Samsung Acknowledges and Fixes Bug On 840 EVO SSDs

Lucas123 writes: Samsung has issued a firmware fix for a bug on its popular 840 EVO triple-level cell SSD. The bug apparently slows read performance tremendously for any data more than a month old that has not been moved around on the NAND. Samsung said in a statement that the read problems occurred on its 2.5-in 840 EVO SSDs and 840 EVO mSATA drives because of an error in the flash management software algorithm. Some users on technical blog sites, such as Overclock.net, say the problem extends beyond the EVO line. They also questioned whether the firmware upgrade was a true fix or if it just covers up the bug by moving data around the SSD.

Comments

What about the Dell knockoffs?

[CaptainStumpy]

So are they going to fix the Samsung SM841 SSD or are we just screwed because we bought Dell?

Wonder what brand is best now... Intel?

[mlts]

This gets me wondering what brand of SSDs is best these days. I've read a lot of good about Intel brand drives, but wonder what is decent these days.

Re:Wonder what brand is best now... Intel?

[tibit]

I've had plenty of success with Crucial and their M500 and M550.

Re:Wonder what brand is best now... Intel?

[ArcadeMan]

I'd rather go with stable than EXTREME, so I go with Intel. It might not be the fastest around, but we rarely hear about Intel SSD problems.

Available soon below my post, someone with a story about failed Intel SSDs.

Re:Wonder what brand is best now... Intel?

[fuzzyfuzzyfungus]

They had that one a while back where the drive would mysteriously decide that it had a capacity of 8MB, though that has been quashed for some time.

The tricky thing (and I'm not actually certain where they stand on this now) is that Intel's initial reputation was founded on the superior performance and reliability of the in-house controller design that they used in their x-18 and x-25, especially dramatic back when there was some utter garbage floating around (JMicron controllers, OCZ living up to their reputation) and the safe options were comparatively slow and extremely expensive.

Then, for some reason, they just sat and stagnated on that controller design for several generations, and eventually shipped a Marvell controller in order to have something with SATA 6Gb support. Since then, they've released some Sandforce based stuff, and some of their own; but it isn't as clear exactly what "Intel" on the label means anymore.

Re:Wonder what brand is best now... Intel?

[MetalliQaZ]

I really like my OCZ SSD.

Re:Wonder what brand is best now... Intel?

[tlhIngan]

I'd rather go with stable than EXTREME, so I go with Intel. It might not be the fastest around, but we rarely hear about Intel SSD problems.

For SATA SSDs, there's no more extreme. All modern SSDs saturate a SATA-3 bus. If you wonder why they all benchmark at 540MB/sec reads and writes, that's why - SATA is the bottleneck, not the SSD.

PCIe SSDs are where the "extreme" ones go, and even the most conservative ones are pretty damn fast - the old MacBook Air's SSD clocks in at 750MB/sec read and write. I think the newer ones can hit 1GB/'sec now easy.

As for what to buy, well, Samsung, Intel and Toshiba are the general safe bets. Even with this bug, Samsung is still stable, just slow.

Intel's got a history of failure as well, but they seem to have gotten beyond it, and while they're not stunners, they generally are solid.

Toshiba's on the slower end of the scale, but Apple uses them, so they can't be TOO bad.

And yes, I say Apple, but you can see what Dell uses as well. The big OEMs that ship lots of units will generally pick ones that give the least warranty and support issues and thus are more conservative. Plus, recalls are expensive.

If you want to follow someone - pick Apple. Given the way news coverage is, if there's a problem with someone somewhere and their SSD in their Apple product, the whole world would know in a nanosecond. Someone as heavily scrutitinized as Apple (where even one failure in millions of computers sold would probably bring about SSD-gate) means if there is a real problem, you'd already know.

Re:Wonder what brand is best now... Intel?

[LinuxIsGarbage]

SSDs will saturate SATA-3 for sequential reads and writes. My Crucial M550 gets 500MB/s vs 150MB/s on my Western Digital. Over a 3 fold improvement!

However where SSDs really shine is random reads and writes. This is why SSD's make PC's more responsive. My Crucial gets 26MB/s vs. 0.66MB/s on the WD. Almost 40 fold improvement, but not near saturating SATA-3. So there is still improvements to be made on random read/write performance.

More and more I see PC's slowing to a grind, and it's due to the Hard drive thrashing crazily at less than 1MB/s! Put an SSD in (any SSD) and it speeds right up.

Re:Wonder what brand is best now... Intel?

[LinuxIsGarbage]

I've had plenty of success with Crucial and their M500 and M550.

I recently had a new PC built. The shop was offering Kingston V300. A quick search showed that the Sandforce controller runs like crap with incompressible data, and Kingston changed suppliers after media did all their benchmarks, so all new units performed like crap.

So searching for alternatives Samsung 840 EVO was a top pick. I was very close to pulling the trigger when I saw all these dire warnings about performance deteriorating with "old data". I knew a firmware fix was pending, but I didn't trust the fix so I kept looking.

I ended up with a Crucial M550 from Amazon (good price). I hope I don't regret it.

In the research I've done, Intel has an excellent reliability record, and OCZ has amazing performance, but questionable reliability.

Re:Wonder what brand is best now... Intel?

[Anonymous Coward]

It's unfair to single out OCZ for the JMicron controller problem.
At the time, most vendors did not provide firmware that properly managed free blocks - once you filled the drive up, the performance nosedived.
OCZ was the first non-Intel vendor to fix the problem, with prompting by Anandtech.com.
For a time, OCZ was the ONLY vendor that you could get a reasonably priced SSD from.
I personally have not had any problem with OCZ SSDs, and I have been buying them since the JMicron days. They developed a reputation for poor quality, and then released their Vector 150 line, which they claimed was designed to correct that reputation. This seems to be a stellar SSD by all accounts.
After some financial problems (improperly accounting for rebates), they were saved from bankruptsy by being bought by Toshiba.
I switched to the Samsung 840 EVO series, but was having misgivings about support being provided by a 3rd party in Florida (only, I think), that was reported to be less than responsive. Now, this deteriorating performance with data that is one month "old".
I wish I had stuck with OCZ.

Re:Wonder what brand is best now... Intel?

[Luckyo]

It's completely fair to single out OCZ as related to JMicron controller and sandforce controller issues. All of these controllers had a large set of options which could be tuned by manufacturer. OCZ was known for tuning for pure performance, disabling all reliability related functions in controllers whenever it could give them even a little bit of more performance on benchmarks.

As a result they typically crushed others on benchmarks but drives had absolutely atrocious reliability. Business model was apparently to sell as much as possible so that massive warranty claims could be accounted for through growth. This business model failed and they went bankrupt.

Re:Wonder what brand is best now... Intel?

[drinkypoo]

More and more I see PC's slowing to a grind, and it's due to the Hard drive thrashing crazily at less than 1MB/s! Put an SSD in (any SSD) and it speeds right up.

while that's true, at least half the time they're swapping. Put in enough RAM and disable swap and they speed right up, too. Maybe not as much as SSD, but more RAM benefits you whether you have SSD or not.

A read disturb problem

[TFoo]

This is almost certainly a firmware bug with their read disturb compensation. At least they're owning up to it - but wow.

DOS version?

[CurryCamel]

"Dos version for MAC, Linux users ... Will be released on end of Oct."
http://www.samsung.com/global/business/semiconductor/samsungssd/downloads.html?CID=AFL-hq-mul-0813-11000279/
Let me guess - the source for that firmware patch is stored on a Samsung EVO 840 disk?

Re:DOS version?

[tibit]

A modern Intel Mac will boot into FreeDOS, no problem. It's more like a PC without the BIOS Setup, and supports booting straight into OS X :)

Re:DOS version?

[LinuxIsGarbage]

The same way that you're suppose to run their DOS executable on a Linux?

The same way you run it on Windows?

64 bit Windows will not even pretend to run 16 bit DOS/Windows 3.1 applications. 32 bit Windows NT (/XP/Vista/7/8) will, but it's in an emulator so it can't access the hardware.

You need a freeDOS bootdisk. You can make it boot from CD or USB since most modern PC's don't have floppy drives.

A DOS executable is almost preferable since it doesn't require a proprietary OS.

Re:DOS version?

[Just Some Guy]

The current firmware update ships as a bootable ISO. Burn it to a CD/DVD (or a flash drive if you can work it out), hold down "option" at boot, and you'll be looking at a DOS prompt in no time. I verified this two days ago when I misread the firmware version on the website and downloaded an updater for the version I already had.

Anandtech had a WAY better article

[ashpool7]

More technical detail as to what is going on.

http://www.anandtech.com/show/...

Classic Samsung...

[Andy Dodd]

Couldn't write a proper wear levelling algorithm if their life depended on it.

First the MAG4FA/KYL00M/VYL00M data corruption bug that affected the Galaxy Nexus - https://android.googlesource.c...

Then (actually BEFORE it, Google found it during Galaxy Nexus development but Samsung kept it hush-hush - but it became a public issue much later) - the infamous Samsung Superbrick fiasco (If you fired a secure erase command at the chip, it had a chance of permanently corrupting the wear leveller data to the point where the chip's onboard controller would crash until you power cycled it any time you accessed that region of flash). - https://git.kernel.org/cgit/li...

Then pre-release 840 PRO devices suffer from the SAME DAMN BUG SAMSUNG HAD BEEN AWARE OF FOR OVER A YEAR - http://www.anandtech.com/show/... - While this only affected review devices, the fact that this was a known bug since before the release of the Galaxy Nexus (a year earlier) is inexcusable.

Then there was the Galaxy S3 "Sudden Death Syndrome" issue in late 2013... - https://github.com/omnirom/and...

Then there were a few other issues - http://wiki.cyanogenmod.org/w/...

Now this...

Re:Windows only; NTFS only

[KozmoStevnNaut]

There will be a bootable DOS disk image for Mac and Linux users. It's supposed to be released late October, according to the download page:

https://www.samsung.com/global...

VNAND

[sonicmerlin]

VNAND run at current 1X node levels should provide 32x the capacity for similar cost. Instead Samsung is using their tech to release 4X node level SSDs with similar capacity but double the cost of 1X node level 2D NAND. When the heck are we going to have some competitors come in with their own VNAND tech and bottom out the SSD market? They should even be able to achieve greater cost per byte effectiveness than HDDs.

misalignment also results in crappy performance

[Browzer]

In my case, based on hdparm -t on xubuntu and centos, the difference between a properly aligned Samsung EVO and an improperly aligned Samsung EVO is 510 MB/sec and 182 MB/sec respectively

http://cillian.wordpress.com/2... has some good info on setting up Samsung EVO properly on linux