I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
Regards,
Keith
It wouldn't harm you to do this?
Jack
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Keith Roberts Sent: Wednesday, November 03, 2010 9:48 PM To: centos Subject: [CentOS] Memtest86+ running time
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
Regards,
Keith
-- In theory, theory and practice are the same; in practice they are not.
This email was sent from my laptop with Centos 5.5 _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Keith Roberts Sent: Wednesday, November 03, 2010 9:48 PM To: centos Subject: [CentOS] Memtest86+ running time
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
Regards,
Keith
----- Message Text ----- On Wed, 3 Nov 2010, Jack Davies wrote:
To: 'CentOS mailing list' centos@centos.org From: Jack Davies me@jackdavies.org.uk Subject: Re: [CentOS] Memtest86+ running time
It wouldn't harm you to do this?
Jack
OK.
I'll let each memory module run overnight if they don't show any errors.
I dusted and hovered out the machine case, and hovered the memory slots using a small paint brush and vacuum cleaner.
The gold contacts I degreased with methylated spirit on a cotton bud. Just to make sure they are squeaky clean.
Keith
On 11/3/2010 4:47 PM, Keith Roberts wrote:
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
Yes, I've seen a machine where it ran over a weekend before catching an error (which turned out to have corrupted the filesystem randomly too).
On Wed, Nov 3, 2010 at 2:47 PM, Keith Roberts keith@karsites.net wrote:
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
I had a system that started crashing randomly. I ran memtest overnight (about 10 hrs) but it did not report any errors. Next time I extended the run to 18 hrs or so and finally saw errors. Replacing the RAM solved the crash problem. So, I would recommend running memtest for one full day.
Akemi
On 03/11/10 22:03, Akemi Yagi wrote:
On Wed, Nov 3, 2010 at 2:47 PM, Keith Roberts keith@karsites.net wrote:
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
I had a system that started crashing randomly. I ran memtest overnight (about 10 hrs) but it did not report any errors. Next time I extended the run to 18 hrs or so and finally saw errors. Replacing the RAM solved the crash problem. So, I would recommend running memtest for one full day.
Akemi _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
I agree. Usually, when systems crash due to possibly memory related errors I let them run testing for 24 hours, or if they have lots of memory possibly longer (48 hours) to allow all memory to be fully tested. :-)
Jake Shipton wrote:
On 03/11/10 22:03, Akemi Yagi wrote:
On Wed, Nov 3, 2010 at 2:47 PM, Keith Roberts keith@karsites.net wrote:
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
I had a system that started crashing randomly. I ran memtest overnight (about 10 hrs) but it did not report any errors. Next time I extended the run to 18 hrs or so and finally saw errors. Replacing the RAM solved the crash problem. So, I would recommend running memtest for one full day.
I agree. Usually, when systems crash due to possibly memory related errors I let them run testing for 24 hours, or if they have lots of memory possibly longer (48 hours) to allow all memory to be fully tested. :-)
And make sure you button the cabinet back up, with all covers in place and put it back on the rack where it normally sits. Running this test with the case open or sitting in free air is a waste of time unless that is how the system usually operates. All memory testing has to be done under normal working conditions to get the maximum benefit.
Bob McConnell N2SPP
Bob McConnell wrote:
Jake Shipton wrote:
Akemi Yagi wrote:
Keith Roberts wrote:
I have run one full test and got no errors on the memory module. Is it worth keeping it running overnight, just to see if temperature changes will afect the test?
I had a system that started crashing randomly. I ran memtest overnight (about 10 hrs) but it did not report any errors. Next time I extended the run to 18 hrs or so and finally saw errors. Replacing the RAM solved the crash problem. So, I would recommend running memtest for one full day.
I agree. Usually, when systems crash due to possibly memory related errors I let them run testing for 24 hours, or if they have lots of memory possibly longer (48 hours) to allow all memory to be fully tested. :-)
And make sure you button the cabinet back up, with all covers in place and put it back on the rack where it normally sits. Running this test with the case open or sitting in free air is a waste of time unless that is how the system usually operates. All memory testing has to be done under normal working conditions to get the maximum benefit.
And, you're using reasonable precautions against ESD, right? http://www.intel.com/design/packtech/ch_06.pdf The sad thing is, most ESD problems don't show up right away.
On Wed, 3 Nov 2010, Bob McConnell wrote:
_snip_
And make sure you button the cabinet back up, with all covers in place and put it back on the rack where it normally sits. Running this test with the case open or sitting in free air is a waste of time unless that is how the system usually operates. All memory testing has to be done under normal working conditions to get the maximum benefit.
Bob McConnell N2SPP
Of course. I noticed that with the side cover off the hard drive is not as cool. Probably due to the air flow not being drawn over the drive, but in from the side of the cabinet.
It's an ASROCK K7S8XE Rev. 3.01 M/B with 3 DDR(1) memory slots.
2 x 1GB PC2100 DIMMS, and a 512MB one.
I tested and replaced the BIOS battery as well. That was getting a bit low, just under 3 volts. Been in the machine since Jan 2006.
So far, testing each memory module in slot 1 has clocked no errors. Each test ran for at least 20 hours.
I now have the 512MB module in slot 2, to test that slot.
Then I'll put it in slot 3 and test that one.
The first 1GB DIMM was tested with the side cover off.
So I might test that again, with the side cover on this time.
Once these test are completed, I'll post the results.
Thanks for all the help so far.
Keith Roberts
On Sun, 7 Nov 2010, Keith Roberts wrote:
To: CentOS mailing list centos@centos.org From: Keith Roberts keith@karsites.net Subject: Re: [CentOS] Memtest86+ running time
On Wed, 3 Nov 2010, Bob McConnell wrote:
_snip_
And make sure you button the cabinet back up, with all covers in place and put it back on the rack where it normally sits. Running this test with the case open or sitting in free air is a waste of time unless that is how the system usually operates. All memory testing has to be done under normal working conditions to get the maximum benefit.
Bob McConnell N2SPP
Of course. I noticed that with the side cover off the hard drive is not as cool. Probably due to the air flow not being drawn over the drive, but in from the side of the cabinet.
It's an ASROCK K7S8XE Rev. 3.01 M/B with 3 DDR(1) memory slots.
2 x 1GB PC2100 DIMMS, and a 512MB one.
I tested and replaced the BIOS battery as well. That was getting a bit low, just under 3 volts. Been in the machine since Jan 2006.
So far, testing each memory module in slot 1 has clocked no errors. Each test ran for at least 20 hours.
I now have the 512MB module in slot 2, to test that slot.
Then I'll put it in slot 3 and test that one.
The first 1GB DIMM was tested with the side cover off.
So I might test that again, with the side cover on this time.
Once these test are completed, I'll post the results.
Thanks for all the help so far.
I have done the following tests on the memory.
Tested each memory module in DDR1 slot for at least 20 hours.
Tested each slot with the 512MB module for another 20 hours. Should I have used the 1GB module for this?
Put a 1GB module into slot 3, and tested again for 20+ hours OK.
Added another 1GB module to slot DDR2 and there are literally millions of errors!
Do I need to put the memory modules back in the lowest numbered slots?
I rebooted the machine, and it gave me a stream of errors within 30 seconds.
Rebooted again, and has run for 10 minutes without any errors.
Would be an indication of an unstable power supply?
Keith
On Wed, 3 Nov 2010, Bob McConnell wrote:
_snip_
And make sure you button the cabinet back up, with all covers in place and put it back on the rack where it normally sits. Running this test with the case open or sitting in free air is a waste of time unless that is how the system usually operates. All memory testing has to be done under normal working conditions to get the maximum benefit.
Bob McConnell N2SPP
Continuing on with my memory testing saga ;)
The Memtest86+ program I'm using is from the UltimateBootCD.
Could it be Memtest86+ 4.10 that is corrupted?
Reason being it locked up after running for 25 minutes. There were no patterns being written to memory, and no errors on the screen. The '+' was flashing but that's all.
So I did a warm reboot, and started Memtest86+ again, and this time it did not even fully load into memory and start running. I just had an incomplete Memtest86+ screen sitting there doing nothing. Should have took a picture - sorry.
Next time rebooted it ran fine. Something really weird is going on here.
I'm now running my version of Memtest86+ (4.00) that I downloaded and md5sum checked myself. This is the version of Memtest86+ I originally used to check the memory in 2006 when the memory was purchased.
These recent memory tests were done with the Memtest86+ from the UBCD, as I could not find the CD I burnt. Dug it out now :)
Regards,
Keith
Tested each memory module in DDR1 slot for at least 20 hours.
Tested each slot with the 512MB module for another 20 hours. Should I have used the 1GB module for this?
So it appears that each memory module works fine if it's the only module installed - in whatever slot.
I found this helpfull guide for using Memtest86+:
http://www.overclockers.com/forums/showthread.php?t=409152
Memtest86+ website: http://www.memtest.org/
Memtest86+ Forum: http://forum.canardpc.com/forumdisplay.php?f=73
Cannot find a mailing list though.
My machine has been known to freeze on occassions, which I put down to software issues while running Fedora.
The 2 1GB memory modules are 1 low density and 1 high density. Both are PC2100.
They work fine on their own, but don't want to work together.
I'm now running Memtest no. 5 using the 1GB low density module, and the 512MB module, which is also low density.
I have put a digital multimeter on the power supply, and overnight while testing it registers ~ 4.97 - 4.98 volts on the 5vdc line, with a stable reading. ie it changes slowly between those 2 values. So it looks like the PSU is fine.
I have a sneaking suspicion that it's the mismatched low and high density memory modules that are the problem?
Kind Regards,
Keith Roberts
On 11/12/10 10:46 AM, Keith Roberts wrote:
I have a sneaking suspicion that it's the mismatched low and high density memory modules that are the problem?
that sounds quite likely. some of those modules have two banks on the single module, and I seriously doubt a system is going to like a single bank and a dual bank module in a dual channel environment.
I got in the habit a long time ago of *always* using matched memory modules.
On Fri, 12 Nov 2010, John R Pierce wrote:
To: CentOS mailing list centos@centos.org From: John R Pierce pierce@hogranch.com Subject: Re: [CentOS] Memtest86+ running time
On 11/12/10 10:46 AM, Keith Roberts wrote:
I have a sneaking suspicion that it's the mismatched low and high density memory modules that are the problem?
that sounds quite likely. some of those modules have two banks on the single module, and I seriously doubt a system is going to like a single bank and a dual bank module in a dual channel environment.
I got in the habit a long time ago of *always* using matched memory modules.
I've learnt my lesson now John.
Well at least I won't mix hi and lo density modules again.
Here's the link for my Asrock m/b:
http://www.crucial.com/uk/store/listparts.aspx?model=K7S8XE%20R3.0
It does not support dual channel memory.
I'll keep y'all posted on the outcome of this saga :)
Kind Regards,
KR
On 11/12/10 12:21 PM, Keith Roberts wrote:
On Fri, 12 Nov 2010, John R Pierce wrote:
that sounds quite likely. some of those modules have two banks on the single module, and I seriously doubt a system is going to like a single bank and a dual bank module in a dual channel environment.
I got in the habit a long time ago of *always* using matched memory modules.
I've learnt my lesson now John.
Well at least I won't mix hi and lo density modules again.
Here's the link for my Asrock m/b:
http://www.crucial.com/uk/store/listparts.aspx?model=K7S8XE%20R3.0
It does not support dual channel memory.
no, but look at these constraints on the memory vs cpu FSB speeds http://www.asrock.com/mb/overview.asp?Model=K7S8XE
*CPU* - Socket 462, supporting AMD Athlon, Athlon XP, Duron - FSB 400/333/266/200 MHz *Chipset* - SiS^® 748 *Memory* - DDR non-ECC, un-buffered memory - DDR400, Max. capacity of system memory: 1GB (FSB 400/333 MHz)* - DDR333, Max. capacity of system memory: 2GB (FSB 400/333/266/200 MHz) - DDR266/200, Max. capacity of system memory: 3GB (FSB 333/266/200 MHz)
*According to SiS^® official document, SiS^® 748 chipset has limitation DDR supports: CPU at FSB 200MHz mode, it will NOT support DDR400 CPU at FSB 333MHz mode, it will NOT support DDR200 CPU at FSB 400MHz mode, it will NOT support DDR200/DDR266
I don't know which CPU you have, but, for instance, if its a 400Mhz CPU, you can't use 200/266 memory, and if you're not using 200/266 memory, you can't have 3 dimms at all. Your PC2100 is 266Mhz memory.
And let just toss out that I personally would scrap any motherboard containing SiS (or VIA) chipsets in a blink if it was giving me any trouble at all.
On Fri, 12 Nov 2010, John R Pierce wrote:
To: CentOS mailing list centos@centos.org From: John R Pierce pierce@hogranch.com Subject: Re: [CentOS] Memtest86+ running time
On 11/12/10 12:21 PM, Keith Roberts wrote:
On Fri, 12 Nov 2010, John R Pierce wrote:
that sounds quite likely. some of those modules have two banks on the single module, and I seriously doubt a system is going to like a single bank and a dual bank module in a dual channel environment.
I got in the habit a long time ago of *always* using matched memory modules.
I've learnt my lesson now John.
Well at least I won't mix hi and lo density modules again.
Here's the link for my Asrock m/b:
http://www.crucial.com/uk/store/listparts.aspx?model=K7S8XE%20R3.0
It does not support dual channel memory.
no, but look at these constraints on the memory vs cpu FSB speeds http://www.asrock.com/mb/overview.asp?Model=K7S8XE
*CPU* - Socket 462, supporting AMD Athlon, Athlon XP, Duron
- FSB 400/333/266/200 MHz
*Chipset* - SiS^® 748 *Memory* - DDR non-ECC, un-buffered memory
- DDR400, Max. capacity of system memory: 1GB (FSB 400/333 MHz)*
- DDR333, Max. capacity of system memory: 2GB (FSB 400/333/266/200 MHz)
- DDR266/200, Max. capacity of system memory: 3GB (FSB 333/266/200 MHz)
*According to SiS^® official document, SiS^® 748 chipset has limitation DDR supports: CPU at FSB 200MHz mode, it will NOT support DDR400 CPU at FSB 333MHz mode, it will NOT support DDR200 CPU at FSB 400MHz mode, it will NOT support DDR200/DDR266
I don't know which CPU you have, but, for instance, if its a 400Mhz CPU, you can't use 200/266 memory, and if you're not using 200/266 memory, you can't have 3 dimms at all. Your PC2100 is 266Mhz memory.
And let just toss out that I personally would scrap any motherboard containing SiS (or VIA) chipsets in a blink if it was giving me any trouble at all.
I wanted to install the maximum amount of memory, so I'm using a 1.3GHz Athlon, with a 133 FSB speed IIRC.
I think the problem is mixing the Hi and Lo density modules.
I've run Memtest86+ for over 40 hours without errors, using the 1GB & 512MB sticks together.
I'll probably put in the maximum amount of Lo density PC2100 memory, which is 3GB.
Also I'm in the process of moving all my websites onto some cloud hosting that has 99.99% uptime.
Then I'll just use this machine for development, and ftp things to the cloud host.
That way my websites will stay online if I get any more problems with my home based server.
Apologies for my websites being down for so long.
And thanks for all the replies so far.
The ultimate boot cd has a bootable Linux on it with lots of usefull diagnostic tools on it.
Regards,
Keith
Keith Roberts wrote:
On Fri, 12 Nov 2010, John R Pierce wrote:
From: John R Pierce pierce@hogranch.com On 11/12/10 12:21 PM, Keith Roberts wrote:
On Fri, 12 Nov 2010, John R Pierce wrote:
that sounds quite likely. some of those modules have two banks on the single module, and I seriously doubt a system is going to like a single bank and a dual bank module in a dual channel environment.
I got in the habit a long time ago of *always* using matched memory modules.
I've learnt my lesson now John.
Well at least I won't mix hi and lo density modules again.
Here's the link for my Asrock m/b:
http://www.crucial.com/uk/store/listparts.aspx?model=K7S8XE%20R3.0
It does not support dual channel memory.
no, but look at these constraints on the memory vs cpu FSB speeds http://www.asrock.com/mb/overview.asp?Model=K7S8XE
<snip>
I wanted to install the maximum amount of memory, so I'm using a 1.3GHz Athlon, with a 133 FSB speed IIRC.
I think the problem is mixing the Hi and Lo density modules.
<snip> May I recommend looking at the docs that tell you what order the DIMMs need to be put into the board in? If you've got four slots, for example, I'll bet that it's 1 and 2, then 3 and 4, or maybe 1 and 3, then 2 and 4. I'd feel safe putting identical DIMMs in each pair, so that the first two are the bigger ones, and the second pair are smaller. A number of our servers *require* pairs, so if I start getting a lot of ECC errors, they send me a pair to replace them.
mark
On Tue, 16 Nov 2010, m.roth@5-cent.us wrote:
To: CentOS mailing list centos@centos.org From: m.roth@5-cent.us Subject: Re: [CentOS] Memtest86+ running time
<snip>
May I recommend looking at the docs that tell you what order the DIMMs need to be put into the board in? If you've got four slots, for example, I'll bet that it's 1 and 2, then 3 and 4, or maybe 1 and 3, then 2 and 4. I'd feel safe putting identical DIMMs in each pair, so that the first two are the bigger ones, and the second pair are smaller. A number of our servers *require* pairs, so if I start getting a lot of ECC errors, they send me a pair to replace them.
mark
Thanks for that Mark.
The board acually only has 3 DIMM slots. Each slot will take 1GB maximum of RAM.
I do understand what you are saying about matching DIMM's in pairs on a 4 slot board.
Main reasons for choosing this board was:
a) price b) 5 PCI slots - so plenty of room for adding other stuff.
Kind Regards,
Keith Roberts
--- In theory, theory and practice are the same; in practice they are not.
This email was sent from my laptop with Centos 5.5
Keith Roberts wrote:
On Tue, 16 Nov 2010, m.roth@5-cent.us wrote:
From: m.roth@5-cent.us
<snip>
May I recommend looking at the docs that tell you what order the DIMMs need to be put into the board in? If you've got four slots, for example, I'll bet that it's 1 and 2, then 3 and 4, or maybe 1 and 3, then 2 and 4. I'd feel safe putting identical DIMMs in each pair, so that the first two are the bigger ones, and the second pair are smaller. A number of our servers *require* pairs, so if I start getting a lot of ECC errors, they send me a pair to replace them.
Thanks for that Mark.
The board acually only has 3 DIMM slots. Each slot will take 1GB maximum of RAM.
I do understand what you are saying about matching DIMM's in pairs on a 4 slot board.
I think I'd check the docs anyway - the order that you put DIMMs in may matter, and it may be happy with 1G,1G,512M, and not 512M, 1G, 1G, or whatever.
Main reasons for choosing this board was:
a) price b) 5 PCI slots - so plenty of room for adding other stuff.
Well, if it's yours, of course, price. Slots... *sigh*, gonna have to look at PCI-Express soon....
mark
-
Akemi Yagi -
Bob McConnell -
cpolish＠surewest.net -
Jack Davies -
Jake Shipton -
John R Pierce -
Keith Roberts -
Les Mikesell -
m.roth＠5-cent.us