Intel Updates SSDs, Supports TRIM, Faster Writes

MojoKid writes "Intel has just released a firmware update for their 34nm Gen X25-M solid state drives that not only boosts sequential write performance, but adds support for the TRIM command as well. A performance optimization tool is also being released today, for users of Windows Vista and XP, who won't be able to take advantage of TRIM. After being flashed with the new firmware update, Intel's 34nm Gen 2 X25-M 160GB drive offered increased performance in a myriad of benchmarks shown here, and sequential write performance was increased on the order of 30%."

A better write up at anandtech

[Zebadias]

http://www.anandtech.com/storage/showdoc.aspx?i=3667 [anandtech.com]

Re:Great

[Anonymous Coward]

TRIM is not in the file system, TRIM is in the operating system. The file system keeps track of where the files are at, and independently, the TRIM command sent from the operating system permits the drive's logic unit to keep track of where the files aren't.

As a result, it is file system independent.

Re:Great

[tepples]

Now what file systems support TRIM?

Any file system whose checking program supports retrieving a list of cluster ranges that aren't in use can be made to support TRIM. These include any FS that uses a "bitmap" to record sector allocation (e.g. HFS or NTFS), as well as any that use a linked list of cluster numbers (e.g. FAT32, exFAT).

Re:Great

[TheRaven64]

Well, not exactly. The filesystem driver in the OS needs to support it. It's not something that you can just support for every filesystem easily, because each filesystem keeps track of which blocks are in use and which are not in a different way. An OS will support it by exporting a trim command from the block driver level, which will be ignored by older drivers and mapped to the SATA TRIM command on drivers for newer controllers. Each filesystem will then be responsible for actually issuing these commands.

Re:Great

[bill_mcgonigle]

Now what file systems support TRIM?

ext4, but the block layer needs to handle it too. There was some LKML traffic a couple months ago about smart designs for this - it's probably not in current distro releases yet. TRIM can be very expensive if not well-optimized (the non-optimized demo took a half hour to delete a kernel tree with TRIM on a supporting SSD) and the right thing to do may depend on drive model capabilities. The moral is it's not worth doing poorly, and doing it right may require some re-plumbing. But the upside is that Linux ought to be very fast and efficient about it when it lands because smart folks are making sure it ships when it's ready, not by some arbitrary date.

Re:Great

[sverdlichenko]

There is no trace of commits to linux filesystems, but article about Microsoft claims NTFS was updated.

Re:OSX ?

[Anonymous Coward]

Not yet. The first place you'll see it is on OS X Server. Keep in mind that TRIM is not a ratified standard yet. Until it is, Apple has stated that OS X isn't going to support it.

Re:A better write up at anandtech

[Vigile]

Here is PCPer.com's post on this update:

http://www.pcper.com/article.php?aid=805 [pcper.com]

Re:TRIM

[clarkn0va]

The first mistake most people make when talking about SSDs vs HDDs is comparing cost/GB. SSDs are not for storing data. They're for installing your OS and programs, while your data goes to the fileserver or the secondary drive in the workstation. Almost every system builder I've talked to fails to recognize that for your typical home or office user, spending an extra $100-200 on a solid-state system drive, even if it means reducing your CPU budget correspondingly, will show huge gains in system usability and responsiveness. I've built several systems on this philosophy and my customers couldn't be happier.

Re:They still are crap compared to Fusion-io

[tlhIngan]

That's a ridiculous comparison. FusionIO cards are not anywhere near the same class at the Intel drives. They have a certain purpose and are geared toward a certain market. The Intel SSDs are a completely different beast. You can't very well use a PCI card in a notebook, which is a typical host for SSDs. The Intel SSDs are coming pretty close to maxing out the SATA 2 port, and that's all that matters for consumer-level systems. It's pretty frelling impressive. I, for one, welcome our performance-improving firmware overlords!

Yes you can.

Many netbooks with SSDs don't have a SATA-connected hard drive. Instead, they have an SSD connected to the miniPCIe slot inside. The card pretends to be a standard IDE hard drive (after all, an Intel SSD on a SATA controller is the same thing - the SATA controller sits on the PCIe bus), and BIOS boots from it when it enumerates all the storage controllers and storage devices attached. Linux etc. see it as a standard disk and nothing special, too.

I'm referring to SSDs that connect via miniPCIe via the PCIe interface - there are a bunch that are just thumbdrives and use the USB side of the miniPCIe slot. (Like ExpressCard, miniPCIe has both PCIe x1 and USB 2.0, WWAN cards use the USB side, as do card readers and the like).

The reason for stuff like SATA SSDs is because it's familiar to more people - you have a black box with a SATA interface, it must be a hard drive and acts like a hard drive (or it could be optical, but I'm assuming a modicum of intelligence). Plus, it works in cases where you have SATA but not PCIe.

Re:Intel change is great, but...

[Spad]

Defragging SSDs is not only mostly a waste of time (Seek time is the same regardless of where the data is physically on the drive so unless you're dealing with heavy fragmentation of large files it won't have any effect), but it reduces the lifespan by needlessly reading and re-writing data all over the place; there's a good reason Windows 7 automatically disables defragmentation for SSDs.

Re:Intel change is great, but...

[Fross]

You should not defrag an SSD. It won't give a performance boost, and will just contribute to wearing the drive down. Fragmentation is only an issue where access is not truly random, as it is with an SSD.

Example discussion: http://www.tomshardware.co.uk/forum/page-246283_14_0.html [tomshardware.co.uk]

The controller should do a decent enough job of spreading out the data for you.

Re:Great

[morgan_greywolf]

Actually, you're not correct. Support has been included in LKDEV devices and and ioctl since 2.6.28 [kernelnewbies.org]. TRIM support (called 'discard' by the Linux kernel maintainers), is included the latest kernel builds in the ext4 filesystem (see the LKML, etc.)

Re:OSX ?

[ceoyoyo]

Don't buy an SSD from any major manufacturer as a build-to-order. They almost all use Samsung drives, which have a lot of problems. Let them put a cheap hard drive in your machine and then put your own aftermarket SSD in it - one with either an Indilinx controller or an Intel.

http://www.anandtech.com/storage/showdoc.aspx?i=3631&p=19 [anandtech.com]

Re:OSX ?

[Latinhypercube]

...like Blu-ray ?

Bricked 30% faster than before!

[ozbird]

Oops. [intel.com]

New firmware was pulled from Intel website

[lbschenkel]

Intel pulled the new firmware from its website because it would brick the drive in some machines with Windows 7: http://www.buzzbox.com/preview/intel_pulls_ssd_toolbox_for_killing_drives_under_windows_7/?id=154783 [buzzbox.com]

I have Windows 7 and a X25-M G2 that I was going to update but I gave up after I found via Google a lot of forums posts from people who bricked their drives with the new firmware.

A word to the wise!

[rabtech]

WARNING: Intel has pulled the firmware because there appears to be a chance of bricking the drive. Users report that the firmware updates successfully, but after rebooting Windows detects changed hardware, installs drivers, and after rebooting again the system BSODs and/or won't boot at all. The drives appear to be bricked unless reformatted.

I have one of these drives and I'm holding off until the dust settles.