Memtest86+ hangs

Toggle sidebar Toggle sidebar
J

joex444

Distinguished
Feb 16, 2006
1,279
1
19,960
I'm not sure what to make of this:

Using Gigabyte 965P-DQ6 Rev 3.3 (released in January)
E4300 @ STOCK 1.8GHz
2GB G. Skill DDR2-800 5-5-5-15

When I run Memtest+ 1.70 it will hang within 2 minutes, reporting no errors. It is always during the 2nd test, inversions.

I've tried this at 5-5-5-15 using the following: 800MHz @ 1.8, 1.9, 2.0V (the specs on the RAM are 1.8-2.0V). I also tried 667MHz @ 1.8V, and it wouldn't boot at 533MHz.

Also worth mentioning is that Memtest displays no information for my chipset or ram timings, don't know if this is important or not, but I've always seen screenshots of it with that being displayed.

Is there something else that I can use to test the RAM, or is this enough proof to go ahead and RMA the sticks?
 
Why not go ahead and see if it will install the OS and test for stability with the OS? Memtest is a great tool but, with what you have posted, I would not base an RMA solely on it. Since those CPUs are pretty new, make sure you have the latest bios also.
 
Oh, sorry, I've already installed XP SP2 on it. Also the board came shipped to me with BIOS F9, the latest one from Gigabyte. The board itself is also quite new, Rev 3.3 supports FSB1333 natively, first 965 to do so, released in January/Februrary of 2007.

When I was encoding some MPEG2 videos it would hang at overclocked speeds which is why I wanted to test the RAM (this happened on another machine and it was a CAS latency problem).

The troubling thing is that if I disable USB legacy support and remove 1 DIMM, it will run Memtest86+ all the way through and pass without errors.

It also seems to be a bit more stable overclocked, as I've actually been in Windows at 2.4GHz for about 20 minutes now and it hasn't hung. It would do that randomly before, at anything higher than 2.0GHz, anything that stresses the CPU/RAM - games, encoding, TAT.

So, now, with this additional information, I intend to see how stable it is at 2.4GHz with 1 DIMM, swap out the other DIMM run Memtest86+ and see how stable it is then.
 
...
The troubling thing is that if I disable USB legacy support and remove 1 DIMM, it will run Memtest86+ all the way through and pass without errors.
Click to expand...
I think the legacy USB causing memtest problems is a known issue with some MBs -- there should be more info on the memtest support forum.

Since you have Win running, go ahead and download and run Orthos.
 
...
The troubling thing is that if I disable USB legacy support and remove 1 DIMM, it will run Memtest86+ all the way through and pass without errors.
Click to expand...
I think the legacy USB causing memtest problems is a known issue with some MBs -- there should be more info on the memtest support forum.

Since you have Win running, go ahead and download and run Orthos.The only conflict I know of for sure is with the ASUS boards and it relates to their bios shadowing. However, it cannot hurt to disable for Memtest.
 
In his initial post, he said the RAM was rated at 1.8-2.0 V, so it wouldn't hurt to try 2.1V.

joe - have you changed any RAM settings other than speed and voltage? It wouldn't hurt to make sure that the latencies, etc. are all set to their "auto" settings.
 
The latencies are set to auto, and even tho the dimms SPD reads out 5-5-5-15, they're actually being run at 5-6-6-17. The clock is 398MHz, so it's within specs.

1 DIMM seems to be stable, I'll try the other DIMM then see if I can get both to work.

Also might try that +0.30V setting for 2.1V if I have to.
 
I have a similar issue with a DQ6 v2 and two dimms. Basically, I can't run at 800 MHz in dual channel. One dimm is fine, and dual channel is fine at 667 MHz. Let me know if you have any luck finding a solution.
 
What RAM are you using? Edit: Saw your post, Mushkin or Supertalent.

I've found out this much:

Both DIMMs work seperately. One defaults to 5-6-6-17 and the other runs at 5-5-5-15. I find this odd.

I've tried Dual Channel operation using both Channel 1 and Channel 2, at 5-6-6-17. Neither configuration allowed Memtest to get past 2 min 17 seconds (somewhere in Test #4).
 
Hmm.... Memtest still won't work at 5-6-6-17. I'll try running some test tools in Windows, see what happens.

If that fails, zipzoomfly.com has Corsair 2GB DDR2-800 for $190 with a $40 rebate until the 22nd. It'd end up being about what it would cost to RMA these sticks.

All the reviews I've seen of this board have been run with Corsair 2GB DDR2, and it's one of the approved manufacturers on Gigabyte's site, here. I notice that G.Skill, Supertalent and Mushkin don't appear on the list...
 
Test 4 is where mine goes down too, almost always at 31%.

I really thought the Mushkin woulda been fine as its 1.8v ram ...
 
This is what I get for being cheap.

Even worse, when I put both sticks in different channels they run as dual channel, even though they are not supposed to... I can't even run 2GB single channel 🙁

IIRC, Dual channel at 667 also didnt work, and it wouldnt boot at 533, or 6-6-6-17 (or was it the 2.1V? I changed both at the same time, idiot).
 
I get a dual channel reading when I put the sticks in different channels also. Although I can run dual prime95 on 667 in dual channel, it still hangs on test4 in memtest86+. Hell I can run BF2142 for 5 hours straight on a 3.0 ghz overclock on these settings.

Same is true at 533 also, so it is any 2*1GB situation that hangs memtest86+. Whatever the problem is, it gets worse at higher memory clocks ... independent of the cpu overclock.
 
I've verified your claim that any 2*1GB situation fails.

Even using slots 3&4 to create Single Channel mode, the RAM will fail Memtest86+ within 2 minutes, at 5-6-6-17 and 1.95V 800MHz.

I've RMA'd those sticks and purchased a pair of OCZ 1066MHz 5-5-5-15 2.1V sticks which actually appear on the Gigabyte DQ6 Rev 3.3's supported memory list!

Single channel mode seemed to have a good shot at being stable, though. I was able to play a game for about 20 minutes. In dual channel mode it usually crashed within 30 seconds. This is at stock 1.8GHz, BTW.
 
Yep, the memory controller on the board is as flakey as mine. Are you air cooling? Someone suggested my problem might be related to cooling the mosfets. Nothing I've tried in that area has helped however.
 
Indeed, air cooling. Tuniq tower 120, got 2 x 120mm fans blowing out near the chipset heatpipes and the PSU is seperated (Antec P180 case), so the only heat is from the CPU & the 8800GTS.
 
SOLUTION

In order to use the Tuniq Tower 120, I had to remove the Crazy Cool heatsink on the underside of the board.

Gigabyte changed the mounting system from Rev 2.0 to 3.3!!

In Rev 2.0 boards, they use a plastic clasp to lock into the Crazy Cool. If you unscrew the Crazy Cool then you have to unclasp the tabs, but reclasp them to attach the Nortbridge heatsink. In Rev 3.3, they went with all screws, so if you unscrew the Crazy Cool your Nortbridge isn't attached to the heatsink and you need to find a way around this.

First, I'll try using the stock cooler and the Crazy Cool in place. If that works then I know the board is good and the RAM works. Then, I can try to replace the stock cooler with the Tuniq 120, and use a nut with an electrical insulator to get the Nortbridge screws to put pressure on the Northbridge.
 
So, here's what happened.

1) The replacement DQ6 was shipped to me with 2 corrupt BIOS. When I powered it on it would reboot then display the Award BootLocker v1.0 screen, which would do nothing because I didn't have a hidden HD partition with a backup BIOS on it. Gigabyte claims Quad BIOS, the 4th being on the CD. I never did try that, but somehow I don't think it would have ever gotten to the CD. I did try a floppy, but alas it wouldn't work either.

2) In a moment of frustrated inspiration, I took the motherboard out of the case and just for due diligence proceeded to setup a very basic setup on my desk. Just the mobo, CPU, hsf, RAM, gfx and PSU. With a PCI card it wouldn't ever display anything, so I had to use the PCIe card I've been using all along (this did work in PEG1 and PEG2). Again, the BootLocker displayed rather than a POST.

3) Just to be sure I truly did test everything, I swapped the 2x1GB OCZ 1066MHz sticks (on Gigabyte's list of supported RAM) for a 1x1GB G.Skill 800MHz stick.

4) Totally amazed, a different screen appears stating that the BIOS checksum has an error. It asks me how to proceed, so I hit Del to enter Setup. Then, I leave the BIOS and reboot.

5) Equally impressed, the damn thing POSTS! I shut it down, and install a DVD drive to run Memtest86+

6) Memtest86+ runs for about 15 minutes, then I shut it off. It ran without errors, like before - single channel.

7) The true test: I install the 2nd OCZ 800MHz 1GB DIMM and let it boot into Memtest86+ in dual channel. I'm sitting here right now watching it go. It's at 22 minutes -- the other board would crash within 2 minutes. It's at 76% total and seems like it will finish (if it doesn't, this will be edited otherwise anyone reading this can assume it finished).

I guess the next test is to try the OCZ RAM in single channel, then dual channel. If that works, then take the CrazyCool off, put some insulated nuts under the Nortbridge and see if it can pass Memtest86+ again. Then, if it does I'll put the Tuniq Tower on, since that is a bitch to remove. If the OCZ doesn't work, then I guess I'll have to RMA that. Luckily, I still have about 10 days left on the OCZ; the G.Skill only has manufacturer RMA at this point.

I'd be interested to know what caused the initial BIOS errors, though. Could it be caused by a bad contact on a DIMM? Like, if you had a DIMM that wasn't electrically connected (ie, 238/240 pins are electrically connected)?
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom