Memtest stable, but unstable with anything else

[mc123]

OK, now I've decided to RMA the RAM. Although cleaning the edge connectors was an excellent idea, it didn't help; but now both GoldMemory and memtest86 often fail on Stick B. Not quite consistently, but enough to have allowed me to collect enough evidence to show that Stick B is defective.

 

Stick A is running without problems at 2-2-2-6 1T: I'm actually writing this in Knoppix, with two instances of mprime (the Linux equivalent of Prime95) running at the same time. I haven't been able to reach this level of stability with Stick B in the system. It rather bugs me that Stick B doesn't fail consistently, but I can't think of anything else to do to support or refute the contention that Stick B actually is bad.

 

If the system stays mprime stable for 24 hours with Stick A alone, I guess I'll reinstall Windows and all the apps later in the week, Ghost the partition to a DVD (again), just in case the partition i First Ghosted had been trashed, and RMA both sticks of RAM next week (if I RMA the RAM tomorrow, I most likely won't get the replacement RAM before early next week at the earliest, anyway). This will be the second time I've RMA'ed Ballistix because of a defective module - nevertheless, the Ballistix modules that have worked have been excellent, and I'm willing to give them another chance. But one more strike, and they'll be out for good and instead I'll probably go for Mushkin Redline or some flavour of OCZ.

 

I'll let you know how things pan out with the replacement RAM when I've tested them. And many thanks to Sharp, ExRoadie and -CsA-TAZ for their advice and encouragement, it is very much appreciated. So I guess it's so long for now...

[red930]

Some times the memory sticks may need an abrasive cleaning. I use the eraser from a number 2 pencil. The pink kind!

 

I hold the stick securely on a nice flat surface then use the eraser to lightly polish the contacts. Once both sides are done I clean the edge again with 90% or greater Isopropyl alcohol.

 

Normally this is required on older sticks but has worked on goofy RAM right out of the box.

[mc123]

That's the first time somebody has advised me to use an eraser to clear the RAM

 

I'll have to dig around to find a suitable eraser - these days, erasers are usually of the white, rubbery type, and 2HBs are near-extinct - but I'll certainly give it a try, without making too much gold dust. Thanks for the tip, ExRoadie.

[Sharp]

Hello,

 

I wish I posted this sooner but,

 

Your PWM temps are very high.

74DC / 65DC

 

You should be trying to keep it at 50DC or less.

[mc123]

Hello,

 

I wish I posted this sooner but,

 

Your PWM temps are very high.

74DC / 65DC

 

You should be trying to keep it at 50DC or less.

Sorry, I should have made it clear that the temps I listed were the highest recorded by MBM5 when I was trying to find the max. stable CPU overclock, with two instances of SP2004/Prime95 running over a period of several hours.

 

At idle (but still with OC'ed settings) the average/max temps reported by MBM 5 are typically (degrees C):

CPU: 32 avg, 36 max

PWMIC: 36 avg, 40 max

NF4: 48-49, pretty constant

HDD: 29, pretty constant

 

Ambient temp is 22-24 deg C.

 

But yes, if the PWMIC had been 74 deg C at idle, it would have been much too high, and could well have gotten close to its max. rating of 110 deg C at load. Then I would have earned the title of DFI Global Warmer - as it is, I'm still just a rookie

[mc123]

OK, following ExRoadie's suggestion, I used an abrasive to gently clean the contacts on the edge connectors of both sticks of RAM (each on both sides), and then cleaned them again with isopropyl alcohol. Alas, it didn't help reduce the goofiness of Stick B. Ah well, it was worth a try.

 

For one last check before the final decision to RMA the RAM, I tested each stick in the Asus K8N-based machine that I originally used to validate it, before it ever got near the Ultra-D. Stick B failed miserably, causing BIOS boot failures similar to those I've seen on the Ultra-D with that stick, and often caused the K8N to report that the BIOS checksum was wrong (it probably uses the RAM to calculate and/or store the checksum). In the rare cases I got the K8N to boot memtest86+ from CD, thousands of errors were reported straight away. With Stick A, on the other hand, I didn't experience errors on the K8N (at least not with the relatively limited testing I performed).

 

Stick A ran stably all night on the Ultra-D, at stock CPU/RAM settings and with two instances of mprime, and was still going strong when I left for work.

 

My general impression is that Stick B now fails more consistently, and more severely. The intermittent failures I experienced in the early phases of testing, with long periods of working OK in-between, rather confused the picture and threw me off.

Never assume anything. While your RAM might have been OK a few weeks ago there's no reason to assume it's still OK.

Damn right, ExRoadie! Sad but true.

 

Well, I'll just finish off testing stability with Stick A, and if all goes well I'll get Windows installed and then RMA the RAM. I'm pretty sure I have enough evidence now to successfully pursue my RMA claim. As before, I'll give the new RAM a jolly good beating with memtest86+ - and not least GoldMemory - before accepting it as stable. I'll post my findings - good or bad - with the new RAM here.

[mc123]

I described my problems to the retailer, who agreed that it sounded like a defect in one of the Ballistix sticks, so I successfully RMA'ed the 2x512MB PC3200 Ballistix kit. After all the hassle I've had, I was feeling rather sorry for myself and requested an upgrade to 2x1GB PC4000 Ballistix to cheer myself up. I'll have to pay the difference in price, of course, but I guess my mental well-being is worth it

 

I should be getting the new Ballistix kit within a day or two, and I'll test the living daylights out of it.