Closed Solved

Questions about bad RAM

In an ongoing effort to find out what the hell is wrong with my computer, I think Ive finally isolated the problem to RAM. One of the sticks, or more perhaps, is wearing out and is producing blue screens.

First thought it was my mobo, but doesnt seem that way. Then thought it was new video card, but stressing it in furmark seems to say its OK. Hard drives are good. CPU? Prime95 will do its math thing with the CPU forever... as long as its on the setting that uses no RAM. As soon as RAM is added to the equation, cores start producing fatal errors with wrong rounding within seconds.

Going into Memtest86, it does produce errors, however Im confused about why memtest is doing certain things, and wanted to know if anyone knew the answer.

I have 8GB of DDR2 at 800 Mhz, but memtest is reporting it underclocked at 400. Bios says its at 800. Why is this? Is this because something, like the RAM, is damaged/worn out? Is it actually at 400 while in windows even though BIOS is set to 800?

Again, I have 8GB... but memtest reports doing tests on RAM in the ranges of 8192 MB (all I have) through 8960 MB... and these are where the failures are coming from. What/where is this mystery 768 MB of RAM that memtest is testing and finding errors in that does not exist?

Last question, for some bizarre reason, all these RAM errors began to occur precisely right after I upgraded my video card (4850 with 512 Vram) to a video card with higher Vram. Right now I have a 6870 with 1GB. Using onboard video or the older card with less Vram, the computer doesnt BSOD in games. But as soon as its allowed more Vram is when the main memory issues begin to manifest themselves. I have gone through 4 new video cards (thinking it was the cards), all seem to agitate the worn out / bad RAM and cause it to BSOD. Why is this?

I cant seem to find much information on these questions, or any really. Could anyone explain these things to me?
11 answers Last reply Best Answer
More about questions
  1. use cpu-z to read your mb and ram.. the spd info will give you the info of what speeds your ram can run at.
    with ddr2 and ddr2 ram you have the bus speed (400) and the ram inside spped that going to be 2 time the bus speed.
    so if the bus is 400 the ram is going to run at 800. if the bus speed is 533 the ram going to run at 1066.
    the best way to see if you have bad stick or mb issue is take out 1 of the ram sticks and see if the error get worse or stop..if they stop switch the ram sticks. also move the ram stick to each of the dimm of them could have gone bad.
    with your mb see if there any bios updates. if there are use the mb dos be safer so you dont brick the mb.
  2. Thanks for the reply, I downloaded CPU-Z.

    It says the DRAM Frequency is 401.0, and FSB: DRAM 1:2.

    Is that right for DDR2 ram that the BIOS says its at 800? FSB is at 200 I think... so 2x4 is 8....?
  3. Took out 2 of the 4 sticks and stressing with Prime95, same test that failed within seconds the other 30 last times I did it, reporting math errors... its not failing now at all, even when using all of the main memory available to perform tasks. Sticks in are actually just a couple months old; the other 2 I removed are approaching 4 years old.

    I guess one or both of those old ones has reached the end of its service life.
  4. there could be a bad stick of ram..most time with ram the newer ram is going to run at a lower voltage and a different ram timing. (not the ram bus speed). but the timing between the chips.) most ram will be 9-9-9-9-24-2t or 11-11-11-11--24.
    if one set of ram is 11t and new set is 9t you can have issues of the older ram trying to keep up with the newer ram.
    cpu-z spd will show you the timming and voltage of your new ram. you can also google the old ram info and see if there close.
    also some older mb all the dimms have to be the same size and type of ram. on older mb the bigger dimms should be on the back two slots and the smaller ram in the other two.
  5. dimes and nickles said:
    I guess one or both of those old ones has reached the end of its service life.

    Try running with the two old DIMMs instead, see if errors come back or mysteriously went away.

    If you get errors then you know that one of them does indeed have problems.

    If both pairs work fine separately then you have some other problem.
  6. Ran the two old sticks... both check out in both sets of slots. New ram checks out in both sets as well.

    Apparently they dont like each other. They are slightly different, OCZ and kingston, 5-5-5-18 at 2.1 - 2.2v, and 5-5-5-15 at 2.0 - 2.1v.

    Ive always run the OCZ at 5-5-5-15 anyway... but always kept the voltage on auto. I think that might be the problem. I bought the kingston because it was the closest I could find to exactly the same timings/voltage as the OCZ, and the OCZ platinums I have are long since discontinued. Specifically I was looking at voltages to be the same. Perhaps if I set the DRAM voltage from auto to 2.1, that will solve the problem.
  7. Ok, I thought just for the sake of it, I would switch around these two sets of RAM, and see if for some reason, that would make a difference. Left voltage on auto.

    Apparently, it makes all the difference, and there are no more errors. Dropped the timings to 4-4-4-12 at 800, as I wanted them to be on... no errors.

    The OCZ set was in the first set of DIMMs with kingston in the second. Reversing that, all errors vanish.

    Now Im just plain confused. What the heck was going on?

    edit- Oops, got an error in Prime95 after 8 minutes with all the RAM in use... but I guess thats a little more stable than multiple errors in 10 seconds. Still confused why switching the sets would make a difference.
  8. Best answer
    dimes and nickles said:
    edit- Oops, got an error in Prime95 after 8 minutes with all the RAM in use... but I guess thats a little more stable than multiple errors in 10 seconds. Still confused why switching the sets would make a difference.

    Either because your motherboard loads different timings depending on the order in which it detects DIMMs or because one of the chip/DIMMs has a slight signaling issue on one of the pins that makes the DIMM just-barely-not-100% stable when it needs to drive the extra bus load from the other DIMM. So close to (not) working that a few millimeters closer/further on PCB trace length is enough to affect which side of almost-working it is.
  9. Interesting, thank you.

    Did Prime95 again from 20 minutes before I got tired of watching it; no errors still. Played some long games flawlessly, BSOD wise.

    Bizarre, quirky RAM/mobo problem solved, I guess.
  10. Best answer selected by dimes and nickles.
  11. This topic has been closed by Nikorr
Ask a new question

Read More

Memory RAM Graphics Cards