Memtest passes all 4 sticks of ram when ran solo, but not when together

Question

I've got a computer with some mixed ram. I started getting BSOD's so I suspected a ram issue. Pulled all ram but 1 stick and tested it with memtest, then repeated with the process with the other 3 sticks (still one at a time). All tests passed with no errors.

Their are 2 pairs of ram, each pair coming from their own specific manufacturer. When I test the matching pairs together, they all pass, but as soon as I mix in 1 pair with another, I get 1000's of errors.

Is it generally bad to mix ram manufacturers? The ram is the same type, with similar specs (the cas timings vary slightly, but other than that it's all PC3200 DDR ram).

Answer 1

Yes, it is always a good idea to have all the same ram. Not just manufacture, I like to say manufacture, speed, layout type, and even Batch. Nothing stagnates in the high tech world. The ram you buy today, might not use the same modules or controller of the ram you buy next year.

If at all possible get the closest thing to it you can. Idealy get it all at one time, get it in a "set" if possible (its easier to return :-) And get one of The exact model they listed in the "motherboard compatability chart" for your board.

There is hope though, many boards are pickey when the ram chip ammount is increased, and there are even special adjustments for how it changes Only because the ammount that is there changes. Like Scew or is it Squew?? Some boards the voltage regulation isnt perfect , so changing the voltage up a bit might help. The slight differences in "timing" things can be tweaked around a bit.

I could not get Post even, with a ram combination I used on default settings. But I was a persistant litte cuss , and I knew how pickey my board could be. I went back to the 2 sticks, then way underclocked the ram. Using the ram Divider (666 800 1000 1600 or 1:1 5:4 4:3) Then tweaked the timings so they had longer "wait states" T-ras cas and whatever to higher numbers. Then I put the rest of it in . After I got some slow combination working, I Wrote it Down. Then I just tweaked it back till it failed again, writing down each change I made. Within a few hours I had it up to 99% of where it was originally, and everything working perfect. It was just like a 9 dial combination lock :-)

So it might be possible, and 4 sticks gets pickey for a lot of people. Many people just concider the ram to be bad or incompatable, and take it back. Saving themselves a lot of trouble. Some of us are too thick to give up so easily :-)

Answer 2

Even if you start with 4 of exactly the same sticks of RAM you'll still possibly run into issues, you will need to run memtest86 to check that they are OK when "paired together".

Running sticks that aren't exactly the same AND weren't purchased at exactly the same time is a recipe for disaster... After you get those two things right you'll still need to memtest86 if you have more than 1 stick in your computer (if you want to be able to rely on the stability of your computer).

Also, a good quality power supply AND and UPS (to clean your power) will also help.

Moral of the story, if you want a stable computer you need to buy good quality parts and test your RAM (load testing your CPU e.g. Prime95 and your GPU e.g. 3DMark is also good) is the only way to really be sure you have a good stable PC. Best to do it up front so you can move on and use your computer with confidence.