Hello,

In 20 years I bought 7 or 8 hdd, from 80 MB to 40 GB. Never had a problem.

In late 2008 I built a new computer, I chose a Samsung F1 320 GB. In late 2009, I found out it had some bad sectors.

In february 2010, I bought a Seagate 7200.12 500 GB to replace the Samsung. Some months later, just because I was curious, I checked the SMART data and found out the reallocated sector count value was in yellow. Since I knew software sometimes don't know exactly how to interpret SMART data I wasn't very worried. Moreover I scanned the drive with hd tune pro which found no bad sector. But in october 2010, I got a SMART error message at the boot telling me the drive was going to fail. CrystalDiskInfo and Speedfan reported the reallocated sector count reached the error level. Again no bad sector according to hd tune or chkdsk. I sent it to the after sale service.

In november 2010 I bought a Samsung F3 500 GB to replace the dying Seagate. But in january 2011 I had a weird issue: Many files were corrupted (exe, dll, jpg,...). Files were not gone but the bytes were corruped. Not zeroed but something like random or compressed data. Chkdsk fixed some orphaned files and other errors. It may look like a virus but I have an AV running in background, also did a full scan and it found nothing. I checked the running tasks, the executables ran at startup, the services,...: nothing suspect. Moreover the modification time of the corrupted file were not changed. I thought it may be again the HDD but SMART is ok and hd tune reported no bad block. Finally I replaced the corrupted files (identified thanks to sfc /scannow) from the original windows dvd. I really didn't understand the problem, but well, disk crashes happen sometimes. One month later, it happened again but this time I couldn't even boot! Not a surprise: with the help of the recovery console and an hex editor I found out that something like 100 system files were corrupted! Again I manually restored them (what a painful task!) Then, again, I scanned it with hd tune: no bad block. Since I was using it, I launched the benchmark test and saw the curve was weird:

I redid a surface scan and saw the scanning speed decreases (down to 60 MB/s) while readind some blocks. ESTool by Samsung reports no error but I can see the progress bar slows down sometimes and hear the head making noise on some sectors. HDAT2 reported no error (but I saw some slowdowns during the scan). I tried MHDD but it didn't detect the drive.

So my questions:
- What does mean these slowdowns during the scan but with no error? Are some sectors "weak" but still readable after some retries? And maybe sometimes the drive doesn't correctly read or write these sectors and makes some files corrupted?

- The most important question: Why my drive are always get damaged !?
In a bit more than 12 months, three drives damaged (Samsung F1: bad sector, Seagate: reallocated sector count, Samsung F3: slowdowns during the scan and data corruption) For me it's hard to believe it's just a coincidence.
My box is an Antec P182 so the drive is at the bottom, near the PSU. Is it possible that the magnetic field generated by the PSU damages the plate of the HDD? The PSU is a Corsair 520HX, on a Seasonic base, then it's supposed to be a good PSU.
In the P182 the drives are in a vertical position, not horizontal. Could it be the cause?
It could be electric surges generated by the PSU or the motherboard but I suppose it should damage the PCB, not the plate, am I right?
By the way I never move the box and the drives are never too hot.
Can it be really a coincidence?
More generally, what is the cause for bad sectors? Magnetic defaults on the plate or the head that hit it?

Thank you for your help in advance

There are several parts which I don't agree with, in your analysis & comments, even before reaching the questions in your posting. There are also some pieces of missing data, which would definitely help to confirm or deny some possibilities - some of the missing data is gone forever (e.g. previous "problem" drives), while some other useful data might still be available from you.

I don't know whether it is worth my time to try to assist further, because many first time posters here disappear when they don't get the response that they want or expect. It all depends how much effort, time, and perhaps money, you are willing to invest on trying to diagnose the (one or more) problem(s) affecting your system at the moment. That work might lead to educated guesses about what happened to previous disks, or it might not - we can't be sure at this stage.

I'll reply to a few of your questions, but I'm not promising further responses as I don't have lots of free time:


Yes, that's possible.

It would be useful to see similar benchmark graphs from your previous "problem" disks, but you don't mention that they are available. It would also be a good idea for you to repeat this test several times on the current disk, and see whether the "dips" in the benchmark graph occur at the same parts of the disk each time, to confirm the disk as the likely cause of those "dips".

It is interesting (but inconclusive at this stage), that the dips in the benchmark graphic which you included, are all towards the outside (OD) of the drive. Again, it would be good to compare this pattern with other "problem" drives.

It would also be a good idea to watch for changes over time in that benchmark graph and in the SMART data, so you should be collecting and keeping those results frequently from now onwards, for later analysis.


That's unlikely, for the large amount of corruption you have reported - large scale data corruption tends to have other causes, in my experience. Having said that, mis-correction when using ECC is always a (very very very small) mathmatical possibility.

IMHO you would need to do much more testing (= time and perhaps money), to make progress on investigating the cause of this corruption e.g. using a different disk drive for real data, and dedicating that current Samsung disk drive to further investigation work.

You would also need to establish a better test for identifying that data corruption had occurred, rather than just that the system didn't boot since, as you explained, that took approx 1 month to show.

Also, although you seem to be correcting for corruption in the Windows files, I did not see any mention by you of finding or testing for corruption in your data files. That information would be very useful.


Interesting theory - disks do have a limit to the size of external magnetic field which they are expected to cope with, before it affects normal operation. You would need to use a gaussmeter in order to further investigate that theory directly; or else you could consider a different approach e.g. proactively changing that PSU for something different, which was known not to affect nearby disks.

It would also be interesting to see if the disk benchmark graph changed at all, when that disk is removed from the case and moved as far away from the PSU as possible (if necessary by using max length SATA cables). If the disk benchmark graph did change (and improved i.e. fewer or smaller "dips") that would fit with the possibility of a problem caused by a magnetic field from the PSU. However the lack of a change in the graph would not disprove that possibility.


I haven't checked the datasheets for the 3 drives that you mentioned (you can do that). However all modern disks which I have seen, are specified for operation when mounted both vertically & horizontally.

In summary: IMHO remote diagnosis of the problem(s) on your system, is likely to be very difficult as they appear to be intermittent; some tests (e.g. gaussmeter) can only be done by you; some evidence (previous disks) is not available for further analysis etc. etc.

Good luck

I doubt very much that any magnetic field inside your PC case would be large enough to affect your drive.

See the images in this thread:
fujitsu-mht2060at-how-update-firmware-t12730.html#p116429

Vertical mounting, at least in Seagate's case, is a problem for some drives, regardless of their specifications. For example, people have reported that the ST32000542AS drives perform very poorly when mounted vertically. Reallocated sectors develop on a regular basis.

See this thread where several new ST32000542AS drives developed bad sectors when mounted vertically , but not horizontally:

http://forums.seagate.com/t5/Internal-A ... /m-p/42506

ISTM that the dips in HD Tune's read benchmark graph are probably consistent with the slowdowns in ESTool and HDAT2. To confirm this with MHDD, you need to temporarily reconfigure your SATA controller for IDE, legacy, or compatibility mode in your BIOS setup.

Are you sure your files were corrupted? I'm not using a recent MS OS, but in earlier OSes, SFC (System File Checker) compared the DLL and EXE files against their original installation versions. Hence, SFC was incapable of differentiating between genuine updates and malicious changes or file corruptions. Is it possible that your "corrupted" files were actually MS updates?

_________________
A backup a day keeps DR away.

An interesting data point, thanks for that.

Reading the thread, we see that some of those 7200.12 drives were fine vertically, some died after a couple of months, some died much more quickly. People also reported other models of drives (Seagate & Samsung) working fine vertically.

All I can say is that based on many years experience of large populations of drives (from multiple manufacturers) in vertical mounting chassis, the premature death issues described in that thread are not the norm - but of course, rogue batches or models can occur for all manufacturers (it appears to be the case for some manufacturers even more than others ).

For the problem in your quoted thread to explain everything in the OP's post, then a similar problem would have to affect both the Samsung drives too. That seems unlikely to me, but as I pointed out in my post, we're missing useful data by not having the previous "problem" drives e.g. the OP can't now test his 7200.12 drive horizontally

Anyway, thanks again for the link - I'm just glad not to have seen such behaviour

I'll try to give the missing data


Well I can't promise I will be still here in 12 months (HDD is not my hobby) but I will reply to the questions until we can consider the topic is closed


Ok, it's what I thought


I'm sure the Samsung F1 HD322HJ 320 GB, the Seagate ST3500418AS 500 GB, the Samsung F3 HD502HJ 500 GB had normal benchmark / surface scan when I installed them in the computer.
I still have the F1 somewhere (not in the PC). I don't remember if I benchmarked it after the issue appeared but the curve had very probably the dips because it had bad blocks (red in hd tune) and I could see the test slowing down a lot when on the bad blocks. I name hd tune but I tested other tools and results were similar.
About the Seagate, I don't remember if I benchmarked it. I only remember surface scans reported no bad block. No data corruption/crash either. The SMART reallocated sector count was the only issue. I don't have it yet, I gave it back to the store where I bought it and they sent it to Seagate. After more than 2 months with no news from Seagate they accepted to replace it by a Western Digital Caviar Blue. It's my current drive.


As explained above my current system disk is a WD. But my latest problem disk, the Samsung F3 HD502HJ is still in the box. The results are consistent, the dips are always at the same part, like the capture I posted in my first post.
SMART data when I installed it (2010/10/28):


SMART data today:

The only only interesting data (to my eyes) is the read error rate. But I don't know exactly how to interprete it.


The bad blocks on the Samsung F1 were in the same zone: In my sytem partition (50 GB), 2 groups of bad blocks.
Maybe something interesting to note:
-I noticed the F1 had a problem because it crashed during the boot. After investigations I found out that the registry was damaged. So it seems the registry was stored on the damaged blocks.
-I noticed the F3 had a problem because calc refused to launch (invalid win32 application). I checked the event log and saw that among many errors like "can't launch service", the registry was damaged and that windows switched to the registry backup.
The registry is problably the most read/written file of windows, so probably the sectors where it's stored are the most used on the disk. Moreover I don't need a lot of space of my disks: I have a 50 GB partition system and the rest is a data partition. And on this partition I don't have a lot data (I think I never had more than 100 GB on it) and since I defragment my disks most of the files are at the beginning of the partition. So the zone with no problem probably contains no file. Maybe there's a wear of the most used sectors, I don't know... Just a supposition, I don't know how/why a sector becomes damaged.


I took some copies of the benchmark result and of the SMART data for the WD I currently use.


Does it have something to do with the SMART raw read error rate? I don't get very well it's meaning.


That's why my system is now on the new Western Digital. The Samsung is plugged to the MB but disabled. I can easily enable it for investigation


Well I used sfc /verifyonly because it's an easy way to identify corrupted system files because windows knows the original hashes. For other files files it's less easy. I know that 2 or 3 jpegs in my pictures folder were damaged, that one day I tried to launch an audio editor but 3 or 4 files were damaged (exe, dll...). And one part of a big thunderbird file containing my emails. Unless I open and verify all my files, documents, softwares, it's not easy to find out what is corrupted. Same thing for my data partition: it mostly contain music, videos, ISOs. And copies on my previous damages disks lol. I found no corrupted file on this data partition but I didn't have the patience to watch all the videos...


I don't have a gaussmeter. The PSU is a Corsair HX520, based on a Seasonic. In 2008 this PSU was ranked as one the best (like most Seasonic), including in reviews testing voltage stability, quality of the components, respect of the norms,... This model shouldn't have magnetic problem. Moreover I know I'm not the only one who has this PSU in the bottom of the box, near the disks and I'm not aware of similar problem. It's possible that my PSU as a problem though. But I don't have an other PSU and I don't want to buy want. If someone tells me that it's common some PSU damage disks, I would investigate this more seriously. But according to the members posting in this topic, it's not common.


Good idea. I'll try to do the test.


I'm aware of a forum with a lot of owners of my box, I'll ask if some have problems


Yes I know. I see no obvious reason for my problem. I already posted on some forums (less HDD oriented that hddguru) and most people think it's bad luck. As far as I know (or actually read) bad sector is mostly bad luck, that it's just your sample is a bad one. But three disks in a row, it's hard to believe.


Thanks


Thanks for the link.


Well if the BIOS sets in AHCI, MHDD see no drive. In IDE mode, it sees my DVDROM, the Western Digital but not the Samsung. I didn't test with only the Samsung plugged though. But as you said it's problably consistent.


Yes. Exe, dll, should begin with MZ. The content of the files looks to be compressed data (it can't be compressed by zip or rar). I'm unable to identify really the data, no obvious strings. All corrupted files are similars. I attached a corrupted file (HelpPane.exe).
I know it looks like a virus but I have an AV (Kaspersky) running in background, also did a full scan and it found nothing. I checked the running tasks, the executables ran at startup, the services,...: nothing suspect. Moreover the modification time of the corrupted file were not changed. My supposition is that the MFT or something is corrupted and some files index incorrectly point to wrong sectors (containing jpg, divx,...)

Well you have a good PSU and good ventilation I assume. I think its more just that you are unlucky. You mentioned drives that I avioded as much as possible just because they are know to have problems... bad luck..
A good example i use often is the Samsung S2 500gb that I had to replace 4 times in a month because it just stopped working.. but i eventually got a good one..

Once I did research on how a drives orientation is effected. The conclusion was that drives are tested label up. Dell approach was to put all drives upside down for a long time in their systems and there was some negativity to position drives in vertical positions. But all that is inconclusive because regardless of the orientation the drive performs at constant performance. But since then I only mount drives label up.......

As drives get bigger more sectors can go bad but newer generation drives handle this pretty well so there is nothing to worry about until windows tells you that a sector could not be read or you get a smart error.

A good guideline is to use smaller drives for OS <160GB as they seem to fail in a smaller percent than larger than that. and the rest of the data larger than 1TB... mirrored raid. People always say use the same drives and models your raid... ummm.. maybe.. but having 2 different models makes redundancy less likely to fail from the same problem on a particular model.. at once heheh

We had a thread here recently about research done in Russia. And apparently the Toshiba and Hitachi are the most reliable... Weird hey.

_________________
It seems that the unluckiest people in the world are those that don't backup.
Free Solutions
Picasaweb- Pictures and movies | Dropbox - Documents and other..

In my stats drives fail and work good the same way if mounted horizontally or vertically (BTW : ProLiant server here , 300 GB SCSI drives mounted vertically on the rack unit, ZERO problems until drives reached end of life being 24/7 on. But they are not 50$ drives). Maybe upside down is a bit too much, though.

Power supplies are enclosed on an iron box so the magnetic field outside is unlike to be "that" powerful to affect drive stability , even if there is a reactor as passive PFC.
Drives are built on aluminum / Al alloy chassis and have usually a stainless steel top cover that add additional shield.

So the magnetic thing is total bullshit to me. The problem is elsewhere or EXTERNAL to the PC.

Of course with the right instruments (not freeware programs, I mean HW diagnostics) it can be pinpointed quite easily.
Otherwise we can try to find an exorcist

I don't think this was mentioned, but it could be the power supply your using. If you've been using the same power supply on all these drives or if your power companies voltage is inconsistent and you are using the cheap power supplies, this could be your problem.

I never thought about this, to me a power supply was a power supply, it kicks out x volts and either works or doesn't work. Come to find out the more expensive power supplies weigh a lot more, they are a better buffer for the spikes, ups and downs from the power companies. In other words, having bigger/heavier transformers for the power to go through provides a more consistent supply of power to your hard drives, mother boards and cd/dvd drives.

Just a thought,
Dave

@Bernd - I missed your reply on March 9, due to it being held in moderation until whenever it later actually appeared in the thread...

I've read your reply carefully. Specifically, it is very interesting that your Windows Registry seems to be often affected with corruption, and also that some data files appear to have been affected too - but that this may caused by filesystem corruption (since chkdsk reported problems). (I realise it's difficult to check all data files. )

If I'm understanding your descriptions correctly, I suspect there may be 2 problem behaviours - something affecting your disks causing difficult-to-read blocks, and something affecting another part of your system (RAM? CPU?) causing corruption of the MFT and/or Registry. I reach this conclusion because it is highly unlikely for multiple disks to return so much corrupt data. If we assume (and, of course this is an assumption), that there is one cause (i.e. something common to both sets of effects), then as Dave48838 says, the PSU must be on the list of suspects.

For example, a faulty CPU could not cause difficult-to-read sectors on a disk; nor could faulty RAM or SATA interface etc. cause that. However, poor quality power (dips, ripple etc. etc.) to the internal electronics can cause RAM corruption (which affects the MFT, as that is held in memory) and also affect the disk drive(s).

Of course there are other possibilities (including the possibility that there are multiple faults), but I return to the fact that in my experience, large amounts of data corruption are almost always not caused by the disks. Therefore I suspect that the disks may be victims, not cause, here.

FYI, for the recent Samsung F3 500GB SMART data - I see no obvious evidence of a disk drive fault. Yet the graph shows that there are clearly parts of the disk which are difficult to read (i.e. slow) - this might be because those sectors were somehow "badly" written and are now difficult to read, and not a disk fault at all. Since you've explained that you don't use the whole disk, that explains why the problems are in the OD (lower LBAs).

However, you said that you don't want to get a new PSU (even to eliminate that as a possible cause), even though you agree that your specific PSU might have a problem. Please remember that I told you this type of investigation may need your time & money. By choosing not to spend some money, then you may delay (or prevent) a successful investigation.

I can think of several possible test approaches, depending on availability of other PCs for you to use as test beds - other members might suggest tests too. In any case, I would run some long-term RAM & CPU tests on your current PC - successful tests tell us nothing (they do not actually mean that there is no problem); however a failure would confirm that there may be something affecting the RAM / CPU on that PC (which would then potentially explain the Registry / MFT corruption). That cause might be discovered to be the PSU, or something else.

The first thing that would help you to make progress, IMHO, is to find a failing test.

Another approach would be to confirm "slow" sectors on the Samsung F1 disk (I suggest that disk, so you don't affect the test case of the Samsung F3 disk), then write to all sectors on the Samsung F1 disk in another PC (of course there is the assumption that another system will not have the same problem as yours).

Then do lots of reading (only!) of that disk in your current system. Does it read normally (fast - no dips in the benchmark graph) in your current PC, after being written completely in a different PC, and also when used for read (only!) tests for months in your current PC? If so, then the likely conclusion is that disk has no problem - it was the victim of sectors that were written badly, when it was written in your current system in 2009.

Of course it would be possible to spend time thinking of many other tests, depending on what you have the time & money to do, but I hope those suggestions give you some ideas.

Thanks again to Vulcan and other hddguru members who took the time to reply to me.

Since my last post I haven't worked a lot on the issue. So just some additional informations:

- As I expected, other people owning the same box + PSU + HDD in vertical position confirm they have NO issue
- No problem so far on my new HDD (WD blue), according to hd tune pro. No slow sector, unchanged SMART data
- Already did some stress tests (occt and memtest86), nothing wrong on the cpu/ram. By the way, it's not overclocked.
- The F3 that is still in the box (but in idle mode) is unchanged. Same slow sectors. Not more, not less.
- Unfortunatly the only other PC I have at home is IDE only. Of course I know some people with SATA controllers but I can't ask them to use their computers to do the tests.

One question to Vulcan: You say the sectors are slow because they were badly written. Does it mean I could "fix" them by writting to them (or formating)?

And thanks for this update. Replies below...


Understood - however that does not mean that your specific PSU is working (or has always worked) correctly; it just means that there is no known generic issue with that combination of components.


In one way, that's good news - but in another way, the lack of further problems means that you've not got a "failing test", as I mentioned in my earlier reply, and hence we can't make confident conclusions from this current "no problems" situation about what may have caused the previous problem(s).

One thing I would suggest as a possibility comes back to power - you now have a different disk drive (with different, perhaps slightly lower(?), power requirements), which is apparently working OK. If this "no problem" situation with that disk drive, continues for long enough that you are sure the system is now working better than (for example) when the Samsung F3 was being used, then consider testing what happens if you use that Samsung drive instead again (after re-writing it all - see below)?


Understood - although my comments above regarding unchanged SMART data, apply here too. Lack of errors during testing, does not prove that not problem exists, and does not prove that different tests would also pass.


Interesting. This suggests that the sectors are not deteriorating due to something in the environment (e.g. magnetic field), and that they may have been written in a way that made them difficult to read, due to an unknown (e.g. power(?)) problem at that specific time.


Understood - perhaps you could buy a little PCI SATA interface card? However, again, IMHO what would help most is for you to find a failing test. Therefore although using another PC (e.g. to re-write and then continuously read the Samsung F1 disk), may lead to a suggestion that the disk is not at fault, that's less conclusive than provoking the fault again - which may be easier for you to do on your "main" system, since you know that the problem has definitely been caused there in the past.


At the moment, this is my hypothesis, yes - but until the root cause is found, I can't say that this is definitely the reason.


If my hypothesis is correct, then yes. If you re-write all the slow sectors, or just re-write all the sectors (e.g. using any of the normal disk erasure techniques - but that does not mean all types of Windows formatting), and if the original (currently unknown) cause of the writing problem has been removed (use a different PC?) or is not triggered, and if (as we suspect) the disk itself is not faulty, then the re-written sectors should be readable as normal (i.e. fast) afterwards.

Hope that helps. Good luck, and please let us know when you have time to make any more progress.

Too bad we lost the poster, but I'm not surprised - as Vulcan noted in his first reply.


I just wanted to point out that when a client comes at me with symptoms that include any or all of:
- Frequent replacement of Fluorescent lamps (they should last YEARS not MONTHS)
- Occasional replacement of Displays, monitors, or TVs (again, YEARS not MONTHS)
- Frequent hard drive damage (4 years typical)
- Occasional display adapter replacement (again years not months)

These are "litmus test" indicators of an AC power system problem. These indicators are more subtle than what can sometimes be observed with very serious problems:
- Pulsating room lights or displays (CRT types, not so much for flat-panels)
- Shocks off of metal cases
- Shock or sparks between two computer chassis

Since this person was reporting "frequent" and especially since its getting worse over time. I would expect to find a problem with his building safety ground.

That problem could be its gone - I've seen the ground bonds rusted right off. Or it might be loose somewhere along its path.

More likely, he has connected more electronic equipment on the same branch, and each has a high-speed switching power supply. These things emit current pulses into the Ground or Neutral line as a side-effect of their regulation process. And your typical branch circuit is wired in SERIES.

That emitted current flows down the SHARED wires past all the other devices on its way to EARTH. Each device sees a voltage jump as the current flows past. Each device "thinks" its output voltage is "wrong" for the duration of the pulse and tries to "fix" it. That fix is usually DOWN, and if it goes DOWN enough, parts of the computer jump into standby mode or RESET. Either can result in data corruption.

Whats worse is Neutral or Ground is - by law in Canada and the US - connected to the GND bus of the computer. So all of those emitted currents show up on the GND bus as voltage spikes. If they are above ~0.5V they start causing damage to sensitive components in the Video card, monitor, or hard drive.

Which is why these devices make great indicators of this problem. My quick and dirty check for this problem on a client site is to put my Fluke (and it has to a meter at least as good as a Fluke) across GND and NEUTRAL. If you see MORE than 2 VDC or VAC you likely have a problem in the building premise wiring - or at least in how things are connected/distributed across the branches. Under 0.5 is GOOD, between 0.5 and 2 is a grey area. You should also check the RMS voltage, if it seems high then you can add "harmonic noise" pollution to the list of things to fix. But its time to get an Oscilloscope or power analyzer onsite and do a full check.

_________________
Thanks in Advance!
Rich.

Design of modern hard disks is very different than older hard disks. These Drives need very clean PSU for stating inrush current.
Besides if you are using consumer drives like 7200.12 in 24X7 usage ,then a premature failure is emminent, as these drives are not meant for such heavy use. Due to high RPM drives becomes very hot ,so active cooling is also must.
For serious business use seagate ES.2 drives are recommended.

I am with Dave48838 and rcooke on this 'coincidence' . Power is suspect. Also suggest a small UPS to filter out transients from sources such as refrigerators on same line. Just had a case last week where the computer shut down after a minit or so- power supply tested 'good' on testor for all voltages. Turns out is was generating noise- fzabkar may be able to explain why this supply killed the computer better.