Sign in with
Sign up | Sign in
Your question

Trim working?

Last response: in Storage
Share
a b G Storage
December 8, 2010 7:13:29 PM

I got a highpoint raid controller (model 620) for my SSD. It seems that trim si functioning after I did these benchmarks but I want your input.

Before loading the drive (completely zeroed out)


I ran this test a few minutes after I copied about 55GB of data to the drive and deleted it.


And this test is about 14 hours later.


It appears to me Trim is working but I wanted someone elses input.

The drive is a Kingston SSDNow V100 64GB.

More about : trim working

December 8, 2010 7:19:48 PM

I would say yes its working,best way to know is to go to registry,do a google search to find exactly where to go,Sorry cant remember
a b G Storage
December 8, 2010 7:36:43 PM

I can run via command line to see if trim is enabled which is a "0" which means yes, but that doesnt mean it is being implimented my the controller card unless the drivers support it.

If you are talking about MSACHI in the registry I changed that value to 0 as it should be
Related resources
a b G Storage
December 9, 2010 6:58:58 AM

Looks entirely normal to me. Garbage collection is clearly working to clean up the mess as well.
a c 127 G Storage
December 9, 2010 10:45:43 AM

Probably just garbage collection indeed. But your driver is listed as msahci, while i expected something like a HP-specific driver. the msahci driver supports TRIM, so chances are TRIM is actually working. But no real way to find that out, unfortunately.

The random I/O scores are exceptionally poor for an SSD, though. Do you get the same scores when you connect the SSD to your onboard AHCI-enabled SATA port?
a b G Storage
December 9, 2010 12:21:07 PM

This is a pretty low end SSD (I got it for about 70 bucks). It doesnt need to be high end, this is in a server which just has the OS on an SSD, all the programs and data are loaded onto 2 500GB, 5400RPM drives in a RAID 1, which the OS was stored on at one point.

I get the same scores when it is used ont he onboard controller, and on a different motherboard (Intel P35 chipset, using iastor). I have the pictures saved on a flash drive that is not accessible atm.

I did not install the Highpoint specific driver. I let windows use mashci.

I cant find another specific chart with AS SSD but here is one if you scroll down the Kingston SSDNow V100 128GB while twice the capacity the numbers are close to mine.
http://www.legitreviews.com/article/1469/4/
a c 127 G Storage
December 9, 2010 1:25:46 PM

Ah it has a Toshiba controller. That is rather a funny story; Toshiba buys JMicron NAND controller chips, then laser-etches them to remove the 'JMicron' text on the chip and make it say Toshiba instead. Apparently this move was done after JMicron got a very bad reputation for 'stuttering SSDs' on its JMF-602 NAND controller commonly used in OCZ Core SSDs.

The Toshiba shouldn't have the stuttering issues, but it would be exceptionally poor at random I/O scores considering this is an SSD. It does have very good sequential scores though, and even the random scores are very good compared to a normal HDD. So you still get most of the benefits of this SSD, though the high sequential speeds (200MB/s) are sort of misleading; since SSDs are particularly fast because of their high IOps (random I/O performance). HDDs do about 0,4MB/s of random reads, while SSDs can do up to 200MB/s. Yours is doing 13MB/s which is sort of in between.

Given its price appears to be a decent buy. Please do not that almost all current SSDs can corrupt themselves, which may be important if you run a server that requires data integrity even after a power failure. The solution to this problem is to use newer SSDs with 'supercapacitor'. This includes the new SF2000 Sandforce controller and Intel G3, coming in early 2011.
a b G Storage
December 9, 2010 1:37:31 PM

The server is just for home use, not in a business enviroment. Since all my critical data in on a RAID 1 with mechanical hard drives, I am not worried if the SSD corrupts itself. I just made an image of the drive after i got everything setup and configured as I want it incase of any future problems.

Eventually I will buy a SSD to replace my 3x 6400AAKS in RAID 0 in my desktop, but only when prices drop nicely.

I found your information very informative about the Toshiba and JMicron controller chips. I thought I opened this as a question but apperantly it is a dicsussion and I cant mark you as best answer otherwise I would.
a c 127 G Storage
December 9, 2010 1:44:17 PM

Well thanks anyway. :) 

Don't feel too bad about JMicron/Toshiba; you can buy a much faster SSD in the future and for now this SSD will give you most of the benefits of todays SSDs. In about a year or so SSDs should have surpassed 1GB/s while being quite affordable. These SSDs would have a PCI-express interface intstead of SATA. There are drawbacks though; current PCIe SSD products lack TRIM under Windows OS and there are no native PCI-express NAND controllers yet for the consumer market (FusionIO has some though but very expensive).

So for now your SSD should be fine, and you can get something much better both in performance and datasecurity (supercapacitor) in the near future.

Keeping a backup of important files on your SSD is advisable; especially since many people do not know about the corruption issue on SSDs. SSDs use DRAM chip to store the 'index' of where data is actually being stored. Windows may think sector 1 starts at the beginning of your SSD/storage device, but your SSD may have decided to write it somewhere else. The problem is that when power is interrupted during a 'page mapping operation', there will be corruption next time the SSD is powered on. In normal english: the contents of a book and its index are no longer in 'sync'; the index may say chapter 4 starts at page 119 while in fact it's 855. This still prevents usage of SSDs for very sensitive data, though professional SSDs either do not use remapping or have a supercapacitor to allow the SSD to finish up its tasks whenever power is interrupted, giving them enough time to write everything it needs to the NAND chips which unlike DRAM, can survive its data after power loss.
a b G Storage
December 9, 2010 2:03:39 PM

Dont all drives (SSD or mechanical) have a risk fo data corruption if there is a power failure? I do have a UPS. One of the batteries is toast though, should probably replace that.
a c 127 G Storage
December 12, 2010 6:42:21 PM

HDDs only lose writes, not corrupt existing data. SSDs can corrupt existing/non-written-to data when it loses power. It also has a HPA mapping table that needs to be saved; like an index to where it has actually stored all its data. An outdated mapping table can also cause dataloss/corruption, possibly severe.

Not sure how this translates to reality, but it's been known not to trust SSDs without super capacitor or other method to write safely. This applies particularly to usage in sensitive systems, mostly business/enterprise. But i think home users can benefit too from supercapacitors; the Intel G3 should have it, to be available soon.

Cheers,
- sub
!