Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Samsung 840 Pros on Sonnet Tempo Pro - enable TRIM?

Yes, Yes enable Trim. Your first link. And make a donation, Cindori is one of the good guys.

No downsides BTW.

Lou
 
For 10.9.2, type in the following command in terminal

sudo perl -pi -e 's|(\x52\x6F\x74\x61\x74\x69\x6F\x6E\x61\x6C\x00{1,20})[^\x00]{9}(\x00{1,20}\x54)|$1\x00\x00\x00\x00\x00\x00\x00\x00\x00$2|sg' /System/Library/Extensions/IOAHCIFamily.kext/Contents/PlugIns/IOAHCIBlockStorage.kext/Contents/MacOS/IOAHCIBlockStorage

And then restart, this will activate TRIM for free.
 
Thanks to everybody for input, +plus the terminal command to enable TRIM, I actually found that github page a while back:

https://gist.github.com/bzerangue/2986122

I downloaded TRIM Enabler, which does the same thing as above for free, and then Paypal'd Cindori $10 a few minutes ago for the convenience to upgrade to pro version, because, why not; hoping it continues being updated and remembers to nudge me to upgrade itself as necessary.
 
@Schismz

Good move, Cindori was among the first to contribute to get the GUI app out. BTW, you can enable Trim as others have mentioned, I have been using it on my raid 0 drive for almost 4 months now, I am not sure whether it works or not on raid drives. However, I have not seen any problem with it being enabled. Speed of the raid partition is pretty much same as when I created the raid array. Mine is full at 70% and I still get a good speed. I have 2 Samsung Pro SSDs in Raid 0 config on Velocity x2.
 
kennyman said:
I am not sure whether it works or not on raid drives. However, I have not seen any problem with it being enabled.
Click to expand...

You can tell if Trim is enabled from System Information.

Lou
 

Attachments

  • System Info.jpg
    System Info.jpg
    141.1 KB · Views: 242
@flowrider

Lou, sorry for the confusion, I was just pointing out that there is no way of knowing if trim is really working as it should or not. For example, I can edit IOAHCIBlockStorage.kext to show a Yes or No in System Profiler. It is just a kext edit, takes 1 minute.

But what I really meant is - is it really working on a raid 0 on a marvell chipset with cMP with intel 5520 chipset? Besides, the way the data is flowing is, Intel chipset then to the Marvell chipset and then to the Samsung controller. Sorry for diving into the tech stuffs.

Samsung ssd controller needs trim, I am sure about that. However, there is no way of knowing whether or not it is working as it should. All I can say is that I have seen Samsung SSDs in Raid 0, on many cMPs (up to 8 units at work), 2 have trim enabled and no-trim on the rest. I have not seen any difference, in fact, the ones that have trim enabled are a little bit slower than the ones with trim disabled.

Now, I have enabled trim on my home mac but I do a shutdown/system reboot once every 2-3 months, keep 10-20% of free space. But the rest of the Macs at work are 24H running, SSDs get full and contents get deleted on a daily basis. But Samsung SSD in Raid 0 still works. To some extent better than the ones with trim enabled.

So personally I am not sure about it, and I do not want to claim that it is working 100% by just looking at system profiler. I am tuning stuffs at the moment, I cannot run quick bench or BM benchmark and pretend, `yes, it is perfect, everyone should use it or buy one`...(sorry Lou I am not interested in this type of hype)

It definitely takes time to fully stress-test core components, depends on people as well. But hey - if it makes you comfortable to think that trim is enabled...then why not:)
 
Even if you do not use TRIM directly, all modern SSD's have internal garbage collection (IGC) that does the same thing, albeit at a much slower pace.

TRIM is nice because it can do it all in one fell swoop because the OS knows when the computer is idle and can call a request for TRIM from the SSD controller. But because the controller doesn't know when the computer is idle it does IGC very sporadically as not to affect user experience. But if you only do normal usage on your computer, you will not see a degradation.

This is how SSD's were able to function ok over time in RAID 0 before RAID controllers could pass TRIM commands. The 2 drives just healed themselves over time.
 
westrock2000 said:
Even if you do not use TRIM directly, all modern SSD's have internal garbage collection (IGC) that does the same thing, albeit at a much slower pace.
Click to expand...

Not quite. The GC doesn't know if a chunk of data that the disk thinks is in use is actually in use by the OS. TRIM let's the OS say that "these chunks are free, go ahead and erase them during GC".

Without TRIM, the disk often has to preserve (copy) data that the OS has deleted - since it has no way of knowing that the file system no longer references the blocks on disk.


westrock2000 said:
TRIM is nice because it can do it all in one fell swoop because the OS knows when the computer is idle and can call a request for TRIM from the SSD controller. But because the controller doesn't know when the computer is idle it does IGC very sporadically as not to affect user experience.
Click to expand...

TRIM has nothing to do with the OS being idle. It means that when the file system in the OS deletes a file, it sends an additional command to the disk saying that the blocks occupied by the file are free, and do not need to be copied during garbage collection.


westrock2000 said:
This is how SSD's were able to function ok over time in RAID 0 before RAID controllers could pass TRIM commands. The 2 drives just healed themselves over time.
Click to expand...

"Healed" is an interesting term for an operation that contributes to the drive wearing out, but does improve performance in the short term.
 
No, no, noOoz... This is way too much thought and effort. Variables and unknowns interspersed with random facts; who wants those? Facts only get in the way of things.

I was looking for PUSH BUTTON, ENABLE SOMETHING, FEEL GOOD about sequence of events that has occurred, run Blackmagic speed test and stare at reassuring numbers *mmmmm* move onwards.
 
AidenShaw said:
Not quite. The GC doesn't know if a chunk of data that the disk thinks is in use is actually in use by the OS. TRIM let's the OS say that "these chunks are free, go ahead and erase them during GC".

Without TRIM, the disk often has to preserve (copy) data that the OS has deleted - since it has no way of knowing that the file system no longer references the blocks on disk.
Click to expand...

Wouldn't the file systems "Table of Contents" (generic term) be enough for the controller to know with a good deal of accuracy what should and shouldn't be there? Because the OS does not get to directly address the NAND in the SSD, the controller dictates where the data actually goes in order to keep wear even across the NAND.

if the OS issues a command to write or delete to a particular set of sectors, the controller should be able to have it's own "table of contents" or index that lets it know what sectors/cells are valid or not. I suppose you could have instances where the OS buffers a lot of disk commands and then for whatever reason never gets a chance to flush those commands to the disk.

I'm not arguing, I'm asking. I could see on the old platter drives that the OS index could be unseen to the drive because everything was generally kept in sequential order. But the SSD controller has to purposefully fragment all the data, so I would assume that it has a very good idea of what is valid and what is not as it has to keep tabs on ALL the data.


"Healed" is an interesting term for an operation that contributes to the drive wearing out, but does improve performance in the short term.
Click to expand...

I think (and this is my opinion) that most people would not care about the hit to NAND endurance the IGC or TRIM causes. It should be trivial to total number of writes each cell can take. Although the fear of NAND endurance decreasing as density increases is justified, for most people, the numbers are still astronomical. I have an early SSD that does not have trim nor a very good, if any, IGC. While it still runs just fine after 5 or so years now, the performance is abysmal. It's down to around 20MB/s write now...sequential.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back