Hardware(?) problem corrupting downloaded files

asgerix · 05-04-2009, 09:01 AM

I'm not sure if this is the right forum to ask this question, as I believe my problem to be hardware related, but here goes:

I recently bought a new desktop computer (because my girlfriend had all but taken over my old one :-) )

Here are the specs:

AMD Athlon 64 X2 Dual Core Processor 6000+
Abit motherboard A-N68SV(MCP68S)
4GB DDR2 RAM
500GB MAXTOR STM350032 harddisk
Realtek RTL8101E/RTL8102E PCI Express Fast Ethernet controller
nVidia GeForce 7025

I had a hard time installing an OS. First I tried installing different versions of Windows, but the installation went wrong every time - I can't remember the exact error message, but I think it said the install CDs were bad (they worked ok before and have worked fine after on other systems).
Then I tried Ubuntu Intrepid, and after a couple of tries I finally had a running system.

However, once in a while when I installed or updated applications, the installation would fail or the application would just not work. After much detective work I finally realised that the downloaded .deb packages were corrupt! Here is an example (the wine-gecko package), first lines are from the healthy file, second from the corrupt:

Code:

< 00005240  f6 c1 fb b7 4d 3c bb 03  f3 13 4d be 3a b6 0b 9a
> 00005240  f6 c1 7b b7 4d 3c bb 03  f3 13 4d be 3a b6 0b 9a

< 0000d020  ca 02 8b 27 a4 b7 bd 7b  5d 20 44 dd f8 b9 36 54
> 0000d020  ca 02 0b 27 a4 b7 bd 7b  5d 20 44 dd f8 b9 36 54

< 002785e0  1b 4c 84 cd cd 62 5a 39  c9 cd de ca d5 e5 30 9b
> 002785e0  1b 4c 04 cd cd 62 5a 39  c9 cd de ca d5 e5 30 9b

< 0050f360  54 69 f9 0e 0b 45 22 0d  7e bd 3f 25 7c 4d 52 ff
> 0050f360  54 69 79 0e 0b 45 22 0d  7e bd 3f 25 7c 4d 52 ff

< 00557ae0  ae de b5 3d b6 d3 bb e9  95 5e c4 b6 f3 dc 60 c9
> 00557ae0  ae de 35 3d b6 d3 bb e9  95 5e c4 b6 f3 dc 60 c9

Notice how the error is always a 1-bit switched to a 0, and always in the same position modulo 32 bytes!

My first thought was that the package was corrupt in the repository, but after downloading the same file a couple of times with errors appearing in different places each time (but still the same bit modulo 32 bytes!), that was obviously not the case.

My next thought was a faulty harddisk, so I tried downloading to a usb stick instead - still with the same problem.

Then I tried running memtest86+, and it showed a bunch of errors, so I thought that I had found the problem, but no, after replacing the faulty ram stick, I still had the same problem.
I *do* blame the ram errors for my problems with installing an OS in the first place though.

It should be noted that the problem usually does not show up immediately after a reboot. If I notice that a downloaded package is corrupt, I just reboot the system and download again. Then the problem usually goes away for a couple of hours. This does sound a bit like some hardware part getting too hot - but if so, why would the problem go away for several hours after a reboot? It's not like the computer would cool down much during the 30 seconds it takes to reboot?

Next up for scrutiny: The network card. Now, if the network card was faulty, wouldn't the network driver notice errors in the TCP/IP packet checksums? Or are the checksums performed on the card itself? I do have a spare network card that I can try instead, it just does not seem (to me) a likely source of the problem, so I have not gotten around to try it yet.

I will be very grateful, if anyone can suggest what else I could try - I'm still hoping I won't have to replace the motherboard.

As a side note, I find it strange that some applications were installed without apparant errors, even though I could see that the package was corrupt! For example the wine-gecko_0.1.0-0ubuntu1_all.deb package; when I installed the package (from Synaptic Package Manager), no error was shown in the log, but when I unpacked the file, the contained data.tar.gz archive was corrupt.

Regards,
Asger Grunnet

camorri · 05-04-2009, 09:29 AM

After reading your post, my first thought was ram. Since you have eliminated it as a possibility, I would suspect the system board next. Possibly the connectors where the ram sticks plug in.

Something to try. Take out the ram, ( handle carefully, ground yourself ) Take a clean sheet of bond paper. Double it over, and carefully clean the tabs on the ram. They should look shiny, probably do now, but contamination is hard to see. Re-install and give it another try. Have a look at the contacts on the ram. They are a source of problems.

I know it is a lot easier to swap the ethernet card, you can try that, but I would think such a pattern of failure is created by a serial device.

If you can not affect the bug, then your system board is the best bet. Since this is 'new', have you thought about taking it back to the vendor?

asgerix · 05-04-2009, 01:03 PM

Thank you for your reply. I will try what you suggest.

Regarding taking it back to the vendor: Yes, I have thought of that, but since the problem is hard to verify - you have to let the computer stay turned on for several hours, and then you have to download a large file and check for corruption - I am afraid that they will send it back to me along with a bill and say that the computer works fine.

win_to_lin_migrant · 05-04-2009, 03:35 PM

Quote:

Originally Posted by asgerix

I'm not sure if this is the right forum to ask this question, as I believe my problem to be hardware related, but here goes:

I recently bought a new desktop computer.

<snip>

I had a hard time installing an OS. First I tried installing different versions of Windows, but the installation went wrong every time - I can't remember the exact error message, but I think it said the install CDs were bad (they worked ok before and have worked fine after on other systems).
Then I tried Ubuntu Intrepid, and after a couple of tries I finally had a running system.

However, once in a while when I installed or updated applications, the installation would fail or the application would just not work. After much detective work I finally realised that the downloaded .deb packages were corrupt! Here is an example (the wine-gecko package), first lines are from the healthy file, second from the corrupt:

<snip>

My first thought was that the package was corrupt in the repository, but after downloading the same file a couple of times with errors appearing in different places each time (but still the same bit modulo 32 bytes!), that was obviously not the case.

My next thought was a faulty harddisk, so I tried downloading to a usb stick instead - still with the same problem.

Then I tried running memtest86+, and it showed a bunch of errors, so I thought that I had found the problem, but no, after replacing the faulty ram stick, I still had the same problem.
I *do* blame the ram errors for my problems with installing an OS in the first place though.

It should be noted that the problem usually does not show up immediately after a reboot. If I notice that a downloaded package is corrupt, I just reboot the system and download again. Then the problem usually goes away for a couple of hours. This does sound a bit like some hardware part getting too hot - but if so, why would the problem go away for several hours after a reboot? It's not like the computer would cool down much during the 30 seconds it takes to reboot?

Next up for scrutiny: The network card. Now, if the network card was faulty, wouldn't the network driver notice errors in the TCP/IP packet checksums? Or are the checksums performed on the card itself? I do have a spare network card that I can try instead, it just does not seem (to me) a likely source of the problem, so I have not gotten around to try it yet.

If you had problems during an install from a Windows OS disk then your NIC is not the problem.

Quote:

I will be very grateful, if anyone can suggest what else I could try - I'm still hoping I won't have to replace the motherboard.

<snip>

Regards,
Asger Grunnet

I suspect you have an intermittent hardware problem. The only way I know to diagnosing this is to swap each part of the system one at a time and see if the problem goes away. If you don’t have spare parts to swap out then your only option is to return the system to the vendor with a detailed explanation of the problems you experienced while trying to install and/or run Windows. I would not mention Linux as IMO this would only muddy the waters. I assume the PC was prebuilt but even if you built it yourself I would ask the vendor what your options are about returning the entire system as it may be not be possible to diagnosis without spare parts. Now is the time when you will find out just how good your vendor is. Good Luck.