RAM In Matched Pairs

DexterDexter Vancouver, BC Canada
edited January 2004 in Hardware
Ok, everyone talks about, buy how important is it really to have RAM in matched pairs?

Here's my situation. I am trying to troubleshoot some random PC reboots/freezes on a friend's XP system. It was crashing randomly, alot of IRQL_NOT_LESS_OR_EQUAL's, and PFN's, and if they left the computer sitting for an hour or so, or were doing stuff on it, BSOD (I had them disable the auto-restart so I could get an idea what the errors were.) So I updated their BIOS, adjusted some BIOS settings for optimal performance, downloaded and installed the latest VIA 4-in-1's, and updated their Nvidia drivers.

That seemed to be more stable, it did not crash every hour any more, it was stable after nearly 24 hours idle time. Yaaa me ;)

But, I then wanted to test their system playing HALO, as they had been having problems with the game freezing (see more about the problem in
this thread. )

I did notice that the RAM was installed in 2 sticks: a 256 MB, and a 128 MB, both PC133, but of different brands. The CPU is an Athlon 1700+.

When the Dad bought the computer from a local store, they asked him if he wanted to add extra RAM. He said yes, so they installed the extra 128.

So, how improtant is matching the brand / size of the RAM? Could it be causing problems with the game crashing as detailed in the other thread?

Any opinions on paired RAM welcomed.

Dexter...

Comments

  • primesuspectprimesuspect Beepin n' Boopin Detroit, MI Icrontian
    edited January 2004
    hehehehehehe.......
  • DexterDexter Vancouver, BC Canada
    edited January 2004
    primesuspect wrote:
    hehehehehehe.......


    I ran MEMTEST already Prime, came back with zero errors. ;)

    Dexter...
  • MediaManMediaMan Powered by loose parts.
    edited January 2004
    Simple...easy peasy.

    Matched pairs is always better. Matched in the sense of equal size...not this marketing term of "matched" that goes with the high priced stuff. BUT...generic, cheap, bargain ram is often the culprit. "Made on East Hastings" and sells for cheap is often the cause. MICRON or SAMSUNG is better quality generic ram.

    Try pulling the 128 and run a disk check and other maintenance (defrag) and see if the system runs well on the 256 alone. That could help identify the 128 as a possible culprit.

    Sometimes...and it just happened to me tonight. Certain motherboards and certain ram combinations just don't agree all the time. Sometimes it can be stable in DIMM1 but move the ram to DIMM2...down she goes.
  • DexterDexter Vancouver, BC Canada
    edited January 2004
    MediaMan wrote:
    Try pulling the 128 and run a disk check and other maintenance (defrag) and see if the system runs well on the 256 alone.

    I was just thinking of yanking that 128 and seeing what happens. I'll try that tomorrow, thanks MM :)

    Anyone else want to chime in their 2 cents on this topic?

    Dexter...
  • Geeky1Geeky1 University of the Pacific (Stockton, CA, USA)
    edited January 2004
    IMO, the "matched pairs" of RAM they sell now for dual-channel MB's are a waste of money. Beyond that, I have never, ever had a problem caused by mixing/matching brands, sizes, and speeds of RAM, as long as they all met the minimum requirements for the CPU's bus speed, etc.
  • qparadoxqparadox Vancouver, BC
    edited January 2004
    Tell him that accessing things randomly is bad and thus he should toss out his RAM. Scrape up some money and invest in some LAM (Logically access memory). Ok that was lame. But as MM said pull either stick and test ... then try the other and test again. I'd pull the smaller one as well. I'm assuming if it has PC-133 that its beyond its warranty date. If you wanna try some other pc133 I've got several sticks in various computers that I could trade straight across (assuming its not completely dead)/ loan you (notice the locations) if the guy really wants 384 MB. PM me after you try swapping the sticks if interested..
  • EnverexEnverex Worcester, UK Icrontian
    edited January 2004
    Geeky1 wrote:
    IMO, the "matched pairs" of RAM they sell now for dual-channel MB's are a waste of money. Beyond that, I have never, ever had a problem caused by mixing/matching brands, sizes, and speeds of RAM, as long as they all met the minimum requirements for the CPU's bus speed, etc.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited January 2004
    "Matched" pairs are often better for overclocking since the sticks are generally from the same batch/week. For general use and no overclocking, it's useless.
  • DexterDexter Vancouver, BC Canada
    edited January 2004
    Mackanz wrote:
    "Matched" pairs are often better for overclocking since the sticks are generally from the same batch/week. For general use and no overclocking, it's useless.


    There is no overclocking in this system, just auto settings inthe BIOS. Thanks Mackanz. :)

    Dexter...
  • Straight_ManStraight_Man Geeky, in my own way Naples, FL Icrontian
    edited January 2004
    Ok, here is an oddball sounding thing that has happened over the years to me....

    I would try switching the two sticks, put the one that is in the first socket in the second, then see what happens after you tell the BIOS to update the DMI table by telling it to reallocate all resources. this, among other things, will force BIOS to rewrite timings for RAM, and some boards just detect first stick when this is done. Stick slowest timed (not rated, but timed, as modules can vary and thus fine timings can vary) stick in the first slot is what you want. What it looks like from symptoms, is that you had either bad stick (but the memtest86 successful pass says NOT that) OR the BIOS set overall RAM timing to faster stick (likely, with what you are getting, or bad video card onboard RAM or massive IRQ stacks with video card in one of the stacked IRQs (run System Information and then click in Hardware section the Conflicts\Sharing line to see what Windows thinks is stacked and see what else is in stack)). If no good, the stick in socket 1 (which to BIOS is Dimm0) now is incompatible with the stick in socket 2(which to BIOS is Dimm1) UNLESS this is a 3 or four memory socket board. If this is a 4 socket or three socket board, try using sockets 1 and 3 and not 1 and 2. Some boards with more than two sockets use paired socket banking, and can boht time by bank and have a 256 MB limit per BANK (thus sockets one and two combined limit can be 256 for both on four or three socket boards or boards with two different memory socket types on board).

    The IRQL thing can be too many things on one IRQ, bad video card RAM, or incompatible or failing memory or a corroded RAM socket pin or pins (If you ran Memtest86 and did not run all tests, plese do that and let it hunt bit failures, ok???). It can also be other things, but lets start with these basics.

    A Windows IRQL fail is a Logical IRQ failure. divide by zero on a Logical IRQ is usually caused by a test signal on IRQ returning NULL which results in a mess. I HAVE seen a new CMOS battery resolve this if this happened after RAM sticks worked for quite a while after install and box went UNSTABLE after being stable for quite a time after hardware install-- I do not have a set of times for what happened, how long it took since addon happened for this to occur or if this happened when box was up for a long time or if it happens also with box up for just a short time after being off for a long time. Knowing how cold box was when this happened would let me zero in more after the above things are done.

    John.
Sign In or Register to comment.