My Memories doesn't pass MemTest for Windows

[crustaceo]

Hi,

 

I received a Crucial Ballistix PC3200 2 x 512 Kit in this Monday.

Ok, then I installed it in the proper way (CMOS clear, one stick of RAM, Load Optimized Defaults, two sticks, etc, etc, etc). So I ran it in the BIOS Auto values (timmings 2-3-2-8 - 200Mhz - 2.6V (BIOS read 2.7)). Everything went Ok and Windows loaded without problems. But after 5 minutes I thought that things could be funnier. I read a review in the anandtech's website that these memories could run 2.5-2-2-5 at 250Mhz. I set it using 2.7V(BIOS read 2.8V) using these settings. It booted without problems and windows loaded Ok, and then I played BF2 for about 10 minutes - because my FSB was 250, and my CPU (10x) was overclocked to 2.5Ghz in STOCK COOLER, so i didn't play much with this setting. Then, I decided to run memtest86+ in this configuration, and I had got some errors after just few minutes. After, I set the memories back to default in order to run memtest86 to check if it was really Ok at default configs. And then things went very bad - no boot sometimes, very very instable system, tons of errors in memtest86. I thought I had damaged my memories with this overclock experience and I even contacted my reseller to RMA them. So, after some thoughts, I decided to run these memories in 2.8V (BIOS read 2.9V), and, wow, everything was OK again (thanks God).

After that, I cleared the CMOS again, and set my RAM at 2-2-2-8 (SPD values), 200Mhz, 2.8V(2.9V) and started to run torture tests. The tests I ran is the following:

 

1. 12hrs Memtest86+ default test and lots of #5 and #8 tests - OK

2. 6hrs dual prime 95 - OK

3. 32M dual Super PI - OK

3. 6hrs ORTHOS - OK

 

But, when I run Memtest for Windows (two instances) I get some errors in one instance at ~300-~500% coverage, and this instance pop ups some errors frequently after that. The other instance seems to run OK. So my question is: after ran a lot of other test programs, is it really worth worry about these errors in MemTest for Windows?

 

BTW, these memories really goes very hot, I can't even bear thouching it for more than 3 seconds when i run MemTest for Windows, for example. I have a good air flow in my case (2x12mm fan + 1 x 92mm fan), but I think I have to adapt a fan to blow some air to my memories. I can't run these memories with 2.8V or less. Is it a problem too?

 

Thanks guys, and sorry about my bad engRish, i'm studying it frequently now.

[Praz]

Regardless what stability testing has shown so far, if you are receiving errors in Memtest For Windows, there is definitely a problem.

 

The heat may be the issue. Memory should never get that hot. Point a fan at the modules and see if that help with the errors. It's also possible that with that much heat the memory has already been damaged.

[crustaceo]

i've already bought a 80mm DC fan to try it this night. :]

i hope it's not damaged... i don't think so, cause these memories are recommended to run at 2.8V, so 2.9V wouldn't damage it so much. everyone says these mem heats more than other memories around there.

 

thanks

[krose]

With a gig total of RAM, did you assign 384 or 400MB of memory to each instance of Windows Memtest? Also, did you assign each instance to a separate core using affinity in task manager?

[crustaceo]

I run MemTest correctly (400MB for each instance). MemTest actually balances between cores and doesn't need affinity set like super-pi or prime. :]

Just to update the thread and let you guys know about this issue, a 80mm fan blowing just over the memories did the job. It now passes MemTest with 6000% of coverage with 350MB for each instance (I let it running all last night long) and i can put my finger over the heatspreader until I become tired. :]

I hang the cooler using a ridig wire and screwing it in a hole that my Soprano has just over the MB.

 

and just to know, is my English bad, or very and horrible bad? :confused:

cya

[krose]

Glad you got it sorted out, and your English is fine!