Printable View
As we all know there currently is no way to pass TRIM to a RAID array. I've tried to get some information about this out of Intel, Areca and Adaptec but each time I don't make it very far. While searching on the subject I came across the following:
For those in raid, or those who just want a drive with consolidated free space and clean Nand...bring on TONY-TRIM
I was wondering what the XS community though about "TONY-TRIM." Has anyone from XS used it? Thoughts?
Personally I am running my Raid 0 setup with SLC SSD's and "Garbage Collection" Firmware and I am so far very happy with the performance.
I think that TRIM is 99% for MLC drives to combat their increased internal write latency when compared to SLC drives.
Since I have not migrated to W7 yet not having TRIM is a non issue to me....
Short Answer = no :)
It works :)
Tony actually describes it as manual garbage collection. Maybe Tony Manual Garbage Collection did not have the same ring to it.
Interesting fact that writing FF (all ones instead of zeroes) is doing a block erase really. I wonder if that's NAND fact or just for OCZ SSDs?
But really, consolidating free space and writing data? Nice way to shorten the life-time of SSDs :( (not lashing at Tony/OCZ this time, kudus for even coming up with this for us RAID folks - I am just not happy that at least Intel didn't put TRIM support in RAID drivers)
When I finally get around to setting up my 2x80G G2 RAID 0 setup I plan to break the array, secure erase the drives, rebuild and restore an image when write performance goes down the toilet. Hopefully this won't be a monthly thing. But options are always nice to have.
tony-trim is not a driver , but just a guide on how to setup perfectdisk 10 to clean up free space ? And perfectdisk 10 works with raid 0 . Is this correct ?
So in fact it's not windows7 TRIM , but just a wipe of free space .
Has anyone tried this with windows 7 , RAID 0 and intel drives ?
Update : i tried it , i don't get 95% of the speed back . but it's hard to compare since i did a firmware upgrade of both my drives and they have different write speeds now .
I can tell you my read speeds are definately lower than 95% compared to my first benches . More like 88-90% .
The Trim part of the name is because of what AS-Clean is doing. Although the drive may have plenty of free space if it is in a used state an erase-before-write has to be performed to store data. Cleaning the free space using FFh sends the 2 byte value of 1111 1111. The cells that make up the free space are left at the bit value of "1". As this is also the value the cell represents when in an erased state a write can now be performed without the need for an erase. Combine this process with free space consolidation and it becomes very similar to what Trim is doing.
This might work as they advertice on OCZ drives, but screwes up performance on Intel drives. Every cell is seen as written to after this, so stear clear of this one if you are on Intel Gen. 1 drives. Besides, HDDErase+True Image takes 1/4 of the time this one takes anyway...
Something else is going on. AS-Clean doesn't write anything to the drive. This is a free space equivalent of Secure Erase.Quote:
Originally Posted by Ourasi
AS-Clean might try to do the same as HDDErase secure erase, but it does not work, atleast not within the OS and on Intel Gen. 1 drives, without a valid purge of LBA, and we all know that purging LBA can not be done in OS without TRIM. The Intel Gen. 1 SSD sees every cell as used after AS-Clean, on all 3 of my Intel G1 SSD's, and the cause is probably that every "write" by AS-Clean get a page in the LBA, and as I said, clearing LBA will not happen unless you do a full HDDErase, or have TRIM.Quote:
Originally Posted by Praz
I knew even before I tried AS-Clean, that this would not work, but I had to be sure, so I wasted an hour or so to confirm this. I guarantie you this, this will happen on all Intel Gen. 1 drives, no matter how one theorize or speculate, every "write" by AS-Clean is seen as a write by the SSD, period..
If the fact that some OCZ drives have a manual TRIM tool possebility, and a firmware that support this, "let" AS-Clean do this without touching LBA, I can only speculate, as I currently do not have a OCZ drive to test this on...
The Fastest and easiest way to see the result of an AS-Clean on an Intel Gen. 1 drive, is to do a HDTune Pro 3.5+ Random Access before and after, beware that this might give you a bad evening:eek:
All benches take an significant drop, including IOMeter Random Write, Random Read and Workstation. My adwise, stay away from this if the above conditions are in place..
AS-Cleaner/FreeSpaceCleaner.exe absolutely DOES write to the SSD. (If you're not talking the same app I am, nevermind! )Quote:
Originally Posted by Praz
According to Tony writing FF byte (all 1s binary) it is the same as doing a block erase. Now whether that is true, and whether that is true for non-OCZ SSDs is the question.
On my X25 gen 1's, I wrote zeros to one drive, and FF's to the other. The zeros benched faster. I then wrote zeros to the 'FF' drive, and it got slightly faster.Quote:
Originally Posted by alfaunits
Here's my last test results, using AS SSD bench. These are two Gen 1 80 MB X25's. The first 4 -gaps- in data lines indicate where I ran PD10 consolidate, and Ccleaner wipe free space. The last gap is where I ran ONLY AS-cleaner. On the C drive on right, I ran AS-cleaner first with FF's, then with zeros. The F drive was only zeros.
http://farm5.static.flickr.com/4044/...efa66f5a_o.png