review some 840 evos still vulnerable to read speed slowdowns

Some 840 EVOs still vulnerable to read speed slowdowns

Last year, users discovered a problem with Samsung’s 840 EVO SSD that caused dramatic slowdowns when reading older data. Samsung attributed the issue to an algorithmic error in the management routine that tracks the status of cells over time. A firmware fix and accompanying Performance Restoration utility were issued in October, and they seemed to do the trick. However, new evidence suggests that the problem persists.

A couple of TR readers (thanks Horia and Richard) pointed me to recent entries in the original thread complaining of slow read performance. Those reports come from drives running the supposedly fixed EXT0CB6Q firmware, and they prompted me to test an EVO I’ve been saving for just such an occasion. The results don’t bode well for the TLC drive.

When the initial fix was issued, I patched our 840 EVO 250GB SSD and then filled it with a mix of movies, MP3s, images, and other files. That drive spent more than three months on the shelf before being called up for a round of read speed tests. Here’s HD Tach’s assessment:

35MB/s on a modern SSD? Yeah, that ain’t right.

Next, I ran SSD Read Speed Tester, which tabulates read speeds based on the age of the files. This benchmark indicates that everything on the drive was almost 15 weeks old when the test was run.

The average here is 54MB/s, with some files reading at well over 100MB/s. That’s faster than in HD Tach, but it’s still a far cry from what the drive can do with fresh data.

Slow read speeds didn’t just afflict targeted benchmarks, either. Transfer rates were extremely slow when I copied the EVO’s contents to a secondary SSD in the same system. I then formatted the EVO, loaded it with the original data, and ran the same tests again.

That’s more like it. The drive averages 430MB/s in HD Tach, a 12X increase. The EVO is about 10X faster according to SSD Read Speed Tester, which reports a 529MB/s average.

SSDs typically aren’t left unpowered for months at a time, so it’s possible that hiatus contributed to the slow read speeds exhibited by our sample. However, the other recent reports of read slowdowns come from drives that have been in service since the patch was applied. The issue doesn’t appear to be confined to unused drives.

Bruno, our resident coder, has an 840 EVO in his personal machine. He agreed to run a few tests for me, and his drive isn’t substantially slower to access older files. But NTFS compression was recently enabled for much of the drive’s contents, so I’m hesitant to draw any conclusions based on those results. Compressing the data should effectively refresh the contents of the NAND cells even if there’s no change in the stated age of the files.

We’ve notified Samsung of our findings and are awaiting an official comment from the company. In the meantime, I’m curious if any TR readers have experienced similar slowdowns. Have any of you noticed any read speed issues with patched 840 EVO SSDs?

0 responses to “Some 840 EVOs still vulnerable to read speed slowdowns

  1. My 840 Evo 120GB seems like is not that slow. I use it as system drive and the slower speeds are connected to smaller files rather than older. SSD Read Speed Tester shows an average of 380MB/s. I updated the firmware immediately after the installation of ssd and did a new clean windows installation.

  2. I have never been a fan of the EVO drives and from what I’ve been seeing there’s no reason to buy them now. From iffy data retention to this slowdown problem there’s other drives not made by Samsung in the same price and performance range of the EVO drives now that do not have these issue.

  3. I have and 840 EVO in a system I built about a year ago. Due to a move the PC did not get hooked up for about 3 months. When I got it going again I was having an extremely slow time of booting up the PC but attributed it to spyware, viruses, etc. (By extremely slow I mean 20 min. boot up time and ~45 min getting the keyboard driver loaded!!) Shut downs are sometimes very speedy but most times the system hangs and doesn’t shut down so I have to hard kill it otherwise it recovers and comes back to the Windows sign in screen.

    Found out about the Restoration fix and thought this might be the issue so applied it, but I am still experiencing the same terrible boot up times. Does the problem affect files on boot up or just programs that are loaded onto the PC after it is booted? I’m about to just cut my losses and just reformat the drive.

    One other thing; I read below that some were asking if defragging a SSD would help with this issue. When researching hard drives to put into my new PC I read that SSDs should not be defragged due to the fact that it would greatly reduce the lifespan of the SSD since it is not a mechanical drive.


  4. Potential Temp fix:

    For windows users, use a program called diskfresh. it rewrites to every sector on your ssd

  5. With this issue still in the air and the way samsung is handling this… shouldn’t this have consequences for the review in the 850EVO? As the tech in that drive is also new one can similarly expect a higher chance of problems – so samsung should provide a better guarantee of fixing this.

  6. Same here, like a month ago I notice a slowdown in boot times. The Windows 7 logo COME TOGETHER?! It’s much slower for no reason :(.

    Is there some Samsung blog where they would answer questions about this?

  7. A large chunk of data on my 256Gb 840 Evo is 252 days old (how long ago I installed it) and the vast majority of it is still at 400Mb/s with some down to 100Mb/s according to SSD Read Speed Tester, the drive has lived at 80%+ full its whole life; 252 days is the only data point where data on it scored at under 350Mb/s. I do have a flash disk defrag that runs weekly (though it rarely runs more than 5sec as most of the data is static) and I did run a tool that re-wrote all the data two days before the firmware fix came out which I haven’t installed/run yet (but interestingly that seemed not to reset the age of the data)

    I suspect the firmware just rewrote data, and lowered some retry threshold somewhere, but ultimately the bug wasn’t fixed as it’s a ‘feature’ of the TLC and will recur until to the drive’s death, but pfft 3000p/e cycles rated still means my drive will “officially” run out of writable flash well after the human race has extinguished itself…

  8. I’m seeing a similar slow-down on my 500GB 840 EVO. It’s my boot drive with Windows7/64 and a game or two that I play frequently – only about 90GB loaded. Through most of the 90GB now (I ran the firmware update and performance restoration right after installation) HD Tach shows about 50MB/s, and I can anecdotally tell slowness in Windows boot time and game loading. I tried running the restoration software again and it didn’t seem to do anything. Suggestions? Per the article, would NTFS compression help? Or per some of the comments, a defrag every month or two? This isn’t what I wanted from my purchase of this (and I have another waiting to put in my wife’s laptop computer, and don’t want to see similar slow-downs with her machine).

  9. Just adding a quick note on my 250GB Samsung 840 (non-EVO) that I had mothballed with new data written on the date 2014-09-23, meaning it had not been used and not powered for 132 days. I didn’t plan to do this but just threw some files on there when I first found out of the problems that 840 and 840 EVO appeared to have. At the time my 840 drive that had been playing second fiddle to another did have areas where read speed was about 150 MB/s so at least it appeared partly affected. (a very similar pattern also appeared on my main drive of a different make, but I don’t have enough data to warrant putting it out there)

    Anyway, today after powering it back on after 132 days of disuse there were no signs of a lowered read speed. Tested first with a simple practical copy test of video files from the drive to a RAMdisk. Then with SSDReadSpeedTester. I’m assuming that at least with these older non-EVO 840’s it takes quite a bit of time to the problem to manifest, so if you have one and if there is no improved firmware coming out down the line; you should be fine running the program called DiskFreshHome about twice a year or so.

  10. True, if you run diskfresh once a month with your 3000-cycle TLC NAND, you’ll wear your SSD out in 250 years

    OH NOES!

  11. Right. There is a remedy.

    But the emo-butthurt overflowing around here on this issue is comedy gold,
    and very reflective of modern society.

    Grow up…

  12. have you ever formatted that drive before since you bought it? I notice it sometimes with my 840 Pro’s but i just format them and start a fresh and wallah fixed 🙂 just an idea on how to get some of your performance back maybe.

  13. Check this:


  14. Unfortunately, after I updated the bios on my 840 EVO a few months ago, I was hoping that it was back to full speed, and it did seem faster than it had been for awhile.

    About a month or 6 weeks ago, I began to think it was getting slower again. I play World of Warcraft at night, and when my SSD is going full speed, I am normally the first one to zone into a new zone with a group. But anymore, half or more of the group is already there when I get there. This is probably no more than a 5 to 7 second difference, but it is noticeable.

    Yesterday I rebooted, and Chrome took a long time to open, and its normally almost instant. So something is slower, I just am not sure yet what it is.

  15. Both my 250GB and 1TB 840 evos are suffering from this issue after applying the fix to them.
    I will be anxiously awaiting Samsung’s response.

  16. I too was in the swing that an SSD should never be defragged. But i use MyDefrag and noticed they had an SSD specific defrag method. as they put it:

    “But fragmented files need extra processing time inside Windows, not noticeable on mechanical harddisks but very significant on fast flash memory disks. Even more important is free space optimization. Flash memory is written in large blocks, and if free space is fragmented then Windows has to (read and) write much more data than the size of the file. This takes time, which translates into lower speed.”

    But it goes on to say that achieving these goals is balanced with the notion of using as few cycles as possible.

  17. I’m really intrigued by the posts about how defragging an SSD can improve performance. I thought that degrag had no impact on SSDs, or was bad for them, whatever, I never did it. I also recall some studies in the past about defrag actually having little impact even on HDDs (I recall defragging a lot in the Win98/XP days).

    In any case, this whole EVO 840 has now raised the question of whether SSDs should be defragged as a basic maintenance operation to sustain performance. It would really be helpful to get some definitive answer to this.

  18. Just ran HDTach on my Samsung 500GB EVO and saw some notable dips in performance. Realized I never ran the “performance restoration tool” on that drive and ran it just now. Ran HDTach again and all is well.

    However, my 1TB EVO in the same computer is showing dips down to the 150-200MB/sec range in the first and last “thirds” of the graph. I ran the performance restoration tool on that drive back when it first came out, so I can’t run it again. It is running the latest firmware. RAPID mode has no effect, manually TRIMming the drive from Samsung Magician and in Windows has no effect either.

    Its not nearly as bad as what Geoff is showing on his graphs, but definitely not the near flat results I’m seeing on the 500GB drive.

  19. Defrag not only won’t help a SSD but will wear it down.
    Call me “old fashion” and downgrade me as much as you like, but i prefer a HDD in special because for HDDs there’s a FREE utility to refresh ALL HDD sectors (it doesn’t defrag).

  20. Really, that depends on how full the SSD is. At <40% full, MyDefrag will most likely go though less writes and definitely take less time since it would only rewrite portions of the SSD actually holding data.

    Unless I’m mistaken, DiskFresh will basically rewrite the entire drive, going though the drive’s capacity worth of GBs in writes guaranteed, writing zeros (and not trimming) the free space. OTOH, on a full drive, since MyDefrag will rewrite many files twice, it would definitely go through more writes than DiskFresh.

  21. “If they make it right really quick then great”

    I’d say they missed that boat. This article and a large amount of evidence appears to point to the fact that a sev 1 was found in September in a widely used product and still isn’t fixed. MTTR fail.

  22. I’ve got a Samsung MZMTD 128G which came with my HP x2 Split, got it just over a year ago, and just for fun I just ran HD Tune and got a really ugly graph with scores ranging from 275Mb/sec to ending at just over 25Mb/sec with lots of peaks and valleys in-between.

    Hmmm. Now this is what my two EVO 840 drive graphs in my two home builds looked like before I did the Samsung update patch some months back. I’ve periodically run HD Tune since and get high consistent numbers, no problems. I have noticed no problems opening any old files. So the fact that this occurs on the non-EVO Samsung laptop SSD drive is interesting to me. Is this a broader Samsung problem? A broader SSD problem?

  23. You’re better off running a free tool called diskfresh as a scheduled task once a month.

  24. Interesting theory @


  25. My 250gb 840 Evo was “fixed” on the day that the performance restoration update was released.

    Since then I’ve secure erased and restored at least twice to recover the drop in read speed which starts after about 4 or 5 weeks of daily use.

    Samsung had better come up with some answers and a proper fix soon. I’ve bought and installed dozens of these SSD for my customers, but have now switched back to using Crucial SSDs. I doubt I’ll be going back to Samsung any time soon.

  26. Having owned a pioneer, nope, not even a chance (come on dude pioneer hasn’t made any since 2009). As far as Panasonic goes, their TX-P65VT65 is a fine TV if you have a dark room, my father has one. When compared to my PN64F8500 it might even edge it out in a dark room situation but the Samsung is far better in a lighter lit room (plus no plasma buzz). Besides, Panasonic doesn’t make plasmas anymore and even if the Samsung is ever so slightly behind in IQ it still blows the doors off of any LCD/LED in IQ.

    And BTW, Sony’s don’t get any better ratings than the Samsungs in LCD/LED either. In fact they are usually virtually tied with the nod being given to the Samsung as they are usually 20+% cheaper.

  27. 840 EVO 750GB. My results as measured by SSD Read Speed Tester 2.02:

    – before the fix: 317 MB/s
    – after the fix: 541 MB/s
    – now (14 weeks later: 537 MB/s

    So it seems fine to me. My system has been turned on about 12 hours a day since the fix.

  28. Before applying the fix: [url<][/url<] (10-11 months 24/7 usage) After applying the fix: [url<][/url<] One month after applying the fix: [url<][/url<] (15 Nov 2014) 2 1/2 months after applying the fix: [url<][/url<] (27 Jan 2015) I am using this SSD as a system disk in my main rig, 24/7.

  29. Defragmentation doesn’t shuffle data around only if you’re using a bad defragmenter (which most of the popular ones are).

    And please don’t EVEN CONSIDER switching compression on and off to rewrite the drive. When compression is turned on, it creates thousands of fragments in the NTFS file system, and since MFT records can only hold information about a small amount of fragments, this will cause many additional records to be created, and they cannot be removed without formatting the drive. All of this will only reduce your I/O speed more and consume extra space in the end. Just use MyDefrag and do some good while rewriting those files.

  30. Absolutely yes, and it has been done successfully by others many times (see [url<][/url<] ). And contrary to popular opinion, defragmenting a SSD won't harm it any more than what it has already been through to get those files written to it in the first place. However, most defragmenters on the market today are trash, ranging from not even doing a good job at removing fragments, to fragmenting free space so badly as to be addictive, to removing fragments but ignoring disk layout, to good defragmenters at do it all. You can read more on that here: [url<][/url<] For this job, rewriting everything on the disk, you want to use MyDefrag running the System Disk Monthly script (on disks with Windows installed) or Data Disk Monthly script (on storage disks). Please note that "monthly" has nothing to do with a defrag schedule; rather it is just the name of MyDefrag's most thorough defragmenting script. This script is just about guaranteed to rewrite everything on the disk, as it will sort all files by name/access, meaning that the change in size of just one file will cause everything to need shuffled to fit (which is exactly what we want to refresh a SSD). MyDefrag is free, and you can download it here: [url<][/url<]

  31. SSD Read Speed Tester is a program I wrote when this issue first surfaced. I decided to write it when early on, when people started suspecting that the EVO’s retrieval speeds decreased with file last-write age.

    SSD Read Speed Tester is publicly available (you don’t have to sign in to download it), and its unofficial homepage is this thread on [url<][/url<]

  32. Seeing as the SanDisk Ultra II employs TLC with a SLC cache like the 840 EVO, I wonder if it suffers the same problems. Although, the Ultra II uses a Marvell controller. Might make an interesting comparison piece being able to isolate the problem to NAND or controller.

  33. Yeah. I still wonder if it’s flash data retention problems rearing their ugly head. Someone should try leaving it off for 4-6 months and check both speeds and checksums.

  34. Where’s Krogoth threatening law suits and giving doomsday prophecies for Samsung?

    Surely this ranks higher on the “impact to user” scale than the GTX 970 memorygate non-event

  35. Friends don’t let friends TLC.

    I thought we all knew this. Is it really any surprise that they have flaws and bugs and problems? Or that Samsung is busy covering up the evidence in the now-traditional way of the corporation? By ignoring it and shrugging and grinning and constantly releasing newly minted branding with new numbers and ever-“evolving” software that supposedly fixes things that were supposedly not broken last gen until the next gen shows up?

    Just stick with MLC and honestly stay away from Samsung. I used to love them and in fact I still love my 830 Pro, but man, just stay away from Samsung now. They’re overpriced and they tend to support their hardware now about as well as OCZ.

    It’s a shame. They used to be the most reliable company, but they got so big and got so hung up on saving a dollar with TLC they lost their way.

  36. Yeah was about to point that out. I have a 840 and benchmarked performance is all over the place. Doesn’t seem to matter much in practice, though, but I don’t want it getting worse over time.

  37. I agree with the plasma theory, but Panasonic (and even Pioneer) beats them in that area too.

  38. My results before the fix: [url<][/url<] After the fix: [url<][/url<] And now, after 3 months of normal use: [url<][/url<] It may not work if you shelve your drive, but under normal usage it seems like the fix is doing its job. Edit: This is my laptop running Windows 8.1. RAPID was off for the tests, as was my AV software.

  39. I’ve got three words for you Samsung: Class Action Lawsuit

    You better make good on this one or I won’t be buying any samsung products anytime soon.

  40. Yeah, I realize the consumers typically get squat and the only winners are the lawyers. But at least it (hopefully) acts as a deterrent against future bad behavior.

    IIRC the most I ever got out of a class action settlement was a check for something like $10, in some sort of dodgy credit card interest rate or billing practices nonsense; or maybe it was related to dodgy mortgage escrow account handling. I forget, it was years ago, but it was a “banks behaving badly” sort of thing (go figure).

    In the Iomega Zip drive “click o’ death” class action I received a coupon good for a free box of Zip media. Gee thanks a lot, now I can run out and get some free media for my dead Zip drives! They probably had a warehouse full of media nobody wanted sitting around at that point, so I’ll bet the coupons effectively cost Iomega nothing beyond the cost of mailing them out.

  41. Is “SSD Read Speed Tester” an in-house tool? Or is it publicly available. Google search just brings up message boards mentioning it.

  42. Class actions are not all they are cracked up to be, users get next to nothing and the lawyers get the big chunk of change. They advertise “Up to” speeds, not guaranteed speeds and since the drives are still functional they are still considered working. A class action maybe hard in this case.

  43. The bootable ISO worked fine for me (on an AMD chipset). Just use a burned CD (via external optical drive if necessary).

  44. Heh. Over the years I’ve cycled through hard drive vendors as each one released a dodgy series of drives. The sequence went IBM (now split into HGST/Toshiba, with the HGST bit owned by WD) -> Maxtor (now part of Seagate) -> Seagate -> Western Digital -> currently back to HGST/Toshiba as my “go to” brand, but I’ll also buy WD if the price is right.

  45. That’s gonna piss off a lot of technically savvy people and seems like a good recipe for a class-action lawsuit to me.

  46. Yeah, “Screw Samsung until my next favorite SSD vendor screws up” is more realistic.. 🙂

  47. I was thinking along the same lines. The article points out that this drive was left off for most fo that time. I wonder if they can find another drive that shows the same read speed problems and–before reformatting–just let it sit powered up for a while. See if it fixes itself silently.

  48. I don’t seem to have a problem with my 1TB 840 EVO :

    [url<][/url<] Sure, there are some "slow files", but as you can see, they are usually the small ones, where the speed is rather inprecise.

  49. Na, like my plasma’s way to much. The top LCD’s from Sony and Panasonic (or really anyone else) still pale in comparison for picture quality compared to plasma. If they are ever able to fix the banding on colour gradients, picture smoothness, in LCD I will look at one but still looking at a good plasma vs the best LCD side by side the plasma still wins out in IQ.


  50. Could be, however I think at this point Samsung is taking the silent “if we ignore it, they will go away” approach.

  51. While I understand that sentiment, saying it was a “fake” firmware I feel is a bit of an assumption. Again, I’m going to wait it out and see what they have to say or what others uncover.

    If they did in fact release a faked fix in hopes that no one would notice then I’ll be the first one to light a match and douce them in gasoline.

  52. Just spend the extra money and buy some Sony goodness next time :).
    Ok, or maybe some Panasonic.

  53. did a reinstall after updating the firmware on 840 EVO 250Gb. just found out about the bench a few weeks ago and showed a slight performance dip after 6 weeks. but hardly enough and too early to really tell so keeping an eye on it.

    but i am slightly disappointed as a first time ssd user, thinking reliability of ssds is nothing to be worried about. btw, thanks for that endurance testing TR! lol.

  54. ‘Never buy storage when you don’t need it.’

    Unless your crystal ball is showing a factory fire that disrupts supply. 😉

  55. Just avoid TLC. I didn’t have time to properly test my 840 last round, but just spent an hour to collect some data. This is definitely not 840 EVO exclusive problem. It might as well be common to all TLC nands.

    My post to the original overclock net thread with HD Tach results.

  56. I had a feeling the fix may not be effective. It’s like it was released too quickly. I appreciate the prompt response from Samsung, of course, but it’s only after months of testing and evaluation do we know for sure if the fix works or not.

    Edit – I applied the fix just today. I’m the last person on earth to do it, I know. Still, I guess it’s better than nothing. I hope Samsung releases another attempt to fix this issue soon enough.

  57. Yeah. And Samsung handled it by releasing a fake firmware fix together with utility that didn’t do more than rewrite all data. So basically, they have already handled it once — with more lies.

    Yes, they all have had issues, which they’ve owned up to by replacing SSD’s. Eg. OCZ replaced two of my bad budget SSDs with better ones (this before they were bought by Toshiba).

    Samsung? Well……

  58. 50MB/s to 100MB/s??? That’s like five years old WD Green or similar… (And quite worse then WD Black or current Green)

    And I would be quite likely to hit because much of system drive consists of large programs like Visual Studio or Sony Vega, where rarely you see modifications to them and sit there for a long time unused.

  59. The last time I did a complete reformat of my 840 Evo was in October of last year. So far I haven’t encountered any serious slowdowns, although I’ve noticed that boot up times went from the usual 11 seconds after a fresh reformat to 12 seconds. I don’t know what exactly causes it but I’ll chalk it up to the usual junk build up in Windows. I haven’t flashed my 840 Evo with Samsung’s patch yet because the flash tool fails to work with AMD’s chipset drivers and I plan to do it on my next reformat. Too lazy to do it these days though. If it ain’t broke, don’t fix it. I did try the DOS based method but the flash drive I tried to make bootable didn’t boot, so… what the heck. And now it turns out the supposed fix doesn’t work either, so not being able to flash my drive isn’t so bad after all, unless the current firmware does nasty things to the drive’s long term health. Oh what the heck. With SSDs dropping in price all the time I’m not gonna get too fussy about it. But I might avoid Samsung on my next SSD purchase and will advise folks to look elsewhere as well.

  60. Random thought: For the people who are still seeing this issue, does the drive tend to be on a lot, or does it spend a lot of time powered down? I wonder if the nature of the firmware “fix” was to add a background refresh of the drive that periodically reads and rewrites all of the data. Even if the drive has been in service, if it spends a significant percentage of its time powered down it could affect the ability of the drive to refresh all of the data in a timely manner. (I assume that if the firmware is doing this, it is doing it fairly slowly to avoid additional wear on the flash cells.)

  61. Even if the drive isn’t being used for archival storage, there are going to be a lot of files (e.g. system files) that are written once during OS installation, and then only read after that. I suppose you could argue that Windows Update will eventually rewrite a large percentage of your system files, but that’s not really true either; look at your system folders and you’ll see lots of old files in there that are part of the OS. Game level data isn’t going to change frequently either…

  62. I doubt periodic defrags will help on an ongoing basis. Files that have been sitting around for a while — the very files affected by this bug! — aren’t getting fragmented, so the defrag will tend to leave them alone.

  63. The fact that the original fix didn’t work (at least for some people) means Samsung does not completely understand the nature and scope of the problem.

  64. While it may be a flawed harwdare design (I’m not convinced of that yet), even if that’s the case a lot can be compensated for in firmware. Worst case, they release another firmware update which periodically scans cells in the background and rewrites them when they start to show read performance degradation. An extra write cycle every few months isn’t going to affect overall endurance of the drive in a meaningful way, though it may mean that a drive that hasn’t been powered up for a long time may be a bit slower initially, when it is first used again.

    Edit: Or maybe they are already doing this, and something is preventing it from working in some cases.

  65. Even if they fix it for good *this* time, it sure would’ve been a heckuva lot better if they’d fixed it properly the first time. I recently bought several 840/850 EVOs, on the assumption that this issue was fixed. D’oh!

    Even with their async NAND bait-and-switch BS, at least Kingston drives maintain their (somewhat lackluster) performance over time!

  66. It’s his daily use laptop. I don’t think he rewrote the entire drive just for this test.

  67. defrag doesn’t necessarily “shuffle around” data, unless that data is, in fact, fragmented. if you run defrag regularly, you’ll find that you can accomplish a full defrag in 5 to 15 mins.

    Per the article, switching compression should work, I would imagine that encrypting and unenctypting would work for the same reason…

    That said, just don’t use your SSD for archival. done.

  68. So the real question is whether this is something inherent to TLC in general, or just planar TLC , Samsung TLC, or this range of drives (NAND+controller) in particular. Sandisk finally put out a TLC drive, although I don’t think it is as popular now as the 840 EVO was, so it’s something that could be watched for there.

    Hopefully 3D V-NAND or just the new controllers take care of this, and it’s not something inherent to the nature of TLC that won’t change – the additional voltage states for example. The 850 EVO is a heck of an SSD now that the street prices have brought it down to a reasonable price as opposed to the MSRP. I have a suspicion there is something they’ve done to extend the endurance of at least planar TLC which is causing this, so hopefully 3D TLC won’t have any such issues…although they do rate it higher, so maybe they used the same sort of trick.

  69. The Black Friday Sandisk Extreme Pro 480GB deals were pretty sick. I was tempted a few times but I still have 115GB free on a 240GB SSD so…meh. ‘Never buy storage when you don’t need it.’

  70. Would enabling defragmentation on the drive be a quick and dirty solution? Have the data shuffled around periodically. The hit on durability would be worth it if the performance is this bad. Regretting purchase of two 1TB 840 evos.

  71. had a choice this past BF for an half terabyte 840 evo or extreme pro. considering the issues at hand that existed back then, i am glad that i made the right choice with the extreme pro. still great performance after 35% full with 40MB/s 4k QD1 reads and 75MB/s 4k QD1 writes.

  72. Short product life cycles and the need to be the next great thing are becoming more and more of a liability. Samsung (and many other consumer electronics companies) seem more interested in selling people a new device rather than supporting something people have already paid for. At the end of the day, new products are where the revenue is…just don’t burn too many bridges along the way.

    My GF has a 480GB 840 (vanilla) and even she was complaining about the slowness. Yet Samsung seems to be ok with not acknowledging this as a defect for many of it’s customers. Maybe there just isn’t any way to fix it. Consumers are then left with the options of warranty RMA (assuming this will be allowed) or to just periodically rewrite all your old data. For me, this is another bad experience due to lack of support from the vendor and there are too many comparable options to simply accept such indifference.

  73. Firstly, thanks TR for being diligent and proactive enough to run news stories like this.

    I’m running an mSATA 840 EVO 500GB (OS disk) alongside a Plextor LMT-256 (for storage) in my Razer Blade (2013). I don’t have any pre-update benchmarks, as I flashed the drive with the latest firmware as soon as I installed it in the machine. Data on the drive is probably around 6 weeks old, tops. Today’s HD Tach test:

    Samsung: [url<][/url<] Plextor: [url<][/url<] Does the Samsung seem a bit tardy? CrystalDiskMark is absolutely fine (523/505 sequential), although I understand this writes new data before testing? I will make a note to re-run the benchmarks in a couple of months and post back here... although hopefully Sammy sort their sh*t out by then 🙂

  74. I have an 840 Evo 1TB and my take was that as this bug manifests over time it was too early to say if the fix worked until a decent amount of time had passed. So I’m not surprised that this has raised its ugly head again.
    I planned for it as I was going to migrate my Evo to my desktop but decided to sell it after getting a very good price on a Sandisk Extreme Pro. Glad I also got the Samsung at a very good price otherwise I’d be looking at a loss. Sometimes it’s better to bail whilst ahead so to speak.

  75. Since this is a Samsung firmware issue specific to TLC NAND, somebody should test the 850 EVOs.

  76. I see.

    A file transfer to a RAMDrive might be a good benchmark to run, maybe with TeraCopy instead of the default Windows utility. That way you can select what data gets “benchmarked”.

  77. Hey everyone,

    I just registered after seeing this coverage here.

    My 120gb 840 EVO still has the problem.

    I run the firmware update and the performance restoration tool as soon as it comes out. Reading speeds are fine for the past 2 months but after noticing the original thread on becomes active again, I ran couple of benchmarks (filebench, hdtune pro etc..) and found out my reading speeds getting low again.

    I had 470 mb/s avarage reading speed on filebench (75% of the disc is full) right after the restoration and it was kinda around same during the past 2 months.

    Now it’s around 200 mb/s with some old files @ 20-30 mb/s. A level in BF4 takes minutes to load now while it was just a couple of seconds before.

    If Samsung doesn’t promise to fix or compansate the issue, we should be getting ready for a class action lawsuit agains them. This performance on a modern SSD is just unacceptable!!

  78. My personal vanilla 840 has the problem here. I also have one office that uses a non-trivial number of 840EVOs.

    I’ve been keeping an eye on progress since the first mention of the 840EVO’s bug. Samsung claim it doesn’t affect other products, I think I posted my affected vanilla 840 speeds in the original forum thread on the 840EVO issue here.

    Samsung won’t even acknowledge the issue on non-EVO drives, so I’m just running Diskfresh as a monthly scheduled task to fix the issue. I would boycott Samsung on principle, but I haven’t had any reason to buy a Samsung drive in ages – there are better drives for less money.

  79. Im interesred to see how Samsung handles this latest fiasco. I for one though an not going to say “Screw Samsung forever!!!”. My reaction will be completely dependant on how they handle this issue. If they make it right really quick then great… im curious to know the technical reason for this problem and hiw it was overlooked a second time.

    For those of you acting like Samsung is the only one with issues like this… how about OCZ… or every SSD maker before they had sudden power loss protection in place… or Crucials SMART error bug rendering the drive uselsss after X number of hours of power on time.

    Point is every single manufacturer has issues during a products lifetime, its how they handle it that matters.

  80. HD Tach’s read speed test doesn’t use freshly written data, which is why it showed slower read performance in the initial tests I ran. But it doesn’t identify the age of the data at the sample points it uses, and it doesn’t test every block on the drive.

  81. Just a side-note from another user of a rusty-but-trusty T61: IIRC the SATA interface being gimped to SATA1 is an artificial BIOS bug/feature. I retired my T61 last year, but I believe there’s a hacked BIOS out there that allows the the SATA port to run at SATAII speeds.

  82. [quote<]What should distinguish Samsung from others -- or not, we'll see -- is whether they continue to try to fix the problem.[/quote<] My guess is that it is something that cannot be fixed properly even if they wanted to. Flaw in design.

  83. mine seems fine
    [url<][/url<] its only the lowly 256gb one ran that while having a bunch of stuff loaded up (steam/a game im playing atm/about 8 firefox windows/other junk in the taskbar i use - i was too lazy to do a proper test) although the weird thing is i diddnt have the issue in the first place with my launch 840evo with the first firmware that came with it (i never upgraded it until the problem was known, i should have just left it) edit: yes its old data - its actually same stuff on it as before the "fix" was released

  84. Well I’m glad we had to go around and update hundreds of 840 EVOs in production to fix this issue the first time around… BAH.

  85. Still running my beloved 256GB 830 Pro in my main rig at home. I almost swapped it out for the 250GB 840 Evo I picked up a while back but really saw no real reason to. The 830 has been rock solid and the performance has always been great. I’ll probably run that poor drive into the ground before I abandon it.

    But yeah, it makes me wonder how the 840 will handle OS’s? As I’m sure there are gigabytes worth of files that aren’t touched (read/written) on a daily basis. At least I wouldn’t think. Would a user notice his PC getting slower over time? Just makes ya wonder.

  86. .. .with their own firmware/validation, which apparently makes all the difference 🙂

  87. Nope, I’ve recommended the 840 Evo all over the place as well. It even sits in our companies business desktops.

    If you can’t trust that samsung logo and the army of engineers behind it what CAN you trust? Not much it seems.

  88. I agree that you aren’t going to get perfection anywhere, but don’t try and gloss over the seriousness of this issue. It was a fairly big deal when they didn’t properly test for a 100% reproducible issue that causes your shiny new SSD to perform worse than an ancient HDD but knowing then what the issue specifically was they produced an update to resolve it that they apparently didn’t test well enough either.

    I’m a pretty easy going guy as far as bugs go, but this does cast doubt on the validation process in use here (again, this isn’t some rare thing… it always happens). I don’t think I’ve seen an issue as widespread with Crucial, Intel, etc. for instance. At least it’s not data loss I guess, but still.

  89. Ironically, most Intel 5xx SSDs use the same Sandforce controller that annihilated OCZ.

  90. So what are you going to buy? All manufacturers have bugs, if not in one product line then in another (my 830 Pro and 850 Pro are fine). If you refuse to buy from a maker because they’ve had a bug, then you very quickly aren’t going to be buying anything from anybody.

    What should distinguish Samsung from others — or not, we’ll see — is whether they continue to try to fix the problem.

  91. Depressing times. I bought these drives based on Samsung’s great reputation and the all around excellent reviews.

  92. The 830 series had an early firmware that occasionally caused hangs/BSODs when resuming from sleep; Samsung issued a couple of [url=<]firmware updates[/url<] to fix it. Nobody has an "impeccable" reputation; all manufacturers have bugs sooner or later. The question is how diligently they follow-up and fix the problems that do occur, especially when they seem to be hard to repro and continue to linger.

  93. I’m sure I’m not the only one who has been recommending Intel and Samsung SSDs to clients, friends, etc as the two top-tier options as far as SSD reliability. (Given that for most users, performance reached “good enough” a couple years back.)

    I think Samsung got a free pass from many (myself included) based on how they handled this bug initially; acknowledging the issue and offering a firmware update in a reasonable amount of time.

    The benefit of the doubt is gone, along with quite a hit to Samsung’s reputation. Can’t really unring that bell – just ask the corpse of OCZ.

    Sure glad I had no reason to upgrade my 830 Samsungs and 5xx Intels!

  94. The issue isn’t benchmark speeds, it’s old data reading speeds. Even though your drive is fine with freshly written benchmark data, old data that you haven’t touched in awhile might load slowly.

  95. Just my $0.01 here, but my 840 EVO has the exact read speed in HD tach as when my last benchmark was run several months ago. In fact I had a double take because I only saw one color on the graph.
    This drive has been used now for several months in a daily use laptop; and my read speeds are still showing as 113.1MB/sec … Granted this is a Lenovo T61; so the interface is SATA1. Still not 50MB/s either.

  96. Wasn’t the issue related to TLC? The 840/850 Pro are probably fine, but the 850 TLC drives might not be.

  97. Yes. Sad really, as I knew I was taking a risk with buying the 840 EVO instead of my usual conservative approach to storage media, so will swap it out soon enough and won’t be going with Samsung again for the foreseeable.

  98. Same. I got an 840 and 840EVO. Tired of this bs. Not buying from Samsung anymore and waiting for the class action.

  99. Since the problem is still present after the ‘fix’ I wonder if the 850 series has the same problem.

    I’ve been waiting on the 840 (non EVO/Pro) to have a firmware update issued. At this point it looks like there won’t be one.

  100. Wow. They still haven’t patched the older 840 models which some users reported having the same issue, and the 840 Evo issue is not completely resolved, too?

    Is Samsung the Seagate of solid state disks?

    I know I’m never buying from them again.