Jump to content
Corsair Community

Puzzling Memtest 86+ Test#7 Errors with 4x1GB Dual-Channel X38 MB


jase439

Recommended Posts

Hi all. I could use some collective wisdom from the forum regarding a puzzling phenomenon I have been unable to resolve (or understand).

 

I have 4 sticks of 1600C7DHX in my Asus Maximus Extreme (X38) running an X3350 (Q9450). Memtest86+ has started kicking out errors in Test #7 affecting one bit at 2 specific memory addresses. It is the exact same bit at each memory addresses that fails every time (~281MB and ~2311MB). I presume this is more than mere coincidence and that these bits are mapped to the same approximate lower address space of each respective bank.

 

In the past, with the problem address being consistent from run to run, I'd lay fault on the RAM, but I've swapped these 4 sticks 10 ways to Sunday to no avail having tried them in different banks and slots. I am left to conclude that my problem is in someway related to the memory controller, but I have no idea where to go next with this. Test #7 fails consistently: same bit, same addresses.

 

I've spent the past several days futzing with my vDIMM (currently 1.9), loosening timings all the way up to 9-9-9-24, jacked the NB up to 1.55v, raised the VTT to 1.3-1.4v, disabled all AI twisting and transaction boost, futzed with clock skews, dialed back FSB, tried different FSB/NB straps, jacked up vCore as high as 1.3 (even though CPU is completely stable at 3.36GHz and needs only 1.22v), changed tRFC from 72 to 80 to 88. Temperatures (according to sensors at least) are well within nominal. Checked airflow in the case. Can't get past the darn test #7.

 

I can run OCCP without issue, but the machine is not Prime 95 Blend stable (precision roundoff...0.5 expected 0.4 yada yada after about an hour). The machine is also unstable after coming out of hibernate or sleep and will often lock up shortly after a resume with any bus-intensive activity (OS: Win7 x64). The machine is probably stable enough for day to day operation, but with my drive being RAID-0, I'm not entirely comfortable opening the integrity of my drivers' memory space up to random bit-rot.

 

My system BIOS is 1104. There is an update, but I read somewhere that the newer BIOS have introduced stability issues.

 

Any ideas on where to cast my line next (short of sanding my grout with the motherboard?) Maybe it's time to cut bait and put a new board in this (but what a pain)...but I feel like I've tweaked every tweakable knob and I'm just not getting any love. Maybe I'll find some here :)

 

Thanks for any help or insights.

 

J

 

EDIT: I may end up eating my words WRT to the RAM. I pulled all the sticks and tested each one individually in single-channel mode in each of the MB slots. Peachy; no problems with test 7. Next, I tried one pair of DIMM's in dual-channel (slots 0 and 2) and was immediately pelted with errors (at almost 3x the frequency as when I was running 4x1GB at the same timings). I pulled those sticks and replaced them with the other pair of DIMM's (same timings). Expecting the same badness, I was surprised to find myself error free. So far, I'm running clean with this pair of DIMM's. I'm going to leave this running for a few hours to see how the system fares, but now I'm entertaining the idea that maybe one of these sticks from the first batch just isn't up to par and is dragging the whole lot of 'em down in dual-channel mode.

Link to comment
Share on other sites

Why would two sticks of RAM at DDR-1600 report a bandwidth of 6015MB/s in memtest, but two different but identical sticks (make/model/version) of RAM at the same timings and same slots report only 5700MB/s?
Link to comment
Share on other sites

Try Memtest with Legacy USB Support disabled. Memtest has a conflict with this setting on ASUS MOBOs. Also, when you have the instabilities, have been able to tell if the NB (memory controller) is getting hot? With 4 modules, it's going to get VERY hot and may need extra cooling.
Link to comment
Share on other sites

NB hangs out around 42-43 at idle, and tops out around 48-49 under load. I think I've seen it as high as 51 once. How hot is too hot in your experience?

 

Will try disabling USB legacy and see where that gets me (although I read somewhere that this was fixed in the most recent version of memtest86+...4.10). I'm a little skeptical that USB legacy is to blame since the system is also not prime stable (does USB legacy support affect only memtest or does it bugger all applications?).

Link to comment
Share on other sites

They are. I ordered 2 pairs of CM3X1024-1600C7DHX. I've had them for a couple years though. Here's the specs from the back of the modules (they are all identical):

 

XMS3-1600 1024MB 1600Mhz 7-7-7-20 1.80V ver3.2 08111187

 

Right now, I'm falling back and just trying to get 2 sticks working in dual-channel mode at 400 FSB. I have one pair of sticks running stable in memtest and OCCP with 7-7-7-20 timings, vDIMM 1.80V, NB pared back to 1.45V (to help mitigate any thermal badness), FSTB/VTT is 1.3V, and every thing else set to auto except for tRFC which I needed to increase from 72 to 82 to get dual-channel to memtest cleanly. I'm satisfied with the thermal readings (CPU core temps low to mid 60's under load and NB at 45).

 

Can't get the other set to give me any love at the moment. And I'm perplexed as to why memtest is giving me different bandwidth values at the exact same timings. ???

Link to comment
Share on other sites

At the moment, it isn't even handling 2 (I've backed away from 4 sticks at the moment to give the memory controller as much benefit of the doubt). But I think I have a bead on one particular stick of RAM that is consistently an underachiever in the "plays nicely with others" category. I'm in the process of convincing myself the other 3 are fine.

 

EDIT: so far I've managed to run different pairings with 2 sticks of RAM with the timings in my previous post, and 3 of the 4 DIMMs consistently demonstrate stability. The 4th stick, when paired with any one of the other 3 will readily generate errors in Test 7 and (at times) Test 5. While this stick seems OK by itself in single-channel operation, it doesn't want to play with others - even under the most relaxed of timings. Whether the stick is outright baked, I can't say, but it certainly is not performing to spec (even at DDR-1333).

 

EDIT: just as an additional data point, I have successfully demonstrated stability (memtest, OCCT, and prime) with 3 of the 4 sticks running as 2 stick pairs with one another. I am also prime stable with 3 sticks running in asymmetric dual-channel. Introduction of the 4th stick of RAM in any dual-channel configuration produces immediate errors. I am convinced that this stick is defective or not performing to spec.

 

Not sure how I should proceed. Can I RMA just 1 stick? I am concerned that any replacement might be a different revision than the other 3 leaving me in a different, but equally bad boat. Should I exchange the pair or all four or throw the lot out and just purchase a set of 2x2GB. Pbbbl. Recommendations?

Link to comment
Share on other sites

Thank you for getting back to me.

 

Should I RMA all four sticks? Three of them appear to be just fine. Can I RMA just the defective stick and get a replacement that exactly matches the other 3? If it's better to send back all 4, this will take my primary development workstation out of commission for a while. Can Corsair cross-ship a replacement if the RMA is secured with a credit card?

 

Thanks again.

 

J

Link to comment
Share on other sites

I was advised today that the DHX series is discontinued and that the comparable replacement part is CMP4GX3M2C1600C7. However, I was reading that this DDR3 module was designed for the Intel Core i3/5/7 series processor. I don't know if that is just a marketing testament to the performance of this RAM or if this memory requires a specific CPU/chipset. My dual-channel DDR3 mobo has an X38 chipset and is running a Yorkdale processor, namely the Xeon X3350 (Q9450). Will I be OK with this new RAM?

 

J

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...