Show us your SSD performance 2

[whs]

Lately I've been hearing good things about Intel SSD drives, is there a real difference between Intels and other brands ?

They are definitely more expensive. But they are also very sturdy. I have a couple of Intels since nearly 4 years and never had a problem.

 

[Terronium 12]

I'm rather curious as to why my speeds have fallen off a steep hill. I'm logging what are essentially 3G speeds when I know it's connected to a 6G port (with the accompanying cable). Is that normal? I mean, the drive is about 70-something percent full but does the remaining space really affect speed?

There's a firmware update available but doing so isn't an option as this is a boot drive and I really don't feel like reacquiring files and moving game saves and what have you.

 

[Britton30]

Mike, do you leave your PC on so it can idle and allow TRIM to do its job daily?
With Crucial the firmware update worked fine on a drive with the OS on it, I don't know about OCZ though. If the new firmware doesn't address an issue you have, I'd skip it.

 

[Terronium 12]

Mike, do you leave your PC on so it can idle and allow TRIM to do its job daily?
With Crucial the firmware update worked fine on a drive with the OS on it, I don't know about OCZ though. If the new firmware doesn't address an issue you have, I'd skip it.

Elaborate, please.

And yeah the firmware addresses speed issues but I'm unsure if it needs to be done from a clean slate to be effective. I'll check regarding both though as I'm not sure either.

 

[Britton30]

For TRIM to function the system needs to:
1. Idle on the desktop
or
2. Stay at the logon screen
for two hours or so. It cleans up the deleted files from NAND cells so they can be written to again.
Don't run too many benchmarks, they do a lot of writes to the SSD.

 

[chrysalis]

TRIM is a oddball function, there is actually no way to be sure its actually working, all one can do is check if the windows TRIM command is enabled and make sure AHCI is on and the drive is on a intel controller (which seem to be the requirements), there is debates whether 3rd party controllers it works or not and the reason these debates exist is because its guess work if TRIM is working. Reviewers try to guess by filling up the drive, deleting all files and then testing quickly afterwards which means in their eyes TRIM works quickly after deletion.

When I toggled the windows TRIM command on and off I havent noticed any performance differences, I think on modern drives the built in GC algorithms mean more than TRIM, also intel and samsung owners can manually TRIM their drives whenever they want using tools provided by intel and samsung. So aside from reliability I consider that the 2nd main reason to get samsung or intel.

 

[Britton30]

An Intel controler is not needed for TRIM, GC is not the same and can cause Write Amplification.

 

[Hopalong X]

GC will cause some write amplification but is necessary on SSD's designed to use it.

Without it more blocks will become used or partially used slowing write speed and eventually there is no place to write new info.

GC allows the blocks to be consolidated and others to be deleted available to be TRIM'ed when the TRIM function occurs.

Write amplification - Wikipedia, the free encyclopedia

If you look at the table in the section just above the GC at the above link marked Factors affecting the value you will note GC is a "Good" value.

The two major bad values for WA are Wear Leveling (though necessary) and Random Writes.

Random Writes are the reason running too many Benchmarks are counter productive to SSD life.

My take on things.

 

[chrysalis]

yes he posted a link which didnt prove his point

my point was not saying TRIM was bad but more that I had no real way of knowing if the drive was actually carrying out TRIM commands, and that even if automated TRIM doesnt work samsung and intel allow manual TRIM runs that can be scheduled as well.

TRIM as I understand it just allows an SSD to know earlier which pages are invalid.

With no TRIM at all a ssd would only know a page is invalid when it gets a request to write new data to those pages, then there is a performance loss because at that point it has to the erase at the same time as the write is occuring. If TRIM works then this can be done when the drive is idle so the blocks are pre erased. According to that wiki page a drive would need some kind of GC for TRIM to be effective.

 

[whs]

Here is some literature that might be useful for the discussion: Garbage Collection and TRIM in SSDs Explained - An SSD Primer - The SSD Review The SSD Review

 

[Britton30]

Apparently some people can't comprehend an article. There will be no writes to a page if any of the blocks therein contain data. The SSD won't erase, then write to it, your are describing TRIM.
Outta here.

 

[chrysalis]

I think you have misunderstood TRIM, multiple articles say the same thing even the wiki one you linked to.

It notifies the drive where invalid data is, thats it, nothing more. TRIM doesnt do any erasing. Its GC that does that.

 

[Hopalong X]

Okay it deletes it. Semantics.

Third section down the page.> TRIM - Wikipedia, the free encyclopedia

Operation
The TRIM command is designed to enable the operating system to notify the SSD of which pages of data are now invalid due to erases by the user or operating system itself. During a delete operation the OS will not only mark the sectors as free for new data, but it will also send a TRIM command to the SSD with the associated LBAs to be marked as no longer valid. After that point the SSD knows not to relocate the data from those LBAs during garbage collection. This will result in fewer writes to the flash, reducing write amplification and increasing drive life. Different SSDs will act on the TRIM command somewhat differently so the final performance can also be different between different SSDs.
The command irreversibly deletes the data it affects.

 

[Hopalong X]

I haven't updated this in a long time. Xtremeforums SSD endurance testing.

Samsung 830 is the champ so far. Post #5104 last update.> SSD Write Endurance 25nm Vs 34nm - Page 205

(GiB) 3,718,760
(TiB) 3,631
(PiB) 3.57
---------------------------------------------------------------------------------------------------------------------

For those that followed my earlier posts.
The Kingston 40MB (Intel X25-V rebranded) has topped 1PiB as of August 7, 2012.
It is one of the original SSD's and has been running since May 16, 2011.
I don't have the most recent write total.

It sure isn't the fastest but just keeps on working.

 

[Britton30]

Interesting thread Mike.

 

[essenbe]

I have moved it from the 6gb Marvel ports to one of the 6Gb Intel ports and it has definitely improved, however it is difficult to tell which one comes out best as each fair better than the other depending on which benchmark I use

As an OS drive, you primarially use the 4KB read stats. Both benchmarks show the Intel controller faster on the 4KB read speeds.

 

[chrysalis]

Okay it deletes it. Semantics.

Third section down the page.> TRIM - Wikipedia, the free encyclopedia

Operation
The TRIM command is designed to enable the operating system to notify the SSD of which pages of data are now invalid due to erases by the user or operating system itself. During a delete operation the OS will not only mark the sectors as free for new data, but it will also send a TRIM command to the SSD with the associated LBAs to be marked as no longer valid. After that point the SSD knows not to relocate the data from those LBAs during garbage collection. This will result in fewer writes to the flash, reducing write amplification and increasing drive life. Different SSDs will act on the TRIM command somewhat differently so the final performance can also be different between different SSDs.
The command irreversibly deletes the data it affects.

That last line is misleading. what it means is because TRIM is going to mean deleted files are actually going to be erased much quicker than without TRIM then deleted files are unlikely to be recoverable. The bit you pasted clearly states that TRIM is telling the drive where deleted files so during GC they can be deleted, it doesnt say TRIM itself actually erases the files. We going to have to agree to disagree if we never going to agree with each other, to me it seems very clear tho.

TRIM = SSD knows where deleted files are and can erase before the pages are needed again for writes.
NO TRIM = SSD doesnt know where deleted files are until the OS tells it to overwrite those pages with new data. However manual trimming would still work in maintaining the SSD inactive pages.

 

[chrysalis]

I haven't updated this in a long time. Xtremeforums SSD endurance testing.

Samsung 830 is the champ so far. Post #5104 last update.> SSD Write Endurance 25nm Vs 34nm - Page 205

(GiB) 3,718,760
(TiB) 3,631
(PiB) 3.57
---------------------------------------------------------------------------------------------------------------------

For those that followed my earlier posts.
The Kingston 40MB (Intel X25-V rebranded) has topped 1PiB as of August 7, 2012.
It is one of the original SSD's and has been running since May 16, 2011.
I don't have the most recent write total.

It sure isn't the fastest but just keeps on working.

any thoughts on why the samsung 830 is doing so well with its apparent high write amplification which should increase wear? and sandforce doing poorly on the same tests with its compression decreasing wear?

 

[kbrady1979]

I haven't updated this in a long time. Xtremeforums SSD endurance testing.

Samsung 830 is the champ so far. Post #5104 last update.> SSD Write Endurance 25nm Vs 34nm - Page 205

(GiB) 3,718,760
(TiB) 3,631
(PiB) 3.57
---------------------------------------------------------------------------------------------------------------------

For those that followed my earlier posts.
The Kingston 40MB (Intel X25-V rebranded) has topped 1PiB as of August 7, 2012.
It is one of the original SSD's and has been running since May 16, 2011.
I don't have the most recent write total.

It sure isn't the fastest but just keeps on working.

any thoughts on why the samsung 830 is doing so well with its apparent high write amplification which should increase wear? and sandforce doing poorly on the same tests with its compression decreasing wear?

Samsung makes everything in their SSD's, and they hand pick the components that go into them. They make the controller, as well as the firmware. It's no surprise these drives work as good as they do......it's Samsung

 

[Dave76]

Sandforce drives aren't doing too well because if you write too much in a short period of time the firmware will slow down the drive to keep you from writing so much.

AFAIK Sandforce is the only one doing that, other brands let you decide what you want to do.