How reliable is Memtest x86?

[Suups]

Having problems with my 200 FSB machine. Every so often, particularly under load, it BSODs. Seems to be all over the shop - portcls.sys, bootvid.dll, hal.dll.. but most of the time it's ntoskrnl. If I bring the FSB down to 166 the problem seems to vanish.

Fortunately I have another machine I can swap parts around in, but the CPU is a native 133 FSB, OCed to 166 with 166 RAM so I can't really test it properly. I've tried swapping the CPU and RAM to the other mobo and repeated the error, so it doesn't seem to be linked to the vid or the mobo so I think it's got to be either CPU or RAM. Generally I would associate this kind of error with RAM - I have 3 sticks, 2x512 and 1x1024, but any combination seems to kill it so if it was RAM all three would have gone at once. Plus I've run the latest version of memtest x86 overnight on 200FSB with all 3 sticks in, 8.5 passes with no errors. Which leads me to suspect the CPU, but this isn't the sort of error CPUs usually present... at least not that I've seen... which makes me wonder whether I'm getting a reliable result from memtest.

Running something like Burn In Test will BSOD the system in about 5 minutes or less. I should also mention that I'm not overclocking at all.

Everything is only a few months old, still under warranty so it's purely a case of finding the problem - any suggestions will be greatly appreciated.

 

[madthumbs]

Running the ram synchronous with the front side bus should give you better stability and performance. iows run the ram at 166.

Other things to consider would be the power supply, motherboard bios, chipset, ram /mobo compatability.

 

[Suups]

Running both the CPU (an XP 3200 400FSB) and the RAM (crucial DDR400) at 200 causes the system to BSOD. Running both at 166 seems to be stable. Running the RAM at 200 and the CPU at 166 also seems to be stable. The PSU is an Enermax 350w, with no peripherals at all but it is running 3 disks. The vid card does not require additional power.

I had considered the MOBO and PSU etc., that's why I swapped the CPU and RAM to the other machine. Both are A7N8Xs, and both with the latest BIOS so 200 FSB speeds are supported. Heat is well catered for, even under full load it has never exceeded 45C, plus with ASUS COP on those boards.

Would the fact that it runs smoothly at 166 rule out the PSU? Or could the slower bus speeds allow the CPU to compensate for dodgy power? Hmmm... the other machine I tested the bits in does have a crappy PSU... what would be a good way to test this out?

 

[madthumbs]

Now that you mention the motherboard, I have the a7n8x deluxe with Corsair PC 3200. I have to run it at 166 also. I believe it was a problem with this particular ram. You could probably find out more about it at www.amdmb.com forums.

 

[Fern]

Hey Suups,

Might check the mobo's manual. Other NF2 say 200 FSB not supported with three DIMMs. Further those mobo's don't appear to like 512MB sticks and higher as much as 256MB sticks.

While peeps with three sticks (2 x 256MB and 1 x 512MB) report that they are able to reach and exceed 200FSB. I haven't seen anyone do it with the amount of ram you have.

Peeps with two sticks of 512MB report 200FSB works, but that the OCing suffers. IIRC something being double sided DIMMs at 512MB, or too many modules on the DIMM etc.

You have BOTH 3 sticks AND large DIMMs. I think this is your problem.

Further, 8.5 passes or whatever is no where near enough time to test with Memtest86 IMHO. In OCing I've had my ram fail on the 24th hour. 8.5 passes would only be about 2 hrs or so I guess.

Fern