HD streams have artifacts using unRAID

[guitarlp]

Very cool

 

I got it working and am running some tests right now. This is much better then manually going into command prompt every 20 minutes. Now my wife can watch TV without having me asking her to pause it for 30 seconds so I can run another test

 

Thank you for this!

[guitarlp]

I ran a bunch of tests with my current network setup and all is good.

 

I ran FCB on my main PC to my Windows Homer Server PC - 10 tests - 0 errors

 

I ran FCB on my HTPC to my Windows Home Server PC - 10 tests - 0 errors

 

I ran FCB on my HTPC to my main PC - 10 tests - 0 errors

 

So... all is well right now. New MOBO arrives tomorrow and then I'll get the unRAID server back online and start checking again. But so far, it's nice to know all my PC's are working correctly and any issues from this point forward will definately be related to the unRAID server I'm putting together.

 

I'll update with my results hopefully tomorrow night. If all is well after tomorrow then those who also have the GA-EP35-DS3R MOBO may want to start checking their systems. From what I've been reading around the net Gigabyte has some pretty hot NB's on their recent boards. You may not notice any issues... but you may want to check your data to make sure it matches what's on your PC's.

[guitarlp]

So I finished my build (new, different brand MOBO)... ran memtest for 10 hours and it passed with flying colors. I cleared all my HD's and re-formated them. I copied a movie from my mainPC to the unRAID server and ran the FCB script... it failed and gave me a bunch of errors.

 

Maybe it's the CPU that's bad. I've used it for the past year or so with my main PC and it has always worked great. Maybe when my fan was blowing on the northbridge it cooled the CPU down as well. Getting that just a little cooler allowed it to work correctly.

 

I'll try picking a new one up tomorrow to test. But if that's not it... I have no idea what else it could be hardware wise. If the CPU is replaced then I've tried at least 2 of every diferent piece of hardware.

[guitarlp]

Interesting thing I noticed too... all the errors are write related. I'll get say 7 errors during an FCB check and if I check 10 times I'll get the 7 same errors every time. I've written to a number of different disks both with parity and without and I always get write errors. Interestingly though... I never get read errors (none of the errors I've gotten have been random).

 

Could this possibly be a bad CPU? My case has great cooling and I have a good heatink/fan with arctic silver thermal paste. It says it's running cool while in the BIOS... and I did have that CPU in my XP machine (later switched to Vista) and I never had any issues (that I noticed anyways).

 

It sounds like the last thing I can check hardware wise though.

[NLS]

wish you luck - I hate this kind of problems

 

sometimes a problem is not identifiable because it has to do with certain COMBINATIONS of two or more than two components or even if TWO (or more) components lead to the same problem ... so you change something, thinking it might be the problem, the problem stays, you put the old back, change something else, problem stays so you move away from component 1 and 2 as "innocent", but in fact "1 or 2" create the same problem

 

 

[SSD]

CPU failure is extremely rare.  Have you tried running somehting like Prime95 on each of your cores?  This would stress the CPU.

 

Are you running any addon cards that could be conflicting with the DMA channel?

[limetech]

My setup is as follows:

 

Motherboard: GIGABYTE GA-MA69GM-S2H

PSU: CORSAIR CMPSU-620HX ATX12V v2.2 and EPS12V 2.91 620W

RAM: G.SKILL 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 800 (PC2 6400) Dual Channel Kit Desktop Memory

CPU: AMD Athlon 64 X2 5000+ Brisbane 2.6GHz Socket AM2 65W Dual-Core Black Edition

Back Planes (2 of them): Athena Power BP-SATA3051B 3 x 5.25"

SATA Cards (2 of them): PROMISE SATA300 TX4 PCI SATA II Controller (not being used yet... still only using the SATA off the MOBO.

Flash Drive: Lexar 1GB Firefly

 

First, I didn't notice in the thread if you suspected the power supply.  That PSU has 3 separate +12V rails, each rated at 18A max.  At least one of those rails is dedicated to the motherboard, possibly two of them.  If so, that leaves 18A for all your hard drives.  I also didn't notice how many hard drives you have, but you might try depopulating the server so there's only 2 hard drives.  Assign these as disk1 and disk2, leave parity unassigned.  Then see if write corruption occurs.

 

Second, those Athena backplanes are not very good (they are just rebranded I-Stars, which themselves might be rebranded who-knows-whats).  For test purposes you might try directly connecting the drives to the SATA cables, bypassing the drive cage.

 

Note about the Athena backplanes:

We stopped using those long ago and now use either Supermicro or Chenbro exclusively.  The problem with the Athenas is that the backplane that the drive plugs into is very thin and very poorly mounted.  We found that with some hard drives when you plug them in, the male SATA connector on the hard drive does not fully engage the female connector on the backplane.  The typical result is DMA errors in the syslog.  A workaround is to push on the drive carrier with your thumb a few times after installing it.

 

The attached images show this problem.

img1 is just a view looking down with the rear plastic fan casing removed

img2 is a view after inserting a drive tray - note the connector not fully engaged and backplane bowing

img3 is a view of my hand pushing on the backplane to press the connector in place

 

I should emphasize that not all hard drives cause this problem - but off the top of my head can't remember which ones are problematic 

[guitarlp]

CPU failure is extremely rare.  Have you tried running somehting like Prime95 on each of your cores?  This would stress the CPU.

 

Are you running any addon cards that could be conflicting with the DMA channel?

 

If the few tests I run today aren't sucessfull I'll install Windows on the unRAID server and run Prime95 to test the CPU.

 

The only cards I have are the PCI-E graphics card, Intel Pro/1000 NIC, and 2 Promise SATA300 TX4 hard drive controllers (no hard drives attached to these yet).

[guitarlp]

My setup is as follows:

 

Motherboard: GIGABYTE GA-MA69GM-S2H

PSU: CORSAIR CMPSU-620HX ATX12V v2.2 and EPS12V 2.91 620W

RAM: G.SKILL 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 800 (PC2 6400) Dual Channel Kit Desktop Memory

CPU: AMD Athlon 64 X2 5000+ Brisbane 2.6GHz Socket AM2 65W Dual-Core Black Edition

Back Planes (2 of them): Athena Power BP-SATA3051B 3 x 5.25"

SATA Cards (2 of them): PROMISE SATA300 TX4 PCI SATA II Controller (not being used yet... still only using the SATA off the MOBO.

Flash Drive: Lexar 1GB Firefly

 

First, I didn't notice in the thread if you suspected the power supply.  That PSU has 3 separate +12V rails, each rated at 18A max.  At least one of those rails is dedicated to the motherboard, possibly two of them.  If so, that leaves 18A for all your hard drives.  I also didn't notice how many hard drives you have, but you might try depopulating the server so there's only 2 hard drives.  Assign these as disk1 and disk2, leave parity unassigned.  Then see if write corruption occurs.

 

Second, those Athena backplanes are not very good (they are just rebranded I-Stars, which themselves might be rebranded who-knows-whats).  For test purposes you might try directly connecting the drives to the SATA cables, bypassing the drive cage.

 

Note about the Athena backplanes:

We stopped using those long ago and now use either Supermicro or Chenbro exclusively.  The problem with the Athenas is that the backplane that the drive plugs into is very thin and very poorly mounted.  We found that with some hard drives when you plug them in, the male SATA connector on the hard drive does not fully engage the female connector on the backplane.  The typical result is DMA errors in the syslog.  A workaround is to push on the drive carrier with your thumb a few times after installing it.

 

The attached images show this problem.

img1 is just a view looking down with the rear plastic fan casing removed

img2 is a view after inserting a drive tray - note the connector not fully engaged and backplane bowing

img3 is a view of my hand pushing on the backplane to press the connector in place

 

I should emphasize that not all hard drives cause this problem - but off the top of my head can't remember which ones are problematic 

 

I only have 5 hard drives attached and running so it shouldn't be a PSU issue. When 3 of them are powered down and I run these tests I still get data corruption.

 

I'm running some tests right now with 1 hard drive connected directly to the MOBO (no backplane) and I'll see what happens with that.

 

Between the Supermicro and Chenbro, which do you feel is the better backplane?

[limetech]

I only have 5 hard drives attached and running so it shouldn't be a PSU issue. When 3 of them are powered down and I run these tests I still get data corruption.

 

I'm running some tests right now with 1 hard drive connected directly to the MOBO (no backplane) and I'll see what happens with that.

 

It still could be the PSU - quite possible it's defective, perhaps too much ripple on one of the rails.  Bad power can case extremely strange problems.

 

Between the Supermicro and Chenbro, which do you feel is the better backplane?

 

The Chenbro (SK33502) looks better and perhaps has better EMI shielding, but has unneeded USB ports on the front and costs more.

 

The Supermicro (CSE-M35T-1B) is almost ideal & is what we prefer.

 

[guitarlp]

I only have 5 hard drives attached and running so it shouldn't be a PSU issue. When 3 of them are powered down and I run these tests I still get data corruption.

 

I'm running some tests right now with 1 hard drive connected directly to the MOBO (no backplane) and I'll see what happens with that.

 

It still could be the PSU - quite possible it's defective, perhaps too much ripple on one of the rails.  Bad power can case extremely strange problems.

 

Between the Supermicro and Chenbro, which do you feel is the better backplane?

 

The Chenbro (SK33502) looks better and perhaps has better EMI shielding, but has unneeded USB ports on the front and costs more.

 

The Supermicro (CSE-M35T-1B) is almost ideal & is what we prefer.

 

 

Thanks for the info. The Chenbro is only $10 more then the Supermicro and it does look better. Cooling wise, do they both do well? Supermicro appears to have a bigger, and from what I understand, louder fan then the Chenbro. Also, with respect to noise, which one runs quieter? I see the Chenbro drive cages have springs on them to reduce/stop ratteling... something that annoys me with the Athena backplane I currently have.

[WeeboTech]

Where do you get the chenbro?  Prices I've seen are around 160 and up.

[guitarlp]

http://www.excaliberpc.com/Chenbro_5-IN-3_STORAGE_KIT_SATA/SK33502-BK/partinfo-id-552586.html

 

$128.40.

 

If it's the better one then I'd like to put those in my system.

[guitarlp]

I got 2 of the Chenbro backplanes today and installed them in my server. I have to say, they are leaps and bounds nicer then the Athena backplanes. But it's clear these are made for a server environment because the stock fans are crazy loud . When I say crazy, I mean 5-6 times louder then the loudest case fan I've ever had the pleasure of listening to. I ordered some quieter 80mm fans to replace the stock ones with and after that I should be good.

 

In regards to the data corruption... I'm pretty sure it's a memory stick causing the issue. I've used 3 different sets of memory and 2 sets caused corruption while 1 set passed with no corruption after multiple tests. Interestingly, all 3 sets of memory pass memtest for over 10 hours with no errors. But unRAID must tax the sticks a little more when writing to the disks then memtest does. I bet if I let memtest run for a few days the bad sticks would error out eventually.

 

So, from what I've gathered so far, my data corruption was due to 2 hardware problems. Both my NB overheating and bad memory. I'm still not 100% certain at this point... but once I get some quieter fans in my new backplanes I'll be able to finish up the testing to make sure my new setup is working like it should.

[WeeboTech]

Maybe the louder fans are scaring the computer into working correctly? LOL. Or actually creating more air flow.

I've always found that bad memory shows up in memtest without question. So it may be a number of circumstances.

[Joe L.]

Interestingly, all 3 sets of memory pass memtest for over 10 hours with no errors. But unRAID must tax the sticks a little more when writing to the disks then memtest does.

I'll bet the disk access cause power supply voltage variations, and your memory was working during the memtest, but barely.  The failures showed themselves when  the power supply voltage had some "ripple"  (in other words, poor regulation of its voltages)

 

I noticed earlier you said the bios option for memory voltage was set to "auto"  Perhaps explicitly setting it to the 2.0 to 2.1 volts as indicated on this page would do iit (assuming this is your memory) http://www.newegg.com/Product/Product.aspx?Item=N82E16820231152 

 

I'm glad you are up running, but when you say memtest alone runs fine, but memory failures occur when accessing disks, or when temperatures are slightly higher, I suspect power supply issues are the root cause.  (I don't trust the "auto" voltage setting on the motherboard for the memory voltage. I'd set it explicitly for the specific memory installed)

 

Joe L.

[guitarlp]

Interestingly, all 3 sets of memory pass memtest for over 10 hours with no errors. But unRAID must tax the sticks a little more when writing to the disks then memtest does.

I'll bet the disk access cause power supply voltage variations, and your memory was working during the memtest, but barely.  The failures showed themselves when  the power supply voltage had some "ripple"  (in other words, poor regulation of its voltages)

 

I noticed earlier you said the bios option for memory voltage was set to "auto"  Perhaps explicitly setting it to the 2.0 to 2.1 volts as indicated on this page would do iit (assuming this is your memory) http://www.newegg.com/Product/Product.aspx?Item=N82E16820231152 

 

I'm glad you are up running, but when you say memtest alone runs fine, but memory failures occur when accessing disks, or when temperatures are slightly higher, I suspect power supply issues are the root cause.  (I don't trust the "auto" voltage setting on the motherboard for the memory voltage. I'd set it explicitly for the specific memory installed)

 

Joe L.

 

That makes pretty good sense and would explain why I was having issues. During memtest only the memory is being tested. But once the computer is running and doing multiple things the power going to the ram would vary and that could start to cause random errors if it wasn't getting the correct voltage.

 

With all my sticks of ram I've set the timings manually as well as the vdim voltages. This is the memory I've been trying to use (most of the time):

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231098

 

It's rated for 1.8 to 2.0 volts at 5-5-5-15. I've tried it at both 1.8 to 2.0 and it gave errors at all voltages.

 

I'm now using some Corsair memory that runs at 4-4-4-12 at 2.1 volts. My last 10 tests over the weekend showed 0 errors (uploaded a file 10 times and compared it to the file on my PC 10 times).

 

When you say the power supply could be the root of the issue, are you saying my PSU could be faulty? Or are you just saying that it's not delivering the correct voltages to the ram so I need to enter those manually in the BIOS?

 

Now you've got me curious that maybe the PSU is bad and the RAM is actually ok. Maybe the PSU just isn't delivering the correct and steady volts the ram needs at times.

 

I thought I bought a pretty good PSU though:

http://www.newegg.com/Product/Product.aspx?Item=N82E16817139002

[guitarlp]

I think Tom mentioned it before... but my Corsair PSU has triple 12V rails and he mentioned that the hard drives may only be using one of those rails. From what I've read the PSU should give power from the other rails if one of the rails is maxed out... but maybe that's not working correctly.

 

Searching on this board I've learned the most important PSU factor is the amps running on the 12v rail. For a server with lots of hard drives, this is where most of the power is used. So, I ordered the PC Power 750 watt power supply:

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16817703009

 

It has a single 12V rail rated at 60A. From what I've read on here too, Tom puts PC Power PSU's in the pre-built unRAID servers. I've also read people highly recommend these PSU's.

 

750 watts is probably overkill for an unRAID server... but better safe then sorry I'd say. I almost jumped on the 610w PC Power PSU but I was worried about startup power. I'm sure the 610watt could power 16 drives in an unRAID server... but from what I've read online the server (if full of 15 or so hard drives) would need more then 610 watts when first powering on. The only downside to the 750 watt PSU is the extra energy costs and the difference in pricing between the 610 watt ($20).

[SSD]

That makes pretty good sense and would explain why I was having issues. During memtest only the memory is being tested. But once the computer is running and doing multiple things the power going to the ram would vary and that could start to cause random errors if it wasn't getting the correct voltage.

 

I have never heard of anything like this happening.

 

With all my sticks of ram I've set the timings manually as well as the vdim voltages. This is the memory I've been trying to use (most of the time):

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231098

 

It's rated for 1.8 to 2.0 volts at 5-5-5-15. I've tried it at both 1.8 to 2.0 and it gave errors at all voltages.

 

I use this memory without any problems.  Perhaps your sticks are bad, or perhaps there is an incompatibility with your MB.  I understand the ABIT MB is more sensitive to memory than most.

 

I'm now using some Corsair memory that runs at 4-4-4-12 at 2.1 volts. My last 10 tests over the weekend showed 0 errors (uploaded a file 10 times and compared it to the file on my PC 10 times).

 

When you say the power supply could be the root of the issue, are you saying my PSU could be faulty? Or are you just saying that it's not delivering the correct voltages to the ram so I need to enter those manually in the BIOS?

 

Now you've got me curious that maybe the PSU is bad and the RAM is actually ok. Maybe the PSU just isn't delivering the correct and steady volts the ram needs at times.

 

I thought I bought a pretty good PSU though:

http://www.newegg.com/Product/Product.aspx?Item=N82E16817139002

 

Your prior statment about unRAID pushing your memory harder than memtst is simply not true.  Nothing stresses your memory more than memtst.  If memtst runs without error overnight and then you still get compare errors in unRAID, I'd bet money that it is not your memory's fault.

 

PSUs usually work until they die.  I've buried 8 or 10 of them in my 25+ years of PC expeience (most of them in the early days).  It is the most common and easiest problem to diagnose.  Computer won't turn on, replace the PSU.  Done.

 

But a non-dead but malfunctioning PSU is uncommon.  I've never had this problem personally.  But if your power is not within spec, you could certainly imagine it could do all sorts of bad things - including damaging hardware components I suppose.  There are a ton of protective circuits in place in PSUs to protect your computer.  All this unlikeliness does not mean this is not your problem!  The only way to really know is to try a different PSU.

 

I would also not rule out a bad motherboard.  The motherboard contains voltage regulators and many other components that can cause wacky problems.  MB failures are hard to diagnose.  The only real way to test is to try another one.  I've had 2 motherboards fail in my life.  On both, Windows kept crashing and it turned out to be the MB.  A third apparent MB problem turned out to be a CPU failure.  I experienced infrequent (once or twice a month) unexplained crashes, and could not figure it out.  I finally got frustrated and tried reloading the OS, and got different errors at different points in the install.  (Sometimes the install actually completed successfully).  Moved the CPU to a completely different MB with separate PSU, memory, etc.  The install problem got worse - happened within minutes of starting every time.  Sent the CPU back to Intel (within months of warranty expiration) and got a new one.  Worked like a champ on the original MB.

 

Do you have another computer (e..g, your desktop) that you could temporarily disconnect from its hard disks and hook up your backplane and boot unRAID?  Use the same cables from your unRAID box.  This would rule out, once and for all, any cabling issues, backplane issues, or drive problems.  If you are getting the same kinds of problems you will KNOW you have a problem within your drive subsystem.  That would be my next step.  If this works, you have a working configuration to start from.  This is a good thing (see below).

 

In diagnosing problems, I always try to get to a WORKING configuration and add suspect components into the mix until I induce a failure.  This is the only way to go in diagnosing complex problems.  Starting from a failure and trying to induce success, if it doesn't happen quickly, is a neverending process in my experience.  You never rule out anything.

 

FWIW - when problems get this hard to diagnose, multiple simultaneous causes is sometimes the reason.  Don't limit your thinking to a single problem. 

 

[WeeboTech]

That makes pretty good sense and would explain why I was having issues. During memtest only the memory is being tested. But once the computer is running and doing multiple things the power going to the ram would vary and that could start to cause random errors if it wasn't getting the correct voltage.

 

I have never heard of anything like this happening.

 

Me either.

 

With all my sticks of ram I've set the timings manually as well as the vdim voltages. This is the memory I've been trying to use (most of the time):

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231098

 

It's rated for 1.8 to 2.0 volts at 5-5-5-15. I've tried it at both 1.8 to 2.0 and it gave errors at all voltages.

 

I use this memory without any problems.  Perhaps your sticks are bad, or perhaps there is an incompatibility with your MB.  I understand the ABIT MB is more sensitive to memory than most.

 

 

I've found it to be sensitive also. I just adjusted the memory timing so it was not so aggressive.

Stock bios on the ABIT overclocks the memory slightly, but on the Gigabyte I've not heard of issues.

 

 

Your prior statment about unRAID pushing your memory harder than memtst is simply not true.  Nothing stresses your memory more than memtst.  If memtst runs without error overnight and then you still get compare errors in unRAID, I'd bet money that it is not your memory's fault.

 

I'm going to agree here also.

In all my years of using memtest, if I suspected memory issues, it always brought them into light.

 

 

PSUs usually work until they die.  I've buried 8 or 10 of them in my 25+ years of PC expeience (most of them in the early days).  It is the most common and easiest problem to diagnose.  Computer won't turn on, replace the PSU.  Done.

 

But a non-dead but malfunctioning PSU is uncommon.  I've never had this problem personally.  But if your power is not within spec, you could certainly imagine it could do all sorts of bad things - including damaging hardware components I suppose.  There are a ton of protective circuits in place in PSUs to protect your computer.  All this unlikeliness does not mean this is not your problem!  The only way to really know is to try a different PSU.

 

I have to agree here too. However a bad psu would cause all sorts of spurious errors.

If I remember correctly, the machine was opened, a fan was put to blow on the guts and it passed tests???

If this is so, then it's probably a heat issue.

With only 5 hard drives on such a powerful power supply.... I start to wonder.

I've had 6-8 drives spinning 24x7 on a 300 watt power supply for years before I upgraded the machine.

Half high speed SCSI, Half lower RPM IDE. The machine was my work horse workstation.

Compressing MP3's all the time, VMware with instances of windows and Linux.

I never had any of these issues. Then again, I was using a server class supermicro with registered ram.

 

 

In diagnosing problems, I always try to get to a WORKING configuration and add suspect components into the mix until I induce a failure.  This is the only way to go in diagnosing complex problems.  Starting from a failure and trying to induce success, if it doesn't happen quickly, is a neverending process in my experience.  You never rule out anything.

 

FWIW - when problems get this hard to diagnose, multiple simultaneous causes is sometimes the reason.  Don't limit your thinking to a single problem. 

 

From what I remember of this long thread, Many parts were swapped out.

Almost to the point of 100% parts or so.

Perhaps there is some incompatibility with certain memory which is exasperated with heat on the NB.

Remember back a number of posts, I inserted a link of other people having heat issues on the NB.

[limetech]

As a final conjecture: all data movements between memory, the hard drives, and ethernet are done using DMA.  This is going to use a different set of circuits into and out of memory via North Bridge, than the  CPU<->Memory path which is the only path tested by memtest.  But if I had to bet, I'd agree with Joe and say it's probably the PSU.

[guitarlp]

Your prior statment about unRAID pushing your memory harder than memtst is simply not true.  Nothing stresses your memory more than memtst.  If memtst runs without error overnight and then you still get compare errors in unRAID, I'd bet money that it is not your memory's fault.

 

PSUs usually work until they die.  I've buried 8 or 10 of them in my 25+ years of PC expeience (most of them in the early days).  It is the most common and easiest problem to diagnose.  Computer won't turn on, replace the PSU.  Done.

 

But a non-dead but malfunctioning PSU is uncommon.  I've never had this problem personally.  But if your power is not within spec, you could certainly imagine it could do all sorts of bad things - including damaging hardware components I suppose.  There are a ton of protective circuits in place in PSUs to protect your computer.  All this unlikeliness does not mean this is not your problem!  The only way to really know is to try a different PSU.

 

I would also not rule out a bad motherboard.  The motherboard contains voltage regulators and many other components that can cause wacky problems.  MB failures are hard to diagnose.  The only real way to test is to try another one.  I've had 2 motherboards fail in my life.  On both, Windows kept crashing and it turned out to be the MB.  A third apparent MB problem turned out to be a CPU failure.  I experienced infrequent (once or twice a month) unexplained crashes, and could not figure it out.  I finally got frustrated and tried reloading the OS, and got different errors at different points in the install.  (Sometimes the install actually completed successfully).  Moved the CPU to a completely different MB with separate PSU, memory, etc.  The install problem got worse - happened within minutes of starting every time.  Sent the CPU back to Intel (within months of warranty expiration) and got a new one.  Worked like a champ on the original MB.

 

Do you have another computer (e..g, your desktop) that you could temporarily disconnect from its hard disks and hook up your backplane and boot unRAID?  Use the same cables from your unRAID box.  This would rule out, once and for all, any cabling issues, backplane issues, or drive problems.  If you are getting the same kinds of problems you will KNOW you have a problem within your drive subsystem.  That would be my next step.  If this works, you have a working configuration to start from.  This is a good thing (see below).

 

In diagnosing problems, I always try to get to a WORKING configuration and add suspect components into the mix until I induce a failure.  This is the only way to go in diagnosing complex problems.  Starting from a failure and trying to induce success, if it doesn't happen quickly, is a neverending process in my experience.  You never rule out anything.

 

FWIW - when problems get this hard to diagnose, multiple simultaneous causes is sometimes the reason.  Don't limit your thinking to a single problem.

 

 

I'm not saying the memory isn't being "tortured" enough during memtest... I'm just saying I agree wtih Joe that maybe during memtest the voltages going to the memory are consistant. But once hard drives are being accessed and more stuff is going on the voltages going to the memory aren't consistant and that could be causing the random errors the memory is putting out.

 

PSU's do usually work until they diie... but they can also cause problems while still working (for example... computers shutting down or re-starting randomly). While I have not experienced a PSU causing data corruption (until now maybe), I have read about it several times. In all my years of building PC's the only time I've had a bad PSU is when the computer stopped booting like you said. But while that is common, it doesn't mean a PSU can't be bad for other reasons.

 

I've done a lot of swapping and changing of parts to narrow down the problem. With my last MOBO it was clear that the NB was part of the problem. Cooling it down caused less errors (didn't stop all the errors... just most of them). On that board I would get write errors and read errors (some errors were consistant and some were random). On my new, different brand MOBO, I'm no longer getting read errors (all the errors I see are the same each test after a file is uploaded to the unRAID server. After replacing the memory with a few different sticks and upping the voltage to 2.1 volts, I appear to no longer be getting data corruption.

 

Like you said, I had multiple sources of hardware problems causing my issue. I know for 100% certainty that the NB on my last MOBO was bad. But, there's something memory related going on as well. Either 2 sets of memory out of 3 are bad or the PSU is bad (or maybe even both). Maybe the memory is fine but the PSU isn't delivering consistent voltages so random errors start happening. Maybe with a better PSU the "bad" memory sticks will work fine without any issues.

 

 

[SSD]

 

... Me either.

 

... I've found it to be sensitive also.

 

... I'm going to agree here also.

 

... I have to agree here too.

 

What!?? Agreement from WeeboTech!  <faints>

 

Now if Tom would change the name on the Restore button ... THAT would be a miracle!

 

[WeeboTech]

LOL.. Am I that bad? 

[SSD]