Forgive my ignorance here, but correct me if I'm wrong...
In order for drives to receive a TRIM ATA command behind a hardware RAID controller, that RAID controller would have to support ATA passthrough.
Most RAID controllers do not support ATA passthrough, especially the ones built into motherboards.
If my assumption is correct, then I would bet that even their SSD toolbox program does not work for drives behind a RAID controller -- no firmware updates or anything...am I right?
If so -- Why would Intel be the ones to fix this problem?
The issue here would not be the drives or the firmware on the drives, but which ATA commands that the RAID controller is able to recognize and send to the drives.
If this is the issue, I would suggest finding a RAID controller that allows for ATA command passthrough prior to purchasing your drives.
I am betting this will become big enough of a problem that most RAID controllers from here on out, especially the 6gb sata\sas ones will have to support this.
I know that's not much help now, but I guess my whole assumption could be wrong here...?
After doing some research, it seems the RAID controller actually has to support TRIM, not just ATA passthrough.
It's very unlikely that existing RAID controllers will be updated to support this (what a surprise...).
Too bad windows can't boot from a software RAID 0 -- I never understood why MS didn't add that functionality in.
OS X can boot from software raid just fine.
Regardless...my initial assumption was right -- there's nothing Intel can do about this.
The RAID controller must support the TRIM command and pass it through appropriately to each disk.
Just to re-iterate -- if you are looking to purchase a new setup, definitely call or email the support for the RAID controller you're thinking of using and make sure they support this.
This is an interesting question. Windows 7 Pro supports Software Raid 0 and Raid 1. And Windows 7 Home support Raid 0 software raid.
If one is using the AHCI controller on the motherboard and the standard Microsoft driver, one should be able to make a software Raid with an SSD.
I understand with newer CPUs and motherboards, RAID 0 and 1 have very little performance loss compared to "motherboard raid" solutions.
THE BIG QUESTION IS: --> Do software RAIDs controlled by windows 7 support TRIM??
Maybe this is the easy way to get trim support with RAID 0 or RAID 1 SSDs right now?
Why hasn't anyone tried this yet?
I see no reason why trim would *not* work with Windows Dynamic Disk \ Software RAID.
Even if it didn't work, you can always TRIM the drives manually with the SSD Toolbox.
The problem is when you are using a hardware RAID controller, the OS has *no* idea how many actual disks and what kind of disks are being used to create that RAID...it's all virtualized....but with windows software RAID, it actually knows about the RAID members, and *should* be able to appropriately send the TRIM command to the drives.
Why hisn't this a viable solution? Simple, because you can't boot from a software RAID on windows, which makes this completely useless for most people looking to store their OS \ boot volume on the RAID0. If you're not using it to boot, then it should be fine.
Mac OS X is the only OS I know of that can boot from a Software RAID -- however (lol), Mac does not have native support for TRIM...although I haven't researched it enough, maybe there are utilities to manually trim the drive.
I've been reading a bunch of stuff and apparently Windows 7 Pro/ Ultimate and Enterprise can do software Raid 1 on the boot drive.
The question then is does the Windows software Raid 1 speed up the read speeds in the same way a hardware raid does?
This would theoretically make it possible to create a Raid 1 array as the boot drive and have double the read speeds (probably much less but more than single drive) and the write speeds would be like a single drive. Also you would get the protection of the mirror. And it seems like TRIM would also be automatically be supported...
A couple people thought that software RAID 0 could be used for the boot drive but that doesn't seem realistic...
Here's the examples of people that got RAID 1 to work as the boot drive in Windows 7 Software RAID:
Go to the 10th message written by shadow3081
Just to add my experience, I'm running win7 enterprise and have mirrored my boot volume onto a secondary harddrive with no issues (yep I know the original question was for non boot volumes but figured someone may find this post useful). Converted both disks to dynamic, right click on the boot C: partition in disk management, click 'add mirror' select your secondary harddisk to mirror onto and its good to go !
Also the 12th long message by gerwien50 seems to have gotten software raid 1 to work on Windows 7 fpr the Boot Drive.
Here's another success:
Read posts 11 and 13 from "nopolite" - in post 13, he shows an image from disk manager with software Raid 1 on the boot partition.
So this is possible with Windows 7 pro / ultimate / enterprise. It seems Windows 7 home does not support this feature even though that difference is never mentioned on any of microsofts product comparison charts.
So it may be possible to get SSD TRIM Support with Software Raid 1 on the boot drive with windows 7 - right now. The question, as I said above, does windows software raid 1 give any "READ" performance benefit at all over single disk read performance??
As to the original poster of this thread, I think what you want to do may be possible if all you want is added protection.
If I get a chance, maybe I'll play with this a bit.
Yes, I read that too -- apparently Win7 and 2K8 allow you to create a mirror'd dynamic disk group from the boot drive.
"The question then is does the Windows software Raid 1 speed up the read speeds in the same way a hardware raid does?"
Yes, it does. Since the data is mirrored, the OS can read a file from both drives at the same time -- the speed of reads should be linear, just like RAID0.
However, writes will not be increased at all -- they will actually get slower since it has to write the data to both drives at once.
Just some other information for you -- software RAID1 and RAID0 are actually no different (speed wise) than hardware RAID1 and RAID0.
RAID1 and RAID0 do not require any complex algorithms -- the operations required to do so are not CPU intensive at all.
RAID5 is another story. The parity calculations needed by RAID5 (and RAID3,4,6, etc) are very CPU intensive.
Software parity RAID will never be as fast as hardware parity raid; even with an i7 cpu and tons of ram.
IO Processors \ ASICs (RISC cpus) are designed with minimal instruction sets and are very efficient at these operations compared to the CISC cpus that we use in our personal computers.
There's a problem with Windows 7\2k8's RAID1 implementation though -- if one of your drives fail, you can't boot.
I believe you have to boot from the recovery dvd and rebuild the mirror with a new drive before you can get back into the OS.
Really stupid if you ask me! But then again, it's Microsoft...so no explanation needed..but hey, I could be wrong on that since it's never happened to me.
And again, TRIM shouldn't matter with a software RAID...you can either do it manually with the SSD Toolbox app or Windows 7 should do it automatically for you.
I just think it really sucks that you can only boot from a RAID1 -- not RAID0 or RAID5.
Dynamic disks support both RAID0 and RAID5..there's just no way to install an OS volume to one...chicken and the egg kind of thing.
In order for that to work, MS would have to add their full disk management tool and recognize dynamic disks right from the OS installation -- that is what OS X does. Really surprised that they didn't do that with Windows 7...
Thanks for all the answers on this. I think I am going to just continue using my hardware RAID1 configurationin my Win7-Pro-64 setup for now. I may switch to a single 160GB disk when the prices come down a bit and I have more confidence in SSDs...I have seen several fail (as in bricked), but have never seen an Intel unit fail....yet. Having the redundancy of RAID1 does provide some peace of mind.
I will soon have much more experience with these as the XP-based product my company makes is now switching from a 160GB HDD to the X25m-80GB SSD (we only need about 32GB.) This will let me see how well they perform in the field. We only ship about 50/month of this product but that should give me a good idea of the reliability....and since I am in product support, I will be the first to hear of any problems!
This update has nothing to do with TRIM. This is just an update on the reliablity of the SSD in our particular application. If you read my previous post you will see that we are shipping about 50 of these per month to customers. We have been doing this for several months now and have yet to have a field failure. We have had 2 failures in production, but those were because someone tried to hot-swap them; against all of our procedures. So far I am very happy. Previously our mechanical HDD was one of our highest failure rate items and now it is the lowest (or, should I say, non existant.)
Our company makes various types of communications test equipment. These cost $10k-$300k apiece and weigh from 20lbs to almost 100lbs.
Most customers are high tech communications companies, aerospace and defense companies, and some medical companies.
All of our instruments run XP as the operating system; with the instrument application run on top of XP. Most user turn on the instrument on Monday morning and do not turn it off until Friday...if ever. Some are used in automated systems and are run 24 hours a day. Some are wheeled around to test things, such as aircraft. Of course, every customer is different.
The manual TRIM command via the Intel SSD Toolbox does not work with any RAID configuration, software or hardware. That is documented in the SSD Toolbox User Guide, although there is no mention of the two types of RAID implementations.
When and if it ever will be, for either implementation, is unknown, I have never seen an official statement regarding this.
Of course, whom is responsible for accomplishing this is complicated, and it is unlikely that one solution would cover all bases.
caesarv, sounds like a good idea. FYI, the new chipsets for the Intel "Sandy Bridge" LGA-1155 CPU's, the P67 and H67, will have SATA 3 support on two of their six SATA ports. Thank God Intel has a SATA 3 chipset/driver coming out (from what I've read), the few (one? The Marvell) that exist now don't seem to be the greatest, it seems some people have problems with it.
On the other hand, that may just be torture, since I've seen test's of Intel 80GB SSD's in RAID 0 that had sequential read speeds beyond SATA 2 capability, so you want to RAID your 'drives with a good SATA 3 implementation, but then... we're back to the TRIM in RAID issue.
The rumored specs of the new Intel G3 SSD's do not indicate they will have their sequential read speeds much improved over the current ones, but better sequential write speeds being the main difference, as well as greater capacity at lower cost, and longer longevity. Of course, those are just rumors, and the SATA 3 support in the new chipsets was not known before the Sandy Bridge CPU specs were released, so who knows.
If someone could get TRIM working with their RAID implementation, they'd probably get a lot of attention.