According to this paper SSD's are very insecure. One drive reported that the data was gone when in fact all of the data was recoverable.
Which drive was that? This paper does not tell us that. It would be very helpful if Anandtech would replicate this study and tell us which drives performed in what capacity. Much more helpful than whether drive A performed a write/read 10kb/sec faster than drive B.
I might be misunderstanding what's being said here, but unless you do a "secure empty trash" all that's happening when you empty the trash is that you're telling the computer that it can write over the sectors that were "emptied". The trashed data is actually still there until overwritten and thus recoverable. This is true with all drives.
SSD drives do not store data as platter drives do. An SSD drive has a controller on board that is independent of the operating system. Data is stored all over the drive at random and tracked by the controller and firmware. When you use Secure Delete Trash or Eraser on a file stored on a platter drive it will erase the data, but used on an SSD the operating system is blind to the actual location of the file.
The controller removes data using its firmware "garbage collection" to prepare it for new writing. In addition, some drives have significant space that is not accessible by the user that stores and rewrites data, so if you do a full wipe none of that data will be wiped. Also some data was found to be recoverable on one drive after twenty wipes had been performed due to how the wipe was implemented by the controller on that SSD.
To clarify: I was referring strictly to SandForce based SSDs. Only Intel has a custom firmware in their SandForce based 520 and 330 series SSDs - other OEMs use the firmware that SandForce provides.
When we hop off the SandForce train, custom firmwares are much more common. Samsung makes everything from DRAM to firmware, Micron/Crucial uses Marvell controller but everything else is in-house, OCZ uses (possibly custom) Marvell controller in Vertex 4 but in-house firmware, and so on.
How come there's no Plextor M3 Pro review, or at least including it in the charts? That's the one to beat, not the stock M3. It's supposed to be the "best of the best" of Marvell drives and AT has ignored it...
I've had Plextor M3 Pro for a couple of weeks now and review should be up next week at the latest. I didn't have all data yet so couldn't include it in the charts, plus it might have stolen the spotlight so I decided to save all data for a dedicated review.
Thanks for the review. Could AT please post the physical measurements of the drive? Some laptops are moving to the 7mm thick form factor and will not accept standard 9.5mm thick 2.5" drives.
With the Samsung 830 going for nearly a dollar a gig at the critical 128 gig capacity, I really can't imagine why someone would buy a competitor's SSD. They're already way ahead of most on performance and reliability, for much less than the vast majority of alternatives.
Interesting how that old Intel controller in the 320 is still kicking ass where it counts (random reads, latency, IOs at low queue depths). I really feel like most SSDs - all of them, really, besides the 830 and 320 - are designed more to score well in benchmarks than to meet the performance needs of desktop/workstation users.
Really glad we have Anand and his colleagues working on understanding this stuff.
Indeed. I finally sprung for the 256G Samsung 830 last Friday, since it was on sale for only $235 w/ free shipping, and because it's still "best in class" overall. The Crucial m4 was a close 2nd choice due to its reliability rep as well, but the write speeds suffer a bit. (Thanks for the SSD comparison benches AT!)
Trust me, you'll be really glad you didn't go for the M4. After my old Vertex (first gen, Indilinx! haha) was dying, I replaced it first with a 128gb M4, largely because of its reputation for reliability.
Man, was that a mistake. I was on FW 0309 (which had been issued in a hurry to correct a BSOD issue created by the last FW update) and the thing had major issues. Stuttering, hanging, inability to resume from sleep and so on. The Windows event manager was lit up red like a Christmas tree, and this was with a fresh install on a Lynnfield based machine that worked beautifully with my old Vertex.
Turns out the thing had major, unacknowledged (outside of some hard to find forum posts) compatibility issues with Intel's RST an a number of power management features in Windows 7. I guess Intel and Microsoft are really obscure little companies whose products don't merit compatibility testing? Beats me.
SMART says it has only 39% of its (writable) life left after about 2.5 years of use: 209 Remaining_Lifetime_Perc 0x0000 --- --- --- Old_age Offline - 39
If you loaded the article right when it went live, it's possible that you saw two different tables. I updated the prices just before I posted this, but forgot to copy-paste the new table to the final page. Both tables should be up to date now, though.
Lite-On owns Plextor, so it is misleading to say that Plextor buys the SSDs from Lite-On. It would be more accurate to say that the Plextor / Lite-On relationship is like the Crucial / Micron relationship.
At my company we use McAfee Endpoint Encryption, how does this affect the performance of the drives and how much does this change over time? I went for a Plextor M3 256GB and seem to be working fine as for now, but only got it now for 2weeks.
My company also uses the McAfee Endpoint Encryption. I upgraded the WD 250 GB 7200 RPM drive to a Samsung 830 258 GB SSD in my Dell laptop (I5 2540M processor) a couple months ago. It definitely helps, especially with boot time, but McAfee EE adds a lot of processor overhead, which a drive can't help with. It's still slow as a slug compared to the same machine and software configuration without McAfee with a traditional 7200 rpm hard disk.
Anand is waiting for the final version before publishing any data. Apparently there are a few things that OCZ promised will be featured in the final version, but I'm not sure what.
I've been running first gen vertex drives in raid 0 since they were released and since installing the firmware on them that enabled garbage collection they have been fine and they are routinely at 90% capacity.
Corsair has not posted any firmware updates so in this case it's the stock firmware (1.00 I think). I'm not 100% sure what firmwares the other drives are running but you may be able to find that info in the review of a certain SSD. However, some drives were tested quite a while ago so they may not be running the latest firmware. We try to update our charts if there are significant gains from a newer FW, though.
"Neither Plextor or Corsair manufactures the drives they sell. Both companies buy their drives from Lite-On and if you take a look at Lite-On's M3S SSD, it seems a lot like Corsair's Performance Series Pro and Plextor's SSDs, doesn't it? That's why the hardware of the Performance Pro and M3 is the same—they come from the same plant."
In fact Plextor is a part of PLDS, a joint venture between Lite-On IT and Philips.
What about write amplification and the longevity of the drive? I thought that would be a hot topic given the aggressive garbage collection! AT is about the only site where such a thing would even be addressed, so I'm sorry there's no discussion here.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
41 Comments
Back to Article
FunBunny2 - Monday, May 14, 2012 - link
-- Only Intel uses in-house firmware whereas the rest use the firmware that SandForce provides.OCZ seems to say that they do their own firmware. How to know which is which?
SilthDraeth - Monday, May 14, 2012 - link
Doesn't Samsung use their own firmware also?Operandi - Monday, May 14, 2012 - link
Samsung uses their own everything.vol7ron - Monday, May 14, 2012 - link
Can someone do a follow-up to see that those Sequential Read numbers are right for the Vertex 4?It just seems odd that the Vertex-4 bested the Vertex-3 on everything, but was significantly lower with Seq.Read
SSD_Privacy - Thursday, May 17, 2012 - link
Yes Samsung does, but does it erase your data when it says it does? The Corsair, OCZ etc.?http://www.usenix.org/event/fast11/tech/full_paper...
According to this paper SSD's are very insecure. One drive reported that the data was gone when in fact all of the data was recoverable.
Which drive was that? This paper does not tell us that. It would be very helpful if Anandtech would replicate this study and tell us which drives performed in what capacity. Much more helpful than whether drive A performed a write/read 10kb/sec faster than drive B.
appliance5000 - Friday, May 18, 2012 - link
I might be misunderstanding what's being said here, but unless you do a "secure empty trash" all that's happening when you empty the trash is that you're telling the computer that it can write over the sectors that were "emptied". The trashed data is actually still there until overwritten and thus recoverable. This is true with all drives.SSD_Privacy - Saturday, May 19, 2012 - link
SSD drives do not store data as platter drives do. An SSD drive has a controller on board that is independent of the operating system. Data is stored all over the drive at random and tracked by the controller and firmware. When you use Secure Delete Trash or Eraser on a file stored on a platter drive it will erase the data, but used on an SSD the operating system is blind to the actual location of the file.The controller removes data using its firmware "garbage collection" to prepare it for new writing. In addition, some drives have significant space that is not accessible by the user that stores and rewrites data, so if you do a full wipe none of that data will be wiped. Also some data was found to be recoverable on one drive after twenty wipes had been performed due to how the wipe was implemented by the controller on that SSD.
exallium - Monday, May 14, 2012 - link
I believe Crucial uses the Marvell controller on the M4Kristian Vättö - Monday, May 14, 2012 - link
To clarify: I was referring strictly to SandForce based SSDs. Only Intel has a custom firmware in their SandForce based 520 and 330 series SSDs - other OEMs use the firmware that SandForce provides.When we hop off the SandForce train, custom firmwares are much more common. Samsung makes everything from DRAM to firmware, Micron/Crucial uses Marvell controller but everything else is in-house, OCZ uses (possibly custom) Marvell controller in Vertex 4 but in-house firmware, and so on.
Tommyv2 - Monday, May 14, 2012 - link
How come there's no Plextor M3 Pro review, or at least including it in the charts? That's the one to beat, not the stock M3. It's supposed to be the "best of the best" of Marvell drives and AT has ignored it...Kristian Vättö - Monday, May 14, 2012 - link
I've had Plextor M3 Pro for a couple of weeks now and review should be up next week at the latest. I didn't have all data yet so couldn't include it in the charts, plus it might have stolen the spotlight so I decided to save all data for a dedicated review.Mathieu Bourgie - Friday, May 25, 2012 - link
Any update on this? I'm dying to see a AT's review of the M3 Pro :Djwilliams4200 - Tuesday, May 29, 2012 - link
Well, "next week" has come and gone.I see that you have the M3P data in the benchmark database, but where is the review article?
Voldenuit - Monday, May 14, 2012 - link
Hi Kristian,Thanks for the review. Could AT please post the physical measurements of the drive? Some laptops are moving to the 7mm thick form factor and will not accept standard 9.5mm thick 2.5" drives.
Kristian Vättö - Monday, May 14, 2012 - link
It's 9.5mm thick, like most 2.5" drives. I added it to the article so it's no longer unclear :-)Voldenuit - Monday, May 14, 2012 - link
Thanks!How's that for awesome-fast service!
GrizzledYoungMan - Monday, May 14, 2012 - link
With the Samsung 830 going for nearly a dollar a gig at the critical 128 gig capacity, I really can't imagine why someone would buy a competitor's SSD. They're already way ahead of most on performance and reliability, for much less than the vast majority of alternatives.Interesting how that old Intel controller in the 320 is still kicking ass where it counts (random reads, latency, IOs at low queue depths). I really feel like most SSDs - all of them, really, besides the 830 and 320 - are designed more to score well in benchmarks than to meet the performance needs of desktop/workstation users.
Really glad we have Anand and his colleagues working on understanding this stuff.
zcat - Monday, May 14, 2012 - link
Indeed. I finally sprung for the 256G Samsung 830 last Friday, since it was on sale for only $235 w/ free shipping, and because it's still "best in class" overall. The Crucial m4 was a close 2nd choice due to its reliability rep as well, but the write speeds suffer a bit. (Thanks for the SSD comparison benches AT!)GrizzledYoungMan - Monday, May 14, 2012 - link
Trust me, you'll be really glad you didn't go for the M4. After my old Vertex (first gen, Indilinx! haha) was dying, I replaced it first with a 128gb M4, largely because of its reputation for reliability.Man, was that a mistake. I was on FW 0309 (which had been issued in a hurry to correct a BSOD issue created by the last FW update) and the thing had major issues. Stuttering, hanging, inability to resume from sleep and so on. The Windows event manager was lit up red like a Christmas tree, and this was with a fresh install on a Lynnfield based machine that worked beautifully with my old Vertex.
Turns out the thing had major, unacknowledged (outside of some hard to find forum posts) compatibility issues with Intel's RST an a number of power management features in Windows 7. I guess Intel and Microsoft are really obscure little companies whose products don't merit compatibility testing? Beats me.
zcat - Monday, May 14, 2012 - link
Hah. I'm also replacing a 1st gen OCZ Vertex. :)SMART says it has only 39% of its (writable) life left after about 2.5 years of use:
209 Remaining_Lifetime_Perc 0x0000 --- --- --- Old_age Offline - 39
SlyNine - Monday, May 14, 2012 - link
Put a M4 in my GF's computer and she has had no problems at all. Friend also has one and again no problems.Just because you have had issues doesn't mean someone else will, if they were that easily repeatable they would be easy to fix.
ShieTar - Monday, May 14, 2012 - link
Why does this article contain two tables with the exactly same lable "NewEgg Price Comparison (5/5/2012)" but very different numbers?Kristian Vättö - Monday, May 14, 2012 - link
If you loaded the article right when it went live, it's possible that you saw two different tables. I updated the prices just before I posted this, but forgot to copy-paste the new table to the final page. Both tables should be up to date now, though.CyberAngel_777 - Thursday, May 17, 2012 - link
I wish you had two different dates and prices just show the price fluctuationsand then later with updated data also updated prices
every drive should have thickness data recorded
thank you!
Coroder - Monday, May 14, 2012 - link
Does the vertex 4 drive in this test use the new firmware that was made available? It increases the speed in most tests by quite a bit.XJDHDR - Monday, May 14, 2012 - link
I believe they are waiting for the firmware to come out of beta testing.jwilliams4200 - Monday, May 14, 2012 - link
Lite-On owns Plextor, so it is misleading to say that Plextor buys the SSDs from Lite-On. It would be more accurate to say that the Plextor / Lite-On relationship is like the Crucial / Micron relationship.koinkoin - Monday, May 14, 2012 - link
At my company we use McAfee Endpoint Encryption, how does this affect the performance of the drives and how much does this change over time?I went for a Plextor M3 256GB and seem to be working fine as for now, but only got it now for 2weeks.
rlhunts - Monday, May 14, 2012 - link
My company also uses the McAfee Endpoint Encryption. I upgraded the WD 250 GB 7200 RPM drive to a Samsung 830 258 GB SSD in my Dell laptop (I5 2540M processor) a couple months ago. It definitely helps, especially with boot time, but McAfee EE adds a lot of processor overhead, which a drive can't help with. It's still slow as a slug compared to the same machine and software configuration without McAfee with a traditional 7200 rpm hard disk.Movieman420 - Monday, May 14, 2012 - link
In all fairness the Vertex 4 needs to be re-tested with the latest 1.4RC fw...or wait til it's final then flash up.Kristian Vättö - Monday, May 14, 2012 - link
Anand is waiting for the final version before publishing any data. Apparently there are a few things that OCZ promised will be featured in the final version, but I'm not sure what.SlyNine - Monday, May 14, 2012 - link
I plan on running drives in Raid 0. Trim doesn't do anything there so its important how a drive deals with a no trim situation.Coroder - Monday, May 14, 2012 - link
I've been running first gen vertex drives in raid 0 since they were released and since installing the firmware on them that enabled garbage collection they have been fine and they are routinely at 90% capacity.Mars999 - Monday, May 14, 2012 - link
What firmwares are you running on these drives?Thanks!
Kristian Vättö - Tuesday, May 15, 2012 - link
Corsair has not posted any firmware updates so in this case it's the stock firmware (1.00 I think). I'm not 100% sure what firmwares the other drives are running but you may be able to find that info in the review of a certain SSD. However, some drives were tested quite a while ago so they may not be running the latest firmware. We try to update our charts if there are significant gains from a newer FW, though.kstan12 - Monday, May 14, 2012 - link
I would like to see my Mushkin Enhanced Chronos Deluxe 240GB reviewed & compared please!!biodtl - Tuesday, May 15, 2012 - link
I second that notion! The Deluxe is a real performer and could put up some competition here.Kristian Vättö - Tuesday, May 15, 2012 - link
Just sent an email to Mushkin :-)kstan12 - Tuesday, May 15, 2012 - link
you rock!! hope to see it on here soon! thanks!!MarcHFR - Tuesday, May 15, 2012 - link
Hi,You wrote
"Neither Plextor or Corsair manufactures the drives they sell. Both companies buy their drives from Lite-On and if you take a look at Lite-On's M3S SSD, it seems a lot like Corsair's Performance Series Pro and Plextor's SSDs, doesn't it? That's why the hardware of the Performance Pro and M3 is the same—they come from the same plant."
In fact Plextor is a part of PLDS, a joint venture between Lite-On IT and Philips.
Arbie - Tuesday, May 15, 2012 - link
What about write amplification and the longevity of the drive? I thought that would be a hot topic given the aggressive garbage collection! AT is about the only site where such a thing would even be addressed, so I'm sorry there's no discussion here.