PDA

View Full Version : SSDs are incompatible with GPT partitioning?!


techvslife
02-22-2012, 02:09 PM
I'd welcome any advice, thoughts, confirmation, etc. for this problem:
I bought some new Samsung 830 SSDs for use as boot drives in new x79 windows 7 x64 pcs with UEFI bioses, so I had to partition them as GPT. (windows requires GPT partitioning for booting when you have a UEFI BIOS.)

I then discovered that the backup GPT partition area was being corrupted every few minutes by random information.

This is the reply I received from Samsung:
Dear Customer,
It is recommended that the unit be placed in an MBR format as the unit was made to work in a MBR. The cause of the corruption is for this very reason. We apologize for the inconvenience.
Thank you.

I wonder if this happens to any SSD with a GPT format, because of SSD use of unallocated space for "overprovisioning" purposes. The last few sectors of the drive, where the GPT backup partition header and table are, are unallocated space.

I ran the gpt fdisk utility to discover the corruption, but other utilities also discover it, such as mbrwizard. I inspected the last sectors and saw random data being written over the backup GPT partition information. I also confirmed this behavior with the author of the gpt fdisk utility. (it could be the cause is the Intel RSTe drivers instead.)

Here are the links to the windows utilities that confirm ongoing corruption:
http://sourceforge.net/projects/gptfdisk/
http://firesage.com/mbrwizard.php
The command by the way to check with gpt fdisk is simply:
gdisk -l 0:
and with mbrwizard it is:
mbrwiz /list /disk=0
In both cases replace the zero with zero, one, two, etc. depending on the configuration of your GPT SSD hard drive. Typically a boot drive will be zero or one. If anyone has tried this on a Samsung 830 ssd, that would be particularly helpful.

Thanks.

Coup27
02-22-2012, 02:29 PM
I have an 830 and a UEFI capable motherboard and recently I installed my system in UEFI mode as an experiment.

Firstly, if you have the capable hardware, you are not required to install in UEFI mode. If I insert my Windows 7 installation media and go into UEFI, I have two boot options:

Sony Optiarc
UEFI: Sony Optiarc

If I use the first one, Windows will use MBR. If I use the second one, Windows will use GPT.

I ran my system in UEFI mode for 4 weeks and I did not have a single error. My system worked fine. I secure erased my drive and entered the Windows 7 partition editor at the installer stage. I requested to create a single partition which was 12GB less than my overall space. Windows made the 3 necessary partitions and left me 12GB unallocated to use for overprovisioning (as the SSD Magician requests).

My system did work fine. The only thing which did not work was the SSD Magicians TRIM and overprovisioning features. A few emails back and forth with Samsungs contact centre and Samsung finally responded and said:
The current version of Samsung SSD Magician (v3.1.1) does not support GPT partitions .

On PC with GPT partitions "Over provisioning" and "Performance Optimization" features
will not work as these features depends on partition type

We will analyze and implement the support for GPT partitions in next version.This reads to me that the only thing preventing 100% GPT support is their current Magician software.

In summary, I successfully ran my system in UEFI with GPT for 4 weeks without issue so I don't believe that the 830 cannot run in this mode, the fact Samsung have said that in your contact with them may only be a reference to their Magician software.

techvslife
02-22-2012, 02:46 PM
Coup27,
Thank you very much for your reply. The corruption occurs (so far) ONLY in the last few sectors, so the question is:
Is your GPT backup partition header and table valid and uncorrupted?

The easiest way to check is to run gdisk or mbrwiz:
http://sourceforge.net/projects/gptfdisk/
http://firesage.com/mbrwizard.php
I would be grateful if you could check that, if you are still using GPT on any SSD. Typically I see it valid for about fifteen minutes after correction or boot or writing a new partition, then it goes invalid after that.

I don't think the Samsung email was a reference to the Samsung Magician software because my email to them was specific in pointing out where and how the corruption occurs, and their reply mentioned that data corruption would result.

It is interesting that they say to you that overprovisioning won't work with GPT partitions. It is a little weird that low level SSD functions "know" so much about logical OS structures--it seems a recipe for disaster frankly, since those higher-level structures vary by OS and change so much.

(As far as windows, Microsoft says it cannot boot the OS, when you have a UEFI bios, from a hard drive partitioned with MBR--I don't know about installation though:
"Systems that support UEFI require that [the] boot partition must reside on a GPT disk. Other hard disks can be either MBR or GPT.")
http://msdn.microsoft.com/en-us/windows/hardware/gg463525
Perhaps your uefi bios has the option of accessing the internal hard drive without using uefi methods; I don't know much about that.

Thanks.

Jocelyn84
02-22-2012, 03:05 PM
I own an Asus P8H67-I Deluxe and for GPT partitioning, I just hit Shift + F10 at the windows installer then do the following.

Type diskpart, press Enter
Type lis dis , press Enter
Type sel dis X (where X is the number your drive shows up as), press Enter
Type cle, press Enter
Type convert gpt, press Enter
Type cre par efi size=100, press Enter
Type for quick fs=fat32 label="ESP", press Enter
Type cre par msr size=128, press Enter
Type cre par pri align=4096 , press Enter
Type for quick fs=ntfs label="SYSTEM", press Enter
Type ass letter=C noerr, press Enter
Type exi, press Enter
Close command then install to "System" (Name can be changed after install).

Anyway, that's kind of disappointing Samsung's features don't work with GPT disks.

beginner99
02-22-2012, 03:07 PM
in my laptop (X220) I use intel mSata ssd as os drive with UEFI and it is GPT. Haven't noticed any abnormalities.

techvslife
02-22-2012, 03:09 PM
Thanks for that information. If anyone has an SSD drive that is partitioned in the new GPT style, I'm checking to see if you are experiencing corruption of the very last sectors, of the backup GPT partition header and table. The easiest way to check is to run gdisk or mbrwiz:
http://sourceforge.net/projects/gptfdisk/
http://firesage.com/mbrwizard.php

Again, I am checking about corruption specific to this area of the drive with SSDs that are partitioned GPT style, although of course information on corruption in any other sector is helpful. However, so far I have seen corruption only in the last sectors of the drive, where the GPT backup partition is stored.
Thanks.

Burner27
02-22-2012, 03:19 PM
Can someone explain to me why one would want to run in UEFI mode? Isn't that on by default on EFI enabled mobos?

Coup27
02-22-2012, 03:34 PM
techvslife I think there are a few bits in your post that you have gotten the wrong end of the stick of.

I am no longer running in UEFI and GPT. I changed back to BIOS and MBR a little while ago. I did this because occasionally I do like to run my SSD Magician TRIM feature and as this was broken in UEFI and GPT mode, I decided to switch back. It was an experiement to see if there were any differences. Had the Magician supported GPT I would have remained with it.

I am suprised Samsung say it does not work with GPT as mine clearly did. Also the support advice I got back only said it was the Magician which did not support GPT and not the product as a whole.

My comment about overprovisioning refers to the Samsung overprovisioing feature within Magician. This feature allows you to shrink down a partition to create some unallocated space. As the Magician does not work with GPT, you cannot use this feature to shrink down a partition. This is why I made a partition 12GB less at the installer stage.

Regarding Windows, as I said previously if you have a UEFI board you can install Windows in either mode, you simply select which mode to boot from within UEFI. If you choose MBR, then UEFI is acting in "BIOS" mode. If you choose UEFI [boot device] then UEFI is acting in UEFI mode.

What MS is saying is that if you want to boot in UEFI mode with GPT, then Windows must reside on a GPT disk. This is a bit of a silly statement because when you install Windows in UEFI mode then it automatically uses GPT anyway. Any subsequent disks you use for data or whatever else can be MBR or GPT. My storage drive I left in MBR and I had no issue.

Personally unless there is an odd hardware combination error I do not believe the drive is not compatible with GPT.

If you still have no joy debugging further, insert your Windows installation media into your DVD drive and go into UEFI and there should be an option similar to mine on being able to boot the device, or boot the device with UEFI in front of it. If you have this option, I would secure erase the drive and then install in BIOS and MBR and you will have a stable system.

Burner27, in summary UEFI is the replacement for BIOS. While there is no visible difference to the user once your system is loaded, UEFI is a lot more secure and has many behind the scenes differences. BIOS also does not support GPT partitions where as UEFI does. Eventually it will replace BIOS but my Asus board allows me the flexibility to run in either mode depending on how I choose to install Windows.

Jocelyn84
02-22-2012, 03:35 PM
Thanks for that information. If anyone has an SSD drive that is partitioned in the new GPT style, I'm checking to see if you are experiencing corruption of the very last sectors, of the backup GPT partition header and table. The easiest way to check is to run gdisk or mbrwiz:
http://sourceforge.net/projects/gptfdisk/
http://firesage.com/mbrwizard.php

Again, I am checking about corruption specific to this area of the drive with SSDs that are partitioned GPT style, although of course information on corruption in any other sector is helpful. However, so far I have seen corruption only in the last sectors of the drive, where the GPT backup partition is stored.
Thanks.

Is this what you're looking for?
http://img804.imageshack.us/img804/5200/gdisk.jpg

Can someone explain to me why one would want to run in UEFI mode? Isn't that on by default on EFI enabled mobos?

You actually have to choose UEFI: DVD Drive instead of DVD Drive during boot. Coup27 touched on this above.
In term benefits, here is a bit quoted from a rather good guide on OCN - http://www.overclock.net/t/1156654/seans-windows-7-install-optimization-guide-for-ssds-hdds

Benefits of a UEFI/GPT boot disk vs. MBR:
Although not currently applicable to SSDs, GPT disks can exceed the 2.2TB bootable limit of a MBR partitioned drive. MBR drives are limited to four partition table entries, unless a secondary "extended" partition structure is created.
Data critical to platform operation is located in partitions, and not in un-partitioned or "hidden" sectors which in certain instances, can lead to system instability. Data contained in hidden sectors that result in system problems are difficult to debug.
GPT disks use primary and backup partition tables for redundancy and 32-bit cyclic redundancy check (CRC32) fields for improved partition data structure integrity.
A UEFI boot is more secure, and less vulnerable to pre-boot malware.
A system utilizing a UEFI boot, will boot and recover from sleep faster than the same machine using MBR.
UEFI is the future, and as different implementations of UEFI mature, UEFI will be used for much more than just booting a computer.

Burner27
02-22-2012, 03:54 PM
I have an Asus P8P67 Pro and my computer boots to UEFI mode and i installed windows from there. Yes, it looks prettier than regular BIOS but I really don't notice the benefits over a regular BIOS.

Coup27
02-22-2012, 03:56 PM
Burner, go into device manager and expand disk drives. Double click on your boot drive and go into volumes. Click on populate and what does it say in the partition style row?

The noticeable difference between BIOS and UEFI is the graphical interface and boot support for GPT partitions. The rest of the differences are largely behind the scenes.

techvslife
02-22-2012, 04:11 PM
The command by the way to check with gpt fdisk is simply:
gdisk -l 0:
with mbrwizard it is:
mbrwiz /list /disk=0
In both cases replace the zero with zero, one, two, etc. depending on the configuration of your GPT SSD hard drive. Typically a boot drive will be zero or one. If anyone has tried this on a Samsung 830 ssd, that would be particularly helpful.

@jocelyn27,
Yes, that's it, thanks; it seems that you are not experiencing the Samsung corruption issue. One caution is in my testing the corruption only occurs some twenty or thirty minutes into operation, and you get cycles back and forth, but if you run it three or four times over a couple of hours and always pass (no error), then you're probably fine. Your ssd drive is attached to the motherboard Intel SATA port I assume?

I should add that I'm using an Asus p9x79 deluxe uefi motherboard, and two Samsung 830 SSD drives in GPT partition are connected to the Intel SATA6 ports in RAID 1 config. I have a few factors to check out, but Samsung seems to be confirming that it is the SSD drives themselves that are the source of the corruption, which fits into some speculation I have about overprovisioning as a cause.

I didn't select any BIOS emulation mode during installation, and it's too late to change back to MBR without of course reformatting and redoing everything--and not sure how well my uefi supports that mode, so I'll stick with uefi for now. Again, I am NOT seeing general corruption, only limited corruption in the backup GPT sector at the end of the drive, which is not essential for operation. However, I'm of course worried that there's more to this than meets the eye.

@coup27,
I'm referring to manual overprovisioning, not the one done in the Magician. In any case, all the Magician does, I believe, is set unallocated space; which you can manually do by setting partitions to a total size that end up smaller than total drive size (so there is a dead zone at the end). However, I'm a little hazy on how ssds do the overprovisioning magic. Anyway, if Samsung SSD does not know about gpt, it's possible it might not know not to overwrite certain unallocated structures at the end of the drive used by GPT. At least that's my current theory.

Coup27
02-22-2012, 04:22 PM
"Again, I am NOT seeing general corruption, only limited corruption in the backup GPT sector at the end of the drive, which is not essential for operation."

When I did all this, I just used my system and did not notice any errors outside of the Magician software.

What errors are you discovering and are they only present if you go and find them? I don't know the ins and outs on GPT partitioning and after booting a secure erased drive I let the Windows installer take care of everything and then didn't look at the partitions at all once I confirmed it was all in GPT, so I don't actually fully understand the errors you are getting.

Regarding BIOS emulation, my Z68 Asus board supports this feature fine so it's unlikely you will have problems with an Asus X79.

techvslife
02-22-2012, 05:28 PM
What errors are you discovering and are they only present if you go and find them?

The errors are consistent: corruption in the GPT backup partition header and table, those reside in the very last sectors of the disk. The utilities I mentioned will report that the backup partition is invalid or corrupt, or that there are CRC errors. When I inspect, the backup GPT header or table has been overwritten with information, sometimes random, sometimes readable, from other sectors of the drive. If I force a good GPT partition to be written, then it holds for about fifteen or twenty minutes, before the backup partition table & header gets overwritten by junk again.

Since only the backup partition is involved so far, it would not affect normal operation of the drive, but I'm worried other parts of the drive may be affected in a subtle way over time that hasn't showed up yet.

At this point, it would be useful to know if anyone else sees this issue with ssd gpt drives, but it may be limited to samsung 830 ssds, or to my configuration--hard to know.

Burner27
02-22-2012, 10:25 PM
Burner, go into device manager and expand disk drives. Double click on your boot drive and go into volumes. Click on populate and what does it say in the partition style row?

The noticeable difference between BIOS and UEFI is the graphical interface and boot support for GPT partitions. The rest of the differences are largely behind the scenes.

Mine all say MBR but I have the option to select GPT.

Coup27
02-23-2012, 04:29 PM
To be honest I still don't really understand what's going on here.

Looking at a GUID partition structure image, the last section is "Backup GUID Partition table header". Are you saying that this area is being overwrote with jibberish?

Does the system still operate without issue but if you check via some tool or command line, you notice that this area is not how it should be? If so, what are the implications of this? What is it used for? As I said, my 830 "appeared" to work fine but I never checked any partition structure to ensure stability because surely if there was an issue with my partition structure I'm sure a BSOD of Windows error would have came about.

Burner, although you have an UEFI based motherboard you have installed Windows in BIOS emulation mode with MBR as your partition structure. To the best of my knowledge if you convert a disk to GPT it wipes it clean in the process so there really isn't any point. Although next time you clean install you may want to give it a go.

techvslife
02-23-2012, 06:45 PM
Looking at a GUID partition structure image, the last section is "Backup GUID Partition table header". Are you saying that this area is being overwrote with jibberish?

Yes that is it: that area, really several sectors at the end of the drive, are being overwritten with jibberish or with data from other parts of the drive. As appears confirmed by Samsung, the drive overwrites at least the last sector of the drive, perhaps because it is wrongly seen as unused or safe to do so. It seems to be using that area as a scratchpad space or some space to do some sort of housekeeping or garbage collection. If it is by design, it is a very bad or foolish decision, as even with MBR disks, the last sector is often not safe to overwrite (for example, Microsoft uses the last sector to write critical metadata for dynamic disks). However, I have seen this behavior only with Samsung 830 drives using GPT partitions.


Does the system still operate without issue but if you check via some tool or command line, you notice that this area is not how it should be? If so, what are the implications of this? What is it used for?

Writing over the area means you lose one of the main advantages of GPT over MBR: the ability to restore the partition from the backup if the main copy is altered either by accident, or malware, or by a hardware failure. That means you lose an important safeguard against loss of all your data. You would not see a boot failure or operation problem if ONLY the backup partition was overwritten -- though you would if you needed to use it because for whatever reason the primary partition area was not correct or fully readable. Backup copies of file allocation tables operate in a similar way, and those areas should not be overwritten by third parties either.

It is also not clear what other "unallocated areas" of disk the hard drive is writing to; I simply don't know what the drive is doing here, and whether it is by design or some defect. There may be undetected corruption at this point. The only corruption I know for sure is happening so far are on those last sectors, and it is happening regularly.

For now, it would be good to get confirmation from other Samsung 830 drive owners, if they use GPT partitions, of whether their drives are also corrupting these last sectors. The easiest way to do that is to run one of the two GPT utilities I mentioned earlier, checking to see if the backup partition tables are intact. If they do show as corrupt, I would recommend as a minimum saving the primary good GPT partition table and header to a file on a different drive, which those utilities can do.

I also caution that it may turn out that the Intel RSTe drivers are at fault instead, or some other cause; I have done only preliminary testing, as this is a production machine.

Coup27
03-02-2012, 05:08 PM
techvslife, firstly thank you for taking the time to explain that. I did read it the day after you posted it but I haven't had time to do anything lately let alone any forum posting.

From the issues I had with my Magician software and GPT, I made a "friend" at the Samsung contact centre ran by Hanaro in The Netherlands. I emailed her this thread and asked her for her thoughts. She said she did not know but passed it onto Samsung for a response and this is what I got back today:

"Hi please refer to the below (answer from our korean engineer):

To tell the conclusion first, it is not SSD related issue. SSD cannot make any corruption in GPT patition.

The user (techvslife) is worried about SSD can change the data on backup area in GPT partition, because GPT partition makes backup area on last sectors of the SSD.
It is known that the last sectors of SSD is reserved for data recovery, and that's why he is worried.
Howerver, the reserved area is unaccesible actually. Either MBR or GPT, user cannot make any partions on that area.
SSD do not chage backup area but chage reserved area.

As you may know, user can use overprovisioning for increasing the reserved area in MBR. that is additinal reserved area in only MBR. GPT may not support overprovisioining.

The user said that the backup area is overwritten, but it is not related on SSD operation. (It might be a normal operation of GPT)."


I don't know if that helps you at all?

techvslife
03-02-2012, 07:13 PM
Coup27,
Thank you for forwarding that to someone at Samsung; that was certainly above and beyond.

I'm afraid though it doesn't end up shedding much light, perhaps because something is getting lost in the translation(s). (All my replies from Samsung have that problem; I guess I'll have to acquire a friend who knows Korean.) This looks erroneous if I understand it right:
It might be a normal operation of GPT.
GPT is only a partition style and doesn't write anywhere; and its partition tables, even its backup copies, should not be overwritten unless of course partitions are being added, removed, or altered. (that's the same as MBR.) I have no doubt that the backup GPT area is being constantly corrupted with junk data, completely invalid for GPT partitions or any other partitioning structure; every tool I use confirms it is junk, and I've inspected its contents and can confirm that.

I have no idea how Samsung does overprovisioning--the whole area is dark to me. But we do know that the Samsung 830 SSD uses overprovisioning and that if an area is left unallocated in a MBR system, that area of the drive (what is being called the "reserved area" in the reply you received) will be used for overprovisioning and hence presumably will get written with some kind of "garbage" or shuffled data by the SSD. I don't know how the SSD "understands" a higher logical structure like MBR, but it looks like it must in some way. While this is a guess, it may be that, exactly because the Samsung SSD doesn't understand GPT, it is overwriting the backup GPT header and partition table. (That last sector location is also not covered by the GPT's "protective MBR," its fictional MBR partition table to protect non-MBR apps from fooling around with the GPT data). That is, it looks like with GPT partitions, Samsung is mixing up the "backup areas" (where backup GPT header and partitions are, areas that are unallocated AND unsafe to overwrite) with the overprovisioning "reserved areas" (areas that are unallocated AND safe to overwrite). At any rate, the backup GPT area is getting overwritten with random data that is not coming from any app (as far as I can see) but from random other places on the drive.

I still leave open the possibility it has something to do with the new intel rste drivers, since that also could explain random information from elsewhere, though it's somewhat unlikely if the intel rste drivers were the cause that it would be confined to the unallocated areas of the drive.

Anyway, it's true it is minor if that area is the only area getting corrupted; I have a manual backup copy of the GPT tables. I'll post back here if I discover anything else going on.

Coup27
03-31-2012, 04:18 PM
Sorry to thread bump people but techvslife if you're still around these parts, log in and check your PM's. I've been doing my own investigation into SSDs and UEFI/GPT and have sent you a couple of PMs.

Cheers.

nilsoli
09-02-2012, 06:19 AM
Hello,

This thread seems a little dead, but as I've just installed w8 on a samsung 830 in uefI mod, I will share here my experience :

First when installing windows I didn't reserved some unallocated space for overprovisionning.

Install went fine, but as everyone said here, it was impossible to use magician to create some unallocated space on my disk.

I tried to launch gdisk but it doesn't seem to work on w8 (Problem opening \\.\physicaldrive0 for reading!), so I don't know if I have a corrupted disk.


For safety, I used windows's disk manager to reduce my C: partition and create 12GB unallocated space. I hope this is enough for overprovisionning to work.

Can someone tell me if there is another way to check my disk ?

Coup27
09-02-2012, 09:25 AM
According to the communications I had with Samsung's support centre at the time, leaving unallocated space under GPT does not actually get used as spare area aka over provisioning.

The partition structure of GPT is very different than MBR. MBR has nothing at the end of the disk. Either a partition takes up all the space or there is a lump of space at the end. At the end of a GPT disk there are GPT related entries such as GUID partition entries and the GUID backup partition header. See this link. (http://msdn.microsoft.com/en-us/library/windows/hardware/gg463525.aspx)

This is what this thread was about. The OP was claiming the 830 was corrupting the overall GPT disk structure because what it assumed to be spare area, wasn't. Samsung said to me in one email "the unit is designed for MBR use only" but never expanded on that. They did tell me they were working on an SSD Magician release which supported GPT. That was scheduled for April, then July, and now it's pushed back to October I think. The latest version should also display SSD life as Samsung have disabled the SMART values from 3rd party apps displaying this. I would assume this release would be coupled with a GPT aware firmware update.

There is an argument to suggest that over provisioning is working, but at the expense of the GPT structure integrity.

Ultimately I gave up on UEFI/GPT and went back to BIOS/MBR. I think UEFI/GPT is a good 6 months ahead of it's time. Software support needs to catch up.

nilsoli
09-03-2012, 05:25 AM
As my system seems to work as it is supposed, I'll keep running with UEFI/GPT.
I'm not affraid of having to format and reinstall everything.

Just waiting for Samsung now...

BFG10K
09-08-2012, 12:36 AM
Ultimately I gave up on UEFI/GPT and went back to BIOS/MBR. I think UEFI/GPT is a good 6 months ahead of it's time. Software support needs to catch up.
Yeah, I personally wouldn’t bother with GPT unless I had drives bigger than 2TB.

gillesv
09-08-2012, 05:54 AM
I installed my SSD as GPT because I wanted to try UEFI.
I never had any problem though I cannot use 'Over provisioning" and "Performance Optimization" in SSD Magician.
I will install the new firmware and software that support UEFI when available (but I think I will wait one month or two after it is available)

My Config:
Asus P8P67 Pro
SSD 830 256 GB Firmware: CXM01B1Q (I know it is not the latest one)
Dual boot W7 x64 / W8 x64 RTM (MSDN)
SSD Magician ver 3.1 (only installed in W7)

>gdisk 0:
GPT fdisk (gdisk) version 0.8.5

Partition table scan:
MBR: protective
BSD: not present
APM: not present
GPT: present

Found valid GPT with protective MBR; using GPT.

Command (? for help): v

No problems found. 49916525 free sectors (23.8 GiB) available in 2
segments, the largest of which is 49914511 (23.8 GiB) in size.

Command (? for help): p
Disk 0:: 500118192 sectors, 238.5 GiB
Logical sector size: 512 bytes
Disk identifier (GUID): 7951441E-D707-4C3E-B999-11B6E544A3AB
Partition table holds up to 128 entries
First usable sector is 34, last usable sector is 500118158
Partitions will be aligned on 2048-sector boundaries
Total free space is 49916525 sectors (23.8 GiB)

Number Start (sector) End (sector) Size Code Name
1 2048 206847 100.0 MiB EF00 EFI system partition
2 206848 468991 128.0 MiB 0C01 Microsoft reserved part
3 468992 104083455 49.4 GiB 0700 Basic data partition
4 104083456 345755647 115.2 GiB 0700 Basic data partition
5 345755648 450203647 49.8 GiB 0700 Basic data partition

akc
09-25-2012, 01:44 AM
Ultimately I gave up on UEFI/GPT and went back to BIOS/MBR. I think UEFI/GPT is a good 6 months ahead of it's time. Software support needs to catch up.

1. So no one has duplicated techvslife's corruption issue? (though i believe it's real). How about uninstalling Intel's drivers to rule out?

2. As an aside: As a non-Mac owner I've heard Apple does use Samsung 830 SSDs in their recent laptops, and that Apple supports some form of GPT. Wouldn't that combo present with problems?

3. I'm glad I haven't gone ahead to play with GPT on my 830.... (though I'll look into whether Plextor has done better).

Oubadah
11-03-2012, 06:59 PM
I wonder if this happens to any SSD with a GPT format...

I haven't tried GPT on the 830 yet, but I recently formatted my Crucial M4 with GPT:

http://i.imgur.com/AlXrR.png

^ I don't even know how to interpret the results - it it corrupted?

Insert_Nickname
11-04-2012, 06:09 AM
I have run the 830 for 4 months with GPT, without any issues what-so-ever...

Coup27
11-04-2012, 03:17 PM
I have run the 830 for 4 months with GPT, without any issues what-so-ever...
If you read the whole thread then you'll see that in normal use you won't see any issues. The issue came about when somebody checked the integrity of the GPT structure and found it to be currupt.

From my limited knowledge of the subject, GPT contains information at the end of the disk which would allow it to be rebuilt from this information. The 830 possibly overwrites this information during use but unless you ever attempt to rebuild a GPT disk with the backup header than you will never see there is a problem.

Insert_Nickname
11-05-2012, 07:49 AM
If you read the whole thread then you'll see that in normal use you won't see any issues. The issue came about when somebody checked the integrity of the GPT structure and found it to be currupt.

From my limited knowledge of the subject, GPT contains information at the end of the disk which would allow it to be rebuilt from this information. The 830 possibly overwrites this information during use but unless you ever attempt to rebuild a GPT disk with the backup header than you will never see there is a problem.

No, you misunderstand. The GPT structure is perfect here, no corruption. Mine might be a newer revision then the OPs...?

Coup27
11-05-2012, 10:55 AM
No, you misunderstand. The GPT structure is perfect here, no corruption. Mine might be a newer revision then the OPs...?
My bad. I don't know then. There will be no changes in the 830 throughout its life, they'll all be the same. I was able to repeat the OP's problem when I tried on my 830 but with no official word from Samsung on the subject we never really got to the bottom of it.

Be interesting to see if the 840 & 840 Pro are officially supported on GPT.

AdamK47
11-05-2012, 11:05 AM
My Vertex 4 was running MBR until recently. Converted it over to GPT while retaining my Windows install and other data. No problems so far.

greenturtle99
12-11-2012, 10:44 AM
I have been following this thread for some time and had nothing to contribute. However, I want to confirm I have identical problems since I started my new build in late summer. I have two Samsung 830 SSDs that have only the EFI boot partition and Linux ext4 partitions. I have left substantial unpartitioned space (40GB) on each SSD for now since I have been unable to find out exactly how Samsung overprovisions the SSDs. Version 3.2 of the Magician software has been no help.

I have a Windows 7 installation on an HDD primarily for diagnostics since Linux is not officially supported on my Asus P9X79 WS. Mostly I run VMs for daily use of guest OSs. Usually, the back up table corruption happens on the SSD that does not have the EFI boot sector but also happens occasionally on the other SSD. I have had no operational errors otherwise and I keep copies of the GPT partition tables on HDD as well as backups (remember to back up your EFI partition). But low level corruptions (even of backup tables) are always irritating.

This has been an education in the state of implementation of EFI, GPT, motherboards and SSDs. I have followed all the suggestions in the thread, very slowly and iteratively, to try to isolate the problems (boy do I dislike intermittent failures). It has made me very skittish about SSDs that do not list what they support and don't (usually very incomplete and impossible to find out even from the vendor). I bought Samsung because it was one of the few consumer SSDs that claimed support for Linux. I hope the release of Windows 8 will push the maturity of EFI and GPT.

But let me confirm the problem exists.

Amnesia1187
12-11-2012, 11:01 AM
I have been following this thread for some time and had nothing to contribute. However, I want to confirm I have identical problems since I started my new build in late summer. I have two Samsung 830 SSDs that have only the EFI boot partition and Linux ext4 partitions. I have left substantial unpartitioned space (40GB) on each SSD for now since I have been unable to find out exactly how Samsung overprovisions the SSDs. Version 3.2 of the Magician software has been no help.

I have a Windows 7 installation on an HDD primarily for diagnostics since Linux is not officially supported on my Asus P9X79 WS. Mostly I run VMs for daily use of guest OSs. Usually, the back up table corruption happens on the SSD that does not have the EFI boot sector but also happens occasionally on the other SSD. I have had no operational errors otherwise and I keep copies of the GPT partition tables on HDD as well as backups (remember to back up your EFI partition). But low level corruptions (even of backup tables) are always irritating.

This has been an education in the state of implementation of EFI, GPT, motherboards and SSDs. I have followed all the suggestions in the thread, very slowly and iteratively, to try to isolate the problems (boy do I dislike intermittent failures). It has made me very skittish about SSDs that do not list what they support and don't (usually very incomplete and impossible to find out even from the vendor). I bought Samsung because it was one of the few consumer SSDs that claimed support for Linux. I hope the release of Windows 8 will push the maturity of EFI and GPT.

But let me confirm the problem exists.

What exactly are you doing to see the corruption? As it's a different controller the problem may not exist at all on my 840 pro, but I gave been validating my disk in gDisk periodically to check for any issues. How frequently is it happening for you? Any errors or anything in your System Event Logs?

greenturtle99
12-11-2012, 11:24 AM
I am using gdisk too. I have not found any correlation in the system logs. The only way I find it is periodically running gdisk. There is some correlation between dual booting into Windows 7 but it can still happen in Linux (Ubuntu 12.04) without ever running Windows 7 natively. I have been ramping on using GSmartControl (smartmontools - Samsung 830 is in their database), trim in Linux.

I happens perhaps once a week but it really random with several in quick succession and and then I go a week or two. Are you seeing this on the 840 pro's too? The should be up to date on EFI/GPT because of the Windows 8 rollout (I hope). I have not stayed in Windows 7 long enough to correlate anything to this corruption.

In my fstab:

UUID=randomhex / ext4 defaults,relatime,discard 0 1
UUID=randomsmallhex /boot/efi vfat defaults 0 1

I understand the "discard" is important to maintain performance by enabling trim (on supported file systems). I am doing maintenance manually once a week since I don't know the source of the errors. I have not tried changing the EFI partition to "discard".

This weeks experiment is to turn off the Legacy bios compatibility in the Asus firmware making it pure EFI for a long term test (at least a couple of weeks). I gave up doing quick changes and I am being very methodical now.

wpcoe
12-12-2012, 11:07 AM
What MS is saying is that if you want to boot in UEFI mode with GPT, then Windows must reside on a GPT disk. This is a bit of a silly statement because when you install Windows in UEFI mode then it automatically uses GPT anyway. Any subsequent disks you use for data or whatever else can be MBR or GPT. My storage drive I left in MBR and I had no issue.

Interesting. Recently I experimented with Win8 and installed it in an EFI/GPT configuration, but Win8 could *not* see/access my other SSD or HDD when they were MBR.

Only after I converted my second SSD to GPT could Win8 in an EFI/GPT configuration see/access it.

I ended up reverting to BIOS/MBR because I didn't want the hassle of shuffling data off of my 1TB HDD to convert it to GPT.

Is there something unique to Win8 that would prevent me from seeing/using MBR drives when I have an EFI/GPT Win8 boot drive?

geohei
01-20-2013, 10:23 AM
Hi.

I found this thread through Google. I have a similar problem, but no data corruption (yet). Can someone confirm that the problem described below is the same one as discussed in this thread?

--

I installed Windows 7, Ubuntu 12.10 and Arch Linux (in that order) on a WD 1TB HDD using UEFI (Gigabyte Z77 motherboard) and GPT for the HDD. Everything worked fine.

Now ... when I use the Samsung SSD 840 PRO Serie (lastest firmware DXM04B0Q is on) and start installing Windows 7, the installer shows already that Windows 7 can't be installed on that disk because no MBR ... (can't remember the exact phrasing). But then, when I continue the installation, all works fine. EFI partition (100 MB), msftres partition (128 MB) and the Windows 7 partition is installed. Later on I proceed with Ubuntu and Arch Linux ... all works just fine!

GParted shows that the device uses GPT ... so all good!

But ... when I reboot the system, the UEFI Boot Manager doesn't show the SSD to be able to start using UEFI. It only shows the BIOS type entry.

The actually shown UEFI Boot Manager BIOS start entry is: "P0: Samsung SSD 840 PRO Series"
The UEFI Boot Manager UEFI start entry should be: "UEFI: Samsung SSD 840 PRO Series"

The second one, I don't see.

For the previously, for testing purpose used HDD, I saw:
The shown UEFI Boot Manager BIOS start entry is: "P2: WDC WD1000FYPS-01ZKB"
The shown UEFI Boot Manager UEFI start entry is: "UEFI: WDC WD1000FYPS-01ZKB"
That's the way it should be and also what I expected for the SSD to see.

Honestly, I have no clue where to start to look for the error.

Is it the motherboard UEFI firmware which doesn't support Samsung SSD UEFI start?
Is it the Samsung firmware?
Is it something I forgot / didn't know to do prior installing the OSs?

Any ideas?

Many thanks!

geohei
01-21-2013, 09:21 AM
Update ...

I ran one more test and installed Windows 7 again, and the error message that Windows 7 can't be installed on that SSD didn't reappear. No clue why!

But then, I kept on testing. I noticed that "UEFI: Samsung SSD 840 PRO Series" showed after Windows 7 was installed AND the BIOS was reflashed (same firmware version, F14). When I selected "UEFI: Samsung SSD 840 PRO Series", Windows 7 started properly. However when rebooting, I noticed that the "UEFI: Samsung SSD 840 PRO Series" entry in UEFI Boot Manager was replaced by "Windows Boot Manager".

So in fact, what I described in the previous article seems to be normal.

I don't have deep knowledge about the UEFI Boot Manager mechanisms, but the replacement of the SSD UEFI entry seems to be done on purpose by the Windows boot loader.

Hence, I believe I can safely use the system as it is now, without any worries about UEFI/GPT incompatibilities.

Coup27
01-21-2013, 03:08 PM
Interesting. Recently I experimented with Win8 and installed it in an EFI/GPT configuration, but Win8 could *not* see/access my other SSD or HDD when they were MBR.

Only after I converted my second SSD to GPT could Win8 in an EFI/GPT configuration see/access it.

I ended up reverting to BIOS/MBR because I didn't want the hassle of shuffling data off of my 1TB HDD to convert it to GPT.

Is there something unique to Win8 that would prevent me from seeing/using MBR drives when I have an EFI/GPT Win8 boot drive?
Nothing that I am aware of. My Windows 7 test installation detected my MBR HDD absolutely fine.

I am more than likely going to give Windows 8 a wide birth on the desktop front.

gillesv
03-08-2013, 03:42 PM
Samsung Magician 4.0 is available : Completely redesigned, support MBR & GPT.
I've installed it on my PC (W8 x64 / SDD830 GPT) : it's OK.

Can be downloaded from : http://www.samsung.com/us/support/owners/product/MZ-7PC256D/AM