Windows 7 SSD Performance Overview

[geoffry]

From hothardware, I've just glanced at it a bit so can't give any cliffs yet.

http://hothardware.com/Article...-Performance-Analyzed/

 

[geoffry]

It looked like TRIM didn't do much.

But I believe it will slow the speed degredation over the life of the SSD as it gets filled up more than improve raw speed on a fresh drive...correct me if I'm wrong the experts around here.

 

[taltamir]

trim just NOTIFIES the controller of free space. after which it is entirely up to the control as to HOW it manages it and what it preemptively clears... done correctly it COULD completely eliminate any and all speed degradation due to internal fragmentation, making the drive work in "new state" all the time... BUT, it took intel to make the first proper SSD drive controller... who knows when we will have someone make a proper TRIM implementation. I expect early implementations to be primitive and limited.

 

[Rebel44]

I didnt see anything about Wiper (OCZ tool to send TRIM) in that review - they just said that they compared older FW (1275 ??) with FW with TRIM support (1370) - if they didnt use Wiper that would mean that they never used TRIM - they just compared 2 versions of FW.

 

[VaultDweller]

if they didnt use Wiper that would mean that they never used TRIM - they just compared 2 versions of FW.

I thought the idea was that Windows 7 supports TRIM natively making utils like Wiper unnecessary?

 

[Idontcare]

Originally posted by: VaultDweller

if they didnt use Wiper that would mean that they never used TRIM - they just compared 2 versions of FW.

I thought the idea was that Windows 7 supports TRIM natively making utils like Wiper unnecessary?

Yeah, my impression too.

 

[taltamir]

windows 7 DOES support it natively and does NOT require any tools... RC1 of win7 has already implemented final version of trim, it sends trim with every erase (regardless of what drive it thinks it has) notifying the drive of free space. Any drive that doesn't recognize the command ignores it, those that do recognize the command (say, indilinx 1370 firmware SSDs) act on the data it gives them.

 

[Doctor W]

1370 for the indilinx doesn't support the actual TRIM command as far as I know. They implemented a custom vendor-specific command to use the wiper tool (that behaves identically). Rumors are the next one actually supports the real TRIM command.

 

[taltamir]

the wiper does NOT behave identically, lemme quote myself:

TRIM is an upcoming drive controller command meant to inform a solid state drive (SSD) of free space, so that it may preemptively defrag it to increase performance.[1]
SSD defragmentation is a very different process, with different causes, than a spindle HDD defragmentation. Regular defragmentation tools do not work with SSD as they are meant to solve different issues with the storage of data. The root cause of the issue is that SSD drives cannot actually modify data without first deleting it. Furthermore SSDs can write 4kb blocks at a time, but only delete 512kb blocks, often requiring a read-erase-modify-write cycle instead of a simply write operation. To avoid unnecessary work and slowdowns, spindle drives are only informed of data having been erased when they are told to modify the contents of the sectors containing it, not when the user actually deletes the data. On the other hand, preemptively erasing unused SSD blocks allows new data to be simply written without having to read, erase and modify existing data, greatly improving speed. By informing the controller of "free space" it can actually erase it in preparation for future writes.
TRIM has already been implemented in Windows 7 release candidate, but until solid state drives are updated with firmware that can understand the command, it will simply be ignored.
Some drives such as the gskill falcon and the ocz vertex have a specialized tool called wiper.exe that claims to perform on demand "trim", however it is actually a specialized defragmentation tool for SSDs (regular defragmentation tools do not work on SSD).[2] It simply clears unused space, while TRIM actually informs the controller about unused space allowing it to continuously manage resources for best performance.

oh... and wiper tool? it corrupts data on 64bit oses... (and sometimes 32bit oses too)

Also the whole PURPOSE of trim is to allow the drive to continuously manage resources, running a tool now and then completely defeats that purpose as it is a one time clean and you immediately start running out of ready to write space again.

intel drive already performs pretty agressive background defragging from what I have read in pcper. trim will allow it to be more efficient about it. but it does not necessitate higher efficiency, trim only informs the controller, which can then implement a bad cleanup scheme, or a good cleanup scheme based on the data provided by trim.

 

[Doctor W]

The end result of wiper is the same as a system that's been trimming automatically since the filesystem was created. Any unused space is marked as available. After that yeah, it's up to the controller to make use of the information. Given that a single run of wiper for me instantly restores out-of-the-box performance, it appears that Indilinx did a pretty good job with it. Once the drives accept the 'official' (not really yet) command from Win 7, the end result should be the same without having to run the program.

Also, wiper behaves properly. If you turned on Win 7 trim and had a trim-capable SSD you'd get the exact same corruption on the systems that are having problems with wiper. It's a driver/controller bios issue. They're truncating the addresses. And I haven't seen a single case of a 32bit system having problems. I've trimmed my two systems dozens of times, no issues.

 

[taltamir]

wow... you have never seen a problem on TWO SYSTEMS!... the wiper tool DOWNLOAD PAGE warns that it has problems with data corruption, and many have experienced it on various systems.

 

[Doctor W]

The annoyance is people act like wiper is totally random and may suddenly destroy your data. It's an incompatibility with certain drivers & chipsets. Either it will trash a drive the very first time, or you're good. And as I said, the *exact same* systems will have the exact same problems with real trim. It has nothing to do with indilinx really.

 

[taltamir]

do you have any confirmation, articles, something, that discusses the explanation you give?
Because this is the first I have heard that it is a specific bug in certain controllers and extends to any sort of trim.

 

[usernamereserved]

TRIM has to be supported at several levels.
The OS has to support it: Win 7 does.
The controller driver has to support it: The native MS Win7 driver does and Intel have a couple of beta drivers out there for Win7 that does support it.
The SSD controller has support it. Intel X25M does and the Indilinx Barefoot also supports TRIM.
The firmware must support TRIM.

RE OCZ Wiper: It will work on most chipsets if the native Win7 MS SATA drivers are used.
In my experience i can confirm Doctor W statement. If it works once, it does not seem to cause problems later.

 

[taltamir]

The controller driver has to support it: The native MS Win7 driver does and Intel have a couple of beta drivers out there for Win7 that does support it.

The controller does NOT have to support it specifically, the controller just forwards ALL commands without understanding ANY of them.

I heard a claim from one guy that some controllers truncate the data because its 64bit (huh), but that is not a trim issue specifically, those controllers would truncate other data too. If it is even a real case (one guy telling me this on anandtech is no proof it is)