.10
open terminal in Applications->Accessories->Terminal
then do the following steps:
sudo bash
add-apt-repository "deb http://archive.ubuntu.com/ubuntu maverick universe"
apt-get update
apt-get install gawk
apt-get install sleuthkit
wget http://sourceforge.net/projects/hdparm/files/hdparm/hdparm-9.36.tar.gz
tar -xvf hdparm-9.36.tar.gz
cd hdparm-9.36
make
make install
cd wiper
now you can use wiper.sh. "sda1" is the first partition of the first disk. "sdb1" is the first partition of the second disk and so on. do at first a dry run:
./wiper.sh /dev/sda1
if everything looks ok do a real run with:
./wiper.sh --commit /dev/sda1
make a backup before. Here is the output from a testrun on a 100Mb hfsplus gpt partition.
http://paste.ubuntu.com/523327/
You can use eg. "fsstat -f hfs /dev/sda1 " to get further filesystem informations to find the right partition. eg.
http://paste.ubuntu.com/525053/
edit: don't use version 9.36 on a resized filesystem.
hdparm is a disk utility for linux, which can also trim sector ranges. The hdparm package contains wiper.sh, a script to trim filesystems. I made a patch to trim unmounted hfsplus (and ntfs) filesystems for wiper.sh. After a success report from a snowleopard user, i informed the hdparm developer and after further tests, the patch is now in the hdparm-9.36 package.
i have no mac, the patch was made to trim my ntfs partition from linux. The extension to hfsplus was made some month ago, because hfsplus has a similar bitmap allocation file.
macusers with linux knowledge could use/test wiper.sh, but make a backup before. Eg. using the latest Ubuntu 10.10 live cd to trim a unmounted hfsplus filesystem:
boot the Ubuntu 10.10
open terminal in Applications->Accessories->Terminal
then do the following steps:
sudo bash
add-apt-repository "deb http://archive.ubuntu.com/ubuntu maverick universe"
apt-get update
apt-get install gawk
apt-get install sleuthkit
wget http://sourceforge.net/projects/hdparm/files/hdparm/hdparm-9.36.tar.gz
tar -xvf hdparm-9.36.tar.gz
cd hdparm-9.36
make
make install
cd wiper
now you can use wiper.sh. "sda1" is the first partition of the first disk. "sdb1" is the first partition of the second disk and so on. do at first a dry run:
./wiper.sh /dev/sda1
if everything looks ok do a real run with:
./wiper.sh --commit /dev/sda1
make a backup before. Here is the output from a testrun on a 100Mb hfsplus gpt partition.
http://paste.ubuntu.com/523327/
You can use eg. "fsstat -f hfs /dev/sda1 " to get further filesystem informations to find the right partition. eg.
http://paste.ubuntu.com/525053/
edit: don't use version 9.36 on a resized filesystem.
Last edited:
I didn't check the whole thread but did nobody read Anands new article about the brother of the MBA SSD aka Kingston V+100.
It shows 3 interesting things.
1. Random write speed is not as important as many think. In real world scenarios it does very well although it is really bad compared to Sandforce etc. in random workload. I guess for servers this won't hold true.
2. You can get by without TRIM.
3. Not using TRIM has its downsides. Higher power consumption and a lower lifetime of the SSD. The latter and how bad it is we will not know for a while and that is kind of the worst part about it.
The former I guess has also to do with Toshiba trying to reduce idle power consumption to a minimum and doing all the work immediately unlike Idle garbage collection implementations. It is most likely the better choice for Notebook use than Samsungs approach, for Desktop it doesn't matter.
It shows 3 interesting things.
1. Random write speed is not as important as many think. In real world scenarios it does very well although it is really bad compared to Sandforce etc. in random workload. I guess for servers this won't hold true.
2. You can get by without TRIM.
3. Not using TRIM has its downsides. Higher power consumption and a lower lifetime of the SSD. The latter and how bad it is we will not know for a while and that is kind of the worst part about it.
The former I guess has also to do with Toshiba trying to reduce idle power consumption to a minimum and doing all the work immediately unlike Idle garbage collection implementations. It is most likely the better choice for Notebook use than Samsungs approach, for Desktop it doesn't matter.
Ey Antiram,
I'm using a Macbook Pro with an Vertex 2 120GB drive and your tutorial seemed to work. I went through the steps pretty smooth, the only thing which caused some problems was that fdisk didn't saw the correct partition layout. It only showed /dev/sda1. This was however, a small vfat partition in front of the hfsplus partition. Parted saw the correct layout and showed /dev/sda2 (hfsplus), which was the partition that needed Trimming. Snow Leopard is running sweet again and the write speeds has gone up to 220MB/s again!!
I'm also encountering another issue with Ubuntu 10.10 live. Most of the time, when booting off the Ubuntu install cd, it generates the following errors in dmesg, which makes trimming, or even review the partition layout impossible:
Sep 24 15:15:28 localhost kernel: ata1.00: status: { DRDY }
Sep 24 15:15:29 localhost kernel: ata1: soft resetting link
Sep 24 15:15:29 localhost kernel: ata1: EH complete
I've tried to play with different power management options of the drive, without luck. I've also tried to enable some modules with modprobe, such as atalib. The use of Ubuntu 11.04 didn't help either. Is there maybe someone, who knows the cause for this, or how to solve it? Maybe Antiram experienced it sometime?
I can however confirm that your tutorial works great(keeping the ata problem out of mind)! Many thanks for your hard work, really appriciate it!
I'm using a Macbook Pro with an Vertex 2 120GB drive and your tutorial seemed to work. I went through the steps pretty smooth, the only thing which caused some problems was that fdisk didn't saw the correct partition layout. It only showed /dev/sda1. This was however, a small vfat partition in front of the hfsplus partition. Parted saw the correct layout and showed /dev/sda2 (hfsplus), which was the partition that needed Trimming. Snow Leopard is running sweet again and the write speeds has gone up to 220MB/s again!!
I'm also encountering another issue with Ubuntu 10.10 live. Most of the time, when booting off the Ubuntu install cd, it generates the following errors in dmesg, which makes trimming, or even review the partition layout impossible:
Sep 24 15:15:28 localhost kernel: ata1.00: status: { DRDY }
Sep 24 15:15:29 localhost kernel: ata1: soft resetting link
Sep 24 15:15:29 localhost kernel: ata1: EH complete
I've tried to play with different power management options of the drive, without luck. I've also tried to enable some modules with modprobe, such as atalib. The use of Ubuntu 11.04 didn't help either. Is there maybe someone, who knows the cause for this, or how to solve it? Maybe Antiram experienced it sometime?
I can however confirm that your tutorial works great(keeping the ata problem out of mind)! Many thanks for your hard work, really appriciate it!
you have gpt partition style and fdisk can only handle msdos partition style (fdisk see a partition which spans the whole drive).
try "irqpoll" as bootparameter for the DRDY problem. Press F8 at the beginning of the ubuntu boot process and add the parameter.
here is a updated wiper.sh which should work save for resized partitions (resized to be smaller, eg after using boot camp to install other OS)
http://www.ocztechnologyforum.com/f...nux-TRIM-tool)&p=575244&viewfull=1#post575244
try "irqpoll" as bootparameter for the DRDY problem. Press F8 at the beginning of the ubuntu boot process and add the parameter.
here is a updated wiper.sh which should work save for resized partitions (resized to be smaller, eg after using boot camp to install other OS)
http://www.ocztechnologyforum.com/f...nux-TRIM-tool)&p=575244&viewfull=1#post575244
AWESOME!
I can confirm this works on my 2010 17'' MBP + Intel Postville 160GB SSD with bootcamp and the new wiper version on the /dev/sda2 (Mac OS) volume.
I also ran chkdsk in windows to make sure that the Bootcamp partition was fine - it found no errors
Booting seems to be slightly faster in OS-X even thought it shouldn't benefit that much when it's trimmed - I should have made proper benchmarks before (with AJA System test or so) to compare it but according to the wiper script it trimmed a couple of areas.
Antiram,
Your suggestion seemed to work! The chipset, which actually causes the problem is working fine with the provided boot parameter. Many thanks for that. I've read that it changes the way it handles hardware interrupts, but how is this exactly explainable in my situation? Wouldn't it be possible to add hdparm(and necessary tools) with your wiper included, to mac ports? This way it would be easier for mac user to test your excellent software, and provide you feedback.
Your suggestion seemed to work! The chipset, which actually causes the problem is working fine with the provided boot parameter. Many thanks for that. I've read that it changes the way it handles hardware interrupts, but how is this exactly explainable in my situation? Wouldn't it be possible to add hdparm(and necessary tools) with your wiper included, to mac ports? This way it would be easier for mac user to test your excellent software, and provide you feedback.
Any news on TRIM ssd implementation for OS X.
Is TRIM really necessary one wonders? Have a look at this link and in particular the final page, page 7.
http://www.bit-tech.net/hardware/apple/2010/07/01/mac-ssd-performance-trim-in-osx/1
Is TRIM really necessary one wonders? Have a look at this link and in particular the final page, page 7.
http://www.bit-tech.net/hardware/apple/2010/07/01/mac-ssd-performance-trim-in-osx/1
I can confirm the wiper.sh from hdparm-9.36 works as described on a HFS+ (Journaled) partition on a Macbook Pro (4,1 -- Core 2 Duo 2.4 GHz, 6GB RAM, Intel 160GB G2 SSD) with the Ubuntu 10.10 LiveCD following the instructions shown below from antiram. System seems faster and I don't have any stuttering anymore. I tried using DiskTester from DigLloyd but that didn't seem to clear up the problem.
I did have an issue upon restarting after using the LiveCD, I got a "No Bootable Device Found" sort of error message, which is apparently related to the LiveCD and is in no way an effect of the TRIM operation. Removed one stick of RAM, reset the PRAM (Command-Option-P-R on boot, hold until get three chimes) and it fired right up, no data loss.
See: http://hintsforums.macworld.com/showthread.php?p=474928 about the error message.
I did have an issue upon restarting after using the LiveCD, I got a "No Bootable Device Found" sort of error message, which is apparently related to the LiveCD and is in no way an effect of the TRIM operation. Removed one stick of RAM, reset the PRAM (Command-Option-P-R on boot, hold until get three chimes) and it fired right up, no data loss.
See: http://hintsforums.macworld.com/showthread.php?p=474928 about the error message.
my 2009 MBA is as fast as the day I took it home, but the SSD's performance has always been quite slow in comparison to some aftermarket drives (like he stated in the linked article). It offers a pleasing "desktop experience" and boot times, but holds little advantage over a good 7200RPM disk drive when doing some heavy lifting.
I'm in the market for the i7 MacBook Pro and I'm torn between which drive to purchase: 7200RPM HDD, Apple 256GB SSD or aftermarket SSD. Does anyone have any recommendations?
Given the pricing and speed advantage of the aftermarket SSD, I think that part of the decision is easy. The OWC SSD is popular here.
I purchased the 200GB Mercury Pro RE SSD from OWC a year ago and it hasn't slowed down one bit (I honestly run Xbench frequently, expecting it to). Of course, it hasn't seen more than 65GB of usage; as it's the primary drive in my Mac Pro and only running the OS and basic applications.
Do they have some kind of trimmer or garbage collector in the firmware, to prevent degradation or something? I ask because the specs for the MercuryPro RE SSD only mentions trimming support in Windows 7. If so, I'll swap it out with a smaller version from OWC and use the 200 gigger in my MacBook Pro (which will see a lot more usage).
Thanks
Last edited:
The OWC drives use a Sandforce controller which performs much the same function as TRIM in the drives firmware. At least that is the idea. Tests do seem to support this is the case.
http://en.wikipedia.org/wiki/SandForce
Better late than never:
http://www.appleinsider.com/article..._trim_support_to_mac_os_x_10_7_lion_beta.html
Osx obtains trim support 2 years later from windows(7).
http://www.appleinsider.com/article..._trim_support_to_mac_os_x_10_7_lion_beta.html
Osx obtains trim support 2 years later from windows(7).