memtest86+ question with new board and memory...

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

memtest86+ question with new board and memory...

Post by tbessie » Sun Oct 04, 2009 4:54 pm

Hey folks...

Still having "fun" with this new file server build.

As noted elsewhere, my new rig has:

Intel DQ45CB Desktop Motherboard
Intel Celeron E3200 Processor (compatible with above board)
2 sticks of Corsair memory (2gb each), model TWIN2X4096-6400C5 (specs match requirements for board)

Simple enough. The memory was recommended when I bought this board on NewEgg.

Anyway, everything's up and running, no problems there. But I decided to run memtest86+ (4.0) against it, just to check that memory was fine.

After running many passes against it, in many configurations, I get these odd results:

x Some passes get an error on a single bit on a single byte, always in the same place (tho' not all passes, and not just the "random" passes, so the same precise test will succeed sometimes).
x If I try the test on just one stick, whichever slot I put it in, it never gets the error
x If both sticks are installed, I get the error (but again, only occasionally)
x If I swap the sticks in the slots, I get the error at the same location (which wouldn't make sense if the memory itself was bad, would it?)
x When I run the test specifically on the region of memory the error happens on, after 20 passes, no errors.
x I don't remember what happened when I switched to use the 2nd channel... I believe I saw the exact same kinds of things happening, so it wouldn't be slot-dependent.

Folks in other forums have said "Must be incompatible memory triggering some kind of edge-case scenario." Possibly. Can't think of any other reason that I'm seeing this.

Have any of you ever seen this kind of thing happen before? I guess I could get some other memory and try... takes so damn long to test, though. May be temperature/running-time dependent, etc. so it seems to sometimes go away for awhile, then come back, depending on if the machine's been shut down in between.

Arrgh! Never had such problems with a build before!

- Tim

theycallmebruce
Posts: 292
Joined: Sat Jul 14, 2007 10:11 am
Location: Perth, Western Australia

Post by theycallmebruce » Sun Oct 04, 2009 5:22 pm

That's bad luck man, bad RAM is really annoying. As you say, you have to test for such a long time to get a conclusive result.

I've never seen the failure mode you describe before (only a single error, always at the same byte) so I don't know exactly what it might signify, but so far I've found memtest86 to be 100% correct with diagnosing bad RAM. Do you have any other RAM which you could test with?

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Sun Oct 04, 2009 6:13 pm

The underlying problem with your system is electrical. The signals going between the memory controller and the memory sticks is marginal on one of the data or address lines.

Folks who overclock are familiar with this issue, and deal with it by either reducing the clock rate or increasing the voltage, or both.

However, you've built your system with a motherboard that does not have those knobs.

Given that swapping the DIMMs around doesn't change the symptoms suggests to me that it is your memory controller or motherboard that has the problem.

All I can suggest is to make sure you're using the black DIMM connectors (the ones farthest from the the CPU physically), as they have better electrical characteristics, or to RMA your motherboard with a detailed explanation of the symptoms.

line
*Lifetime Patron*
Posts: 338
Joined: Wed Sep 28, 2005 10:07 pm
Location: Israel

Post by line » Sun Oct 04, 2009 7:10 pm

Can you pick up one of the modules and see what's the voltage rating on it? You can find it under the model name.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Sun Oct 04, 2009 7:42 pm

cmthomson wrote:The underlying problem with your system is electrical. The signals going between the memory controller and the memory sticks is marginal on one of the data or address lines.

Folks who overclock are familiar with this issue, and deal with it by either reducing the clock rate or increasing the voltage, or both.

However, you've built your system with a motherboard that does not have those knobs.

Given that swapping the DIMMs around doesn't change the symptoms suggests to me that it is your memory controller or motherboard that has the problem.

All I can suggest is to make sure you're using the black DIMM connectors (the ones farthest from the the CPU physically), as they have better electrical characteristics, or to RMA your motherboard with a detailed explanation of the symptoms.
Thanks for your great analysis! Sounds like you know your stuff!

I, too, have had an inkling it's something to do with the motherboard. This motherboard DOES actually allow me to modify voltages and timings, though not as precisely as, say, some Asus boards might. I tried using manual timings/voltages (to match that on the memory package), which didn't help; I also raised the voltage from that on the package (1.8v) to the next highest allowed increment (1.93 or something along those lines), but that didn't help either.

I've been using Channel A (Blue DIMM connectors); as mentioned, I can't recall what happened when I tried using Channel B (Black DIMM connectors). I'm about to try that now and see if I get the same results.

And yes, in the worst case, I'll see if I can RMA the motherboard. I threw away all the packing materials, unfortunately, which might make that difficult.

You might see me on the street soon wearing a hat over my eyes and saying "Heyyyy kid... wanna buy a muthaboard....? cheap?" :-)

- Tim

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Sun Oct 04, 2009 7:43 pm

line wrote:Can you pick up one of the modules and see what's the voltage rating on it? You can find it under the model name.
Yes, it's 1.8v. The Corsair web page on this memory says 1.9v sometimes, and 1.8v sometimes, but the package itself has 1.8v.

Someone was saying that they may have used the same memory for 2 different marketing strategies, depending upon how well any given stick spec'ed.

- Tim

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Mon Oct 05, 2009 7:24 am

Damn - ran it all night with the memory in Channel B (black sockets), and got the error twice over 8 hours.

Time to call NewEgg. Bleh.

Can these things be caused by the CPU itself, or is it most likely the motherboard?

- Tim

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Mon Oct 05, 2009 5:17 pm

Well, since I threw the packing materials away, NewEgg won't take the motherboard back. They suggested I call Intel. I called Intel. Intel told me that they'd need to see more than memtest86+ failing... eg. they'd need to see, specifically, a BSOD when starting Windows, or other similar kind of typical user experiential problems.

Jeez!

- Tim

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Mon Oct 05, 2009 6:22 pm

tbessie wrote: Can these things be caused by the CPU itself, or is it most likely the motherboard?
No, it wouldn't be caused by the CPU. This kind of failure occurs on the bus between the north bridge (MCH) and the DIMMs. The underlying cause is the north bridge chip, the board traces, or the DIMMs (or DIMM connectors). Most likely, in my opinion, is the MCH chip. Corsair DIMMs usually have lots of margin.

However, if you're willing to spend another $60 on an experiment (after all you can resell on eBay), try a pair of Crucial Balistix memory sticks. These have a huge margin and work perfectly in my system at much lower voltages than any other memory I've tried.

http://www.newegg.com/Product/Product.a ... 6820148216

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Mon Oct 05, 2009 7:00 pm

+1 for everything cmthomson was saying on the cause and diagnostics. I'm vendor/manufacturer neutral on the ram (I have preferences but they aren't strong enough for me to impose).

If you have a friend or coworker who has RAM to swap with you could try almost any ram that is DDR2.

Don't forget to try lowering the ram voltage or speeding up the RAM. It is opposite the recommendation prior in the thread but it is for the same reason.

Every stick of ram wants a different voltage. Often the average is so close to what is given that you don't notice the difference but you could have a board that gives out too much or too little or could have ram that wants less or more. Giving too little is not a risk, giving to much could be dangerous if you try too much.

line
*Lifetime Patron*
Posts: 338
Joined: Wed Sep 28, 2005 10:07 pm
Location: Israel

Post by line » Mon Oct 05, 2009 8:34 pm

I would also try another power supply, especially if the one in use is not a name brand.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Mon Oct 05, 2009 8:54 pm

line wrote:I would also try another power supply, especially if the one in use is not a name brand.
I'm using one of the top SPCR recommended PSU (well, for silence, anyway), the Nexus Value 430. Is that well known enough? My other builds all use Corsair modular PSUs.

- Tim

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Mon Oct 05, 2009 8:56 pm

cmthomson wrote:
tbessie wrote: Can these things be caused by the CPU itself, or is it most likely the motherboard?
No, it wouldn't be caused by the CPU. This kind of failure occurs on the bus between the north bridge (MCH) and the DIMMs. The underlying cause is the north bridge chip, the board traces, or the DIMMs (or DIMM connectors). Most likely, in my opinion, is the MCH chip. Corsair DIMMs usually have lots of margin.

However, if you're willing to spend another $60 on an experiment (after all you can resell on eBay), try a pair of Crucial Balistix memory sticks. These have a huge margin and work perfectly in my system at much lower voltages than any other memory I've tried.

http://www.newegg.com/Product/Product.a ... 6820148216
I think I'll give that a try. It's a lot easier than ripping out the motherboard and trying to find another one I like better (or hoping another of the same type wouldn't have the same problem with the same memory). Thanks!

By the way, what's the difference between those, say, and these ones that Crucial's site actually recommends for this motherboard?

http://www.crucial.com/store/mpartspecs ... F2A5CA7304

Besides the flashy lights, I mean? :-)

I'll also try some of those speed/volting changes... as best as I can, anyway. This motherboard doesn't allow a whole lot.

- Tim

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Mon Oct 05, 2009 11:18 pm

I don't see any Corsair memory on the "parts which passed testing" list on the Intel support site here:

http://www.intel.com/support/motherboar ... 029529.htm

I have not yet found a mainboard that did not work 100% reliable with Kingston ValueRAM.

This is the memory test I recommend:

http://oca.microsoft.com/en/windiag.asp

Microsoft calls it "Windows Memory Diagnostic" which a stupid name as is boots up directly from a floppy or CD using only the mainboard's BIOS. It has a very plain menu system.

Hint: press the T to run extended tests.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Mon Oct 05, 2009 11:47 pm

Wayne Redpath wrote:I don't see any Corsair memory on the "parts which passed testing" list on the Intel support site here:

http://www.intel.com/support/motherboar ... 029529.htm

I have not yet found a mainboard that did not work 100% reliable with Kingston ValueRAM.

This is the memory test I recommend:

http://oca.microsoft.com/en/windiag.asp

Microsoft calls it "Windows Memory Diagnostic" which a stupid name as is boots up directly from a floppy or CD using only the mainboard's BIOS. It has a very plain menu system.

Hint: press the T to run extended tests.
Thanks for all that - I'll take all that into consideration, and run that memory test. Now if that memory test fails, might be a decent stick with which to beat Into into RMA'ing this board. :-)

I'll also check into the Kingston memory. I found the list of tested RAM you quote before, and it seemed faily paltry, and contained little of the memory I ever use (generally Corsair and Mushkin).

This being said, I think I remember getting memory errors with memtest86+ on a previous system I built, but never had a problem with it over years, which has always made me doubt memtest86+'s accurracy for some motherboards.

Anway, will report back.

- Tim

PartEleven
Friend of SPCR
Posts: 279
Joined: Sun May 06, 2007 10:37 am

Post by PartEleven » Tue Oct 06, 2009 7:32 am

Hi, not to thread hijack, but I've got a similar problem. The details can be found in this post. At this point I think that particular set of ram is faulty, but is there a chance that it's the motherboard? Zotac motherboards aren't known for their wide compatibility, and the problem ram passes memtest just fine on another system.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Tue Oct 06, 2009 10:10 am

More joy (ie. lack of)... I've confirmed that the memory effect I'm seeing is indeed real (if there was any doubt). I had rsync'd my whole multimedia collection from one set of disks to another on this machine.

Just as a test, before I wiped out the old storage), I ran a diff -r -q oldHierarhy newHierarchy.

I found 4 files that were different (4 files in some ripped DVDs I'd copied).

I used 'dd' to get the relevant regions of bytes, and compared them, and found that the exact same bit of the same nibble (most significant 4 bits) of a byte were flipped.

Note, this was only 4 files in thousands of huge ones, so this problem doesn't cause MUCH strife. But it scared me much.

I think I'll order the recommended memory from Crucial and try that for yuks, and if THAT fails, swap out the board.

So... does anybody know why Ballistix memory would be better than the standard recommended Crucial memory for this board?

- Tim

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Tue Oct 06, 2009 11:25 am

I don't know if Ballistix memory is better than Crucial memory, but the recommended/tested memory is always your best bet. Next best is Kingston ValueRAM.

Intel says not to use double-sided DIMMs with x16 organization which I think means 16 IC's with 8 IC's each side of the PCB organized logically as a single rank. That would mean that all sixteen IC's are energized at the same time and would present a heavy load on some of the lines on the memory bus.

So then you have options:

> Perhaps the Corsair will work in another system.
> Use only 1 of those 2 Gbyte modules
> Slow your memory down with manual settings and/or twiddle voltages

Personally data integrity is my priority...

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Tue Oct 06, 2009 11:46 am

The reason I suggested Ballistix is simple: they are bin selected for high margin to support outrageous overclocks, and they are cheap. (They are also really gaudy; if that bothers you, put some tape over the LEDs :) )

Many other memories are also binned, so don't take this as anything other than a happy customer saying he's happy with a product.


On a completely different tack, if your BIOS will let you bump the north bridge/MCH voltage by a few percent, you should give that a try. This is a well-tested method used by overclockers to make otherwise marginal systems stable. And, just in case it might be a heat issue, make sure your MCH isn't too hot; you might need to upgrade the TIM or airflow on its heat sink...


BTW, memtest is not the most strenuous program for memory stability. I use it as a quick gateway, before running the real bear: ORTHOS. If your system can run ORTHOS for 24 hours in blend mode, then it is rock solid.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Tue Oct 06, 2009 11:47 am

Wayne Redpath wrote:I don't know if Ballistix memory is better than Crucial memory, but the recommended/tested memory is always your best bet. Next best is Kingston ValueRAM.

Intel says not to use double-sided DIMMs with x16 organization which I think means 16 IC's with 8 IC's each side of the PCB organized logically as a single rank. That would mean that all sixteen IC's are energized at the same time and would present a heavy load on some of the lines on the memory bus.

So then you have options:

> Perhaps the Corsair will work in another system.
> Use only 1 of those 2 Gbyte modules
> Slow your memory down with manual settings and/or twiddle voltages

Personally data integrity is my priority...
Does the comment about x16 organization mean I'd have to put in 4 1gb sticks to meet that requirement? Or are there 2gb sticks that would meet the requirement? I wonder how I can find out this level of detail about the memory I buy... don't remember seeing that anywhere.

Officially, the board specs say:
he Intel® Desktop Board DQ45CB supports the dual or single channel memory configurations defined below.

* Four 240-pin Double Data Rate 2 (DDR2) SDRAM Dual Inline Memory Module (DIMM) connectors with gold-plated contacts
* Support for:
o Non-ECC, 1.8 V, DDR2 800/667 MHz memory
o Serial Presence Detect (SPD) memory only
o Unbuffered, non-registered single- or double-sided DIMMs (except double-sided DIMMs with x16 organization)
* Memory configurations listed below:
o Up to 2.0 GB utilizing 256 Mb technology
o Up to 4.0 GB utilizing 512 Mb or 1 Gb technology
o Up to 8.0 GB utilizing 1 Gb technology
* A minimum of 512 MB of total memory is required
I am not sure what they mean by, for example, "1 Gb technology". Do you?

As for the options.... 1. I hope, 2. I want 4gb memory, as it seems to help in large data transfers, and 3. I may play with that, but I'll try new memory first.

Thanks!

- Tim

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Tue Oct 06, 2009 11:54 am

tbessie wrote:By the way, what's the difference between those, say, and these ones that Crucial's site actually recommends for this motherboard?

http://www.crucial.com/store/mpartspecs ... F2A5CA7304

Besides the flashy lights, I mean? :-)
CL6 vs CL4. A measurable but perhaps not noticeable performance difference. Which reminds me: you might try bumping up the CL value if your BIOS allows that. I'm guessing you're currently using the SPD values, so you might want to switch to manual values if you can.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Tue Oct 06, 2009 12:09 pm

cmthomson wrote:
tbessie wrote:By the way, what's the difference between those, say, and these ones that Crucial's site actually recommends for this motherboard?

http://www.crucial.com/store/mpartspecs ... F2A5CA7304

Besides the flashy lights, I mean? :-)
CL6 vs CL4. A measurable but perhaps not noticeable performance difference. Which reminds me: you might try bumping up the CL value if your BIOS allows that. I'm guessing you're currently using the SPD values, so you might want to switch to manual values if you can.
It allows some tweaking. Right now I'm running at the auto values, 5-5-5-18 (if you're speaking of those values). What would you modify them to in tests? And would you change voltage and "reference frequency" at the same time?

- Tim

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Tue Oct 06, 2009 12:35 pm

tbessie wrote:It allows some tweaking. Right now I'm running at the auto values, 5-5-5-18 (if you're speaking of those values). What would you modify them to in tests? And would you change voltage and "reference frequency" at the same time?
Well, just for a lark, you might try 6-6-6-20, but I doubt if it will help.

Based on what you've posted above (same bit fails even when DIMMs swapped or put in other slots), one of the data lines between the MCH and the memory is marginal. This could be a physical problem with the board (eg, slight nick in the trace or improper soldering), or more likely a marginal driver inside the MCH for that line.

So my best advice at the moment is to try a slightly higher MCH voltage if you can.

I don't know what "reference frequency" means for your BIOS. If it is the CPU:memory clock multiplier, then it might make sense to try changing it. The best performance is with a 1:1 (usually reported in the BIOS as 1:2 since the memory is double-pumped (but then again the CPU bus is quad-pumped... :D )), but many enthusiasts select different ratios because of some unanticipated sweet spot in their motherboard/memory combination.

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Tue Oct 06, 2009 12:38 pm

dhanson865 wrote:Don't forget to try lowering the ram voltage or speeding up the RAM. It is opposite the recommendation prior in the thread but it is for the same reason.
Hmm, good suggestion, even if counter intuitive. Since the problem here is signal integrity on a data line, any change to voltage or frequency could make it better or worse, including lower voltage or faster clock.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Tue Oct 06, 2009 2:06 pm

By the way, I wanted to thank everyone for all their help and suggestions!

This is by far the friendliest and most helpful, knowledgeable forum (all at once) of any of the computer-building forums I've participated in.

Who wouldda known? Isn't it just about silent computing? :-)

Maybe people who appreciate silence are kinder? :-)

- Tim

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Tue Oct 06, 2009 5:59 pm

The problem here is signal integrity on a memory buss line. It is probably not a data line. I know that you are seeing a data error, but if the problem is an excessive buss loading issue then it is probably an address line or secondly one of the control lines.

You said "I try the test on just one stick, whichever slot I put it in, it never gets the error" and "If both sticks are installed, I get the error". You also indicated that you have tried using both the blue sockets and then both the black sockets. Both of these options will give dual channel, also called interleaved, mode which is faster than single channel mode. Dual channel mode also turns on both modules at the same time which probably doubles the loading on some of the bus lines. Single channel mode will turn on only one memory module at a time.

The next 'easy on the wallet' experiment would be to try single channel mode.

As the Intel Desktop Board DQ45CB manual says "Regardless of the memory configuration used (dual channel, single channel, or flex mode), DIMM 0 of Channel A must always be populated" my suggestion is to use the 2 sockets closest to the CPU.

Have fun.

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Tue Oct 06, 2009 6:36 pm

Wayne Redpath wrote:The problem here is signal integrity on a memory buss line. It is probably not a data line. I know that you are seeing a data error, but if the problem is an excessive buss loading issue then it is probably an address line or secondly one of the control lines.
Okay, maybe. Whether it's a data or control/address line, the task is to find a clock/voltage that works reliably. Unless of course a hammer is involved...

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Tue Oct 06, 2009 7:03 pm

Does "The memory was recommended when I bought this board on NewEgg" mean that the NewEgg sales rep recommended this parts combination? The correct recommendation would have not have been hearsay, but reliable technical research. If so then I would hope that they would correct their error.

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Tue Oct 06, 2009 9:42 pm

Wayne Redpath wrote:Does "The memory was recommended when I bought this board on NewEgg" mean that the NewEgg sales rep recommended this parts combination? The correct recommendation would have not have been hearsay, but reliable technical research. If so then I would hope that they would correct their error.
Well no, I was just being lazy (egged on by the fact that I'd never had a serious memory issue in previous builds).

When I added the board to my "shopping cart", NewEgg said "People who bought this board also bought..." and named that memory.

Of course, it might well've been something they were trying to get rid of. I was lazy, as mentioned, so I just got that memory.

In the past, I did a lot of research, and always ended up buying Mushkin memory from the Mushkin website, after discussing the motherboard I was using with one of their forum moderators. Maybe I just like the name "Mushkin". :-)

Anyway, this is what comes of lack of research, most likely.

- Tim

tbessie
Posts: 232
Joined: Tue Feb 05, 2008 10:07 pm
Location: San Francisco, CA

Post by tbessie » Tue Oct 06, 2009 10:08 pm

Wayne Redpath wrote:The problem here is signal integrity on a memory buss line. It is probably not a data line. I know that you are seeing a data error, but if the problem is an excessive buss loading issue then it is probably an address line or secondly one of the control lines.

You said "I try the test on just one stick, whichever slot I put it in, it never gets the error" and "If both sticks are installed, I get the error". You also indicated that you have tried using both the blue sockets and then both the black sockets. Both of these options will give dual channel, also called interleaved, mode which is faster than single channel mode. Dual channel mode also turns on both modules at the same time which probably doubles the loading on some of the bus lines. Single channel mode will turn on only one memory module at a time.

The next 'easy on the wallet' experiment would be to try single channel mode.

As the Intel Desktop Board DQ45CB manual says "Regardless of the memory configuration used (dual channel, single channel, or flex mode), DIMM 0 of Channel A must always be populated" my suggestion is to use the 2 sockets closest to the CPU.

Have fun.
I'm currently running memtest86+ with the memory in 1-blue/1-black (nearest the cpu) configuration. Seems like an abomination of nature, since you don't get the benefit of dual-channel memory from that, but worth a test anyway. :-)

- Tim

Post Reply