New system has boot time issues - memory related

[The Gorn]

Here's what has happened.

My system is as follows: (main system specs of interest to this situation listed)

Motherboard: EVGA X58 SLI
Processor: Intel i7 - 920
Memory: 12 GB, arranged as 6 x 2048 MB. Memory is supplied as a kit, OCZ technology brand OCZ3G1333LV12GS; 9 CAS; PC10666.

The system runs on a voltage controlled UPS (And anyway has never been through an electrical storm yet.)

So what has happened is this:

From New Year's (approx) when I built it through about a week ago the system has been rock solid with no problems. No blue screens ever and no problems starting ever.

Since last week, with no changes to the system that I can recall, the system has developed at first an intermittent and then a very consistent problem with startup. When I turn the PC on the BIOS never beeps and it never progresses to a BIOS screen nor to boot load. It gets stuck in a loop where you hear it go "tic" every 5-7 seconds. When this first started I was able to get the system started by turning the "hardware" power switch off for a few minutes and then restarting. After 1 or 2 cycles it would start. Today it didn't start at all.

So, I dug deeper. I pulled the cover off. This mobo has a two digit LED display mounted on the board. When the system enters this stuck mode on startup the display cycles through the codes 68, 69, and F3 repeatedly.

I tried resetting the CMOS with a button provided on the system board. I also tried pulling out the CMOS lithium battery and waiting for about 1/2 hour. Nothing new with either approach.

So I googled and found that "EVGA X58 SLi POST 68" is a recognized search term. The "68" hex code is apparently related to memory problems on this board.

Several threads I found on overclocker sites and also on the EVGA site recommended RMAing the board. (Nothing I "love" better than ripping apart a PC to pull out a main board... nothing.  )

But some threads I saw were giving me some ideas. The memory angle seemed worth pursuing. This board has a specific sequence in which you populate the 6 memory slots. There are three red and three black sockets. You populate the red ones first, then the black. Also, your DIMMs should be matched in terms of manufacturer lot, in each group of 3.

So I pulled all of the DIMMs out of the machine. The serial numbers were all the same (the memory was a kit of 6) except that the serial numbers ended in -1 through -6.

So I tried some experiments. I first put in -1 through -3. The computer POSTed and booted normally but only reported 4 GB, not 6 GB. I then replaced those DIMMs with the -4 through -6 ones. The system again came up normally and reported 6 GB this time.

Ah-hah.

So I populated all of the slots. Now the computer boots up and reports 10 GB, not the expected 12 GB.

I then tried a supported 4 DIMM configuration. I now suspected 1 through 3 of being bad or contributing. I used 4 through 6 in three slots and one of the first three DIMMS in the remaining fourth slot.

I found that the module ending in -2  would consistently "poison" the system so it would not boot and would get stuck.

I checked the warranty info. OCZ memory has a lifetime warranty. Their RMA process seems to be well regarded by most people I checked with. I just don't want to do without a system for the 1-2 weeks it will take.

Do any of the hard core techies here have any recommendations on this? A few more notes:

1) I ran the Windows memory test at startup a couple of days ago and everything checked out OK. (However, .. yeah, it's a Microsoft memory test.)

2) I do not believe that I have ever flashed this BIOS. Think that would have a bearing?

3) What about the timing parameters like "CAS latency"? I have no clue here. The memory is labeled "9-9-9" but this system supposedly runs at "7-7-7". (I think.) I do not know enough to evaluate this properly. And I have never messed with these parameters in the BIOS nor in the EVGA "tuning" utility they provide.

So right now I am OK. I am running with 8GB (the system supports either 1, 3, 4, or 6 sticks.) Of course I am not getting my money's worth... I am reluctant to RMA until I know that I have exhausted any possibilities with the BIOS configuration.

Thanks for any thoughts or guidance.

[The Gorn]

I cleared the CMOS by reloading defaults.

I flashed the BIOS to the current version (dated 3/30/2010).

From reading it looked like I was inadvertently "overclocking" the memory I am using. I changed the three delays (CL, tRCD, tRP) from 7-7-7 to to 9-9-9.

No difference with the "bad' ram stick. The -2 DIMM still isn't recognized.

[The Gorn]

Surely someone else here has dealt with a similar flavored issue?

<reflecting>This is exactly the type of stuff that you deal with every day as a computer service technician and no wonder I hated that work so much. I observe that my own blood pressure goes through the roof on every issue like this when I know that I am the only one able to deal with it, even when it's not my own computer. And it's much worse when it is. There is an initial strong feeling of helplessness combined with "mission" - "must... make ...this ...work."</reflecting>

[pxsant]

I downloaded the manual for your motherboard for reference.

First a simplified definition of CAS latency. It is not  a specific frequency but is actually the number of clock ticks between the request for data and the availability of stable output for reading.  So the correct CAS latency setting is dependent on the clock speed you are running the memory slots.

I see in the manual that the bios memory speed setting  is either "Auto" or one of three fixed settings.  I assume that you have it set to "Auto" (if not, set it to auto).  Look up the specs of the memory you have and see what the clock speed spec is.  Try setting it to the specific correct frequency instead of auto to see if there is any difference in symptoms.

You can also specifically set the CAS latency (in your bios, this appears to be the tCL setting) to spec for the memory with the clock speed set correctly.

I'm guessing that you have already done all of this.  In fact, you have done pretty much what I would have done to isolate a memory issue.    It seems pretty clear that you simply have a bad module. 

For memory testing I use Memtest from HCI Design at http://hcidesign.com/memtest/.  I use the free version but they also have a pro version.  The new version 4.0 supports 64 bit systems.

[The Gorn]

I downloaded the manual for your motherboard for reference.

The actual manual, what a concept.

I confess to NOT referring to it. I chased postings on message boards that referred to the same components. I zeroed in on the memory kind of late in the process. Initially I thought I had a bad mobo.

I will take a second look at the settings you're describing. I was already aware that the numbers like CAS latency were a clock tick and not a frequency value - IE, lower means "faster" and that is what I suspected with the difference between the modules' specs and what the BIOS and the tuning utility were telling me. I had all three numbers on "auto" before this and they were being realized as 7s at run time. I forced all three manually in the BIOS to 9 and there is no difference either in operation or in the results of trying the memory.

Thanks very much - in buckets - for the validation of my methodology.

[pxsant]

Just a clarification on my "Auto" comment.

Assuming the memory frequency is set to "Auto", just what frequency is it set to?  It might be set to a higher frequency than the module specs but who knows.  I would set this to the specific frequency for the modules you have rather than "Auto".  That way you are absolutely sure you are not overclocking the memory and the CAS settings are actually correct for the memory.

[The Gorn]

Ok. Will do. Thanks again.

[pxsant]

Another useful troubleshooting utility I use is CPUID at http://www.cpuid.com/cpuz.php.

The memory section of this utility will give you good info on your memory including the correct settings and the proper settings if you change the frequency.

[The Gorn]

Another useful troubleshooting utility I use is CPUID at http://www.cpuid.com/cpuz.php.

The memory section of this utility will give you good info on your memory including the correct settings and the proper settings if you change the frequency.

I just ran that. It turns out to be the backbone of the "EVGA E-Leet Tuning Utility" that comes with this motherboard. I guess it is an open source kit that OEMs can customize.

What do you recommend for a really good, comprehensive memory test?

[benali72]

Hiren's Boot CD has several memory testers you can run all day for extensive ram tests.  Information at -- http://www.hiren.info/pages/bootcd

Download it free from -- http://www.hirensbootcd.net/download.html?ver=10.4

I usually just use Memtest86+, which is bundled with both Ubuntu and Hirens.

[pxsant]

The link I mentioned in the earlier post is a pretty good memory tester.  It has the ability to spawn multiple copies to test different memory regions at the same time.  The base version is free but you can buy a version which can make a boot CD for $14.  That way the OS is out of the way and all memory is tested.  The link is http://hcidesign.com/memtest/

[DarkHumour]

Surely someone else here has dealt with a similar flavored issue?

<reflecting>This is exactly the type of stuff that you deal with every day as a computer service technician and no wonder I hated that work so much. I observe that my own blood pressure goes through the roof on every issue like this when I know that I am the only one able to deal with it, even when it's not my own computer. And it's much worse when it is. There is an initial strong feeling of helplessness combined with "mission" - "must... make ...this ...work."</reflecting>

My ocd anxiety riddled nature is either complimentary to tech support as I must find the answer or it will not work.  (AND/OR) This is why I'm miserable much of the time.  "Good enough" is NOT good enough. 

I cannot seem to turn that methodology off completely.  I still wonder about the answers I got wrong on my security+ exam in spite of passing the thing.  At least this particular obsession is fading with time or simply being drowned out by other things...

On topic sortof - it used to annoy me when I found clients mixing and matching memory types and brands in the same system.  Gee, I wonder why it reboots or acts weird ?   One refused to buy system branded ram. (And of course the OEM refused to support anything but their own in a system).  This client got unheard of generic dimms/simms instead.  The OEM and Generics didn't work well together.  My solution?  Every other system had a pair of branded ram and the others had the off-brand. 

If the generics crapped out then that could have indicted their decision to go cheap.  As far as I know there were no problems (except the BIOS underreporting what was installed if you set the memory hole...but that was an OEM issue).

WinPE 3.0 also has a basic memory tester which is likely the same utility built into Vista or Windows 7.0

DarkHumour

[benali72]

pxsant -- thanks for your good links on memory and CPU testing.