Sign in with
Sign up | Sign in
Your question

Dual Chan fails mem test, passes 1 stick at a time

Tags:
Last response: in Memory
Share
July 19, 2003 7:12:07 PM

I am quite new to this and I have read quite a lot of threads on here and other forums without luck (I saw a similar thread on here but the problems seemed different).

I have just bought a new system including the following components
Antec Truepower 430
MSI K7N2 Delta-ILSR
XP2800 Barton fsb333
2 x elixir 512mb DDR pc3200 RAM
Seagate Barracuda V 120GB SATA hdd

I put all components above in the system including a GeForce 3 grfx card, floppy, cd rom etc

Bios settings mostly default or auto except fsb changed to 166. RAM was recognised at 400

I tried to install win2k straight onto the brand new SATA hdd. Every time it would blue screen at some point while copying the files over.

After many hours I removed 1 stick of ram to see if 1 stick was faulty and managed to get an installation up and running. As soon as I put the other stick back in, the system would blue screen at random points during OS startup with various errors.

I got a bootable floppy running docmemory and tested all combinations of 1 stick at a time in each slot, then both together in different slots. 1 stick at a time the test said RAM was fine.
Together with one stick in DDR1 and one in DDR2(single channel on my Mboard) system would simply stop while going into "booting from floppy" stage.
Together with one stick in DDR1 and one in DDR3 system (dual channel) and then one in DDR2 and one in DDR3 (dual channel) RAM fails memory test.
Errors came up on both when at the burst test. I repeated these a few times each to make sure (each set of tests takes 20mins :(  ). I cant remember if the error on the first dual channel combination was exactly the same as the second because I didn't record the exact values on the first one. The second combination gave the following RAM test error.

Burst.......FAIL
Addr:( 111M), Exp.: 0F0F0F0F, Act.: 0F0FF00F

As I understand it, underclocking my RAM is a good idea to sync with the cpu, but should I be getting these errors anyway without them in sync? If anyone can give me some values I can use to underclock my RAM then I would also try that, but I have never really clocked anything before so was a bit wary at this stage.

Any help would be much appreciated.
July 19, 2003 7:47:44 PM

Do the sticks fail the test when you test them one at a time but using a different DIMM? Try that.
Also, it seems that your system has been faulty from the moment you got it and it has noting to do with the components you added, if you bought all the items from one suplier you can go to them and tell them to find what's the friggin problem with the system they sold you, since it's obviosly a hardware problem and nothing you screwed up.

Coppermine, Banias, Opteron, Prescott, Barton... How 'bout investing less in the CPU and more in its name?
July 19, 2003 9:01:26 PM

Well if you got your CPU running 333fsb then the ram will run only in pc2700. Overclocking the FSB to 400 is a cool idea to sync them up and get the most from your ram but you gotta crank the voltage and cpu cooler. When the 2 are out of sync, you could deal with stability issues pending the MOBO. Another thing if you are running in dual channel mode. The ram MUST be pretty much perfectly identical to each other or it will not work. Either way this the job for the seller to figure out so give them the WTF.
Related resources
July 27, 2003 9:47:28 PM

The RAM was the same brand bought on same day so would imagine it to be as near as identical as you can get apart from paired sticks. I may overclock the fsb and cpu to 400 at a later date when I am confident in my cooling but for now have just changed the ram timings in the bios and everything is fine.

Thought I'd let people know how I got on just in case they have the same problem and read this thread.

Thanks for the help.
July 30, 2003 1:52:55 AM

Sometimes "near identical" isnt quite good enough, especially with high Mhz you are dealing with.
(Why do u think corsair released special 'twinex' modules?)

To test for memory errors i suggest you use MEMTEST86... far more reliable than waiting for windows to crash.
See if relaxing the memory timings helps at all.


<b>Regards,
Mr no integrity coward.</b>
!