memtest86 results interpretation

[webbdalan]

This seems like a simple question, but am having trouble finding the answer. In two sequential runs of memtest86 I have had multiple errors reported on test 4 within about ten minutes of beginning of test run.

Question: Does the fact that the blinking cursor is under L1 in the upper left necessarily mean that these errors were in the L1 memory, i.e., the cpu? Another way to word the question ... does the blinking cursor move down from L1, to L2 and then to the memory modules as the tests are performed, or is the location of the cursor meaningless?

My system is turning itself off. I have ordered an upgraded PS, but then thought to test memory as suggested in this forum while waiting.

I understand that even errors reported in L1 could be caused by overheating, voltage etc. At this point just wondering where the test itself is having the problem.

Thanks in advance for helping a newbie.

 

[elpresidente2075]

The location of the cursor is meaningless. It will stay in the same place forever, unless you program it to go somewhere else.

Just so we know, what is the exact address reported by memtest to be failing? It could be useful in determining exactly what is happening. Also, please post the rest of the pertinent specs (cpu model, ram, mobo, etc).

A failure in L1 can be caused by one thing (as far as I know), besides extreme old age, and that's overheating. I would venture to guess that this could be caused by either a malfunctioning/improperly mounted heatsink, or perhaps overclocking beyond the limits of said heatsinking capabilities of your system.

I'm gonna be honest and say that in the hundreds of memtest sessions that I've run, on a similar number of computers, that I've never seen an L1 fail, but there's a first time for everything, huh?

Ok, post those specs and we can move on from there.

 

[webbdalan]

Thanks, elpresidente2075. I had not expected additional help, but appreciate it. Cursor blinking under L1 clearly meant nothing. Did not keep careful track of error addresses since they meant nothing to me, but will collect some more.

Did get 24 errors starting with 9187b9d8 and running sequentially. This one may have been trying to test memory that did not exist. Second run caught 4 errors then 4 more. Did not keep memory address.

I have swapped out memory and video card, added fans, updated all drivers, falshed latest bios. This is a new computer that I should probably ship back, but want to learn from the experience and have specifics for my return ... if necessary.

MB = P5N32-E SLI
PS = 650 Watt GEC Zephyr
CPU = Core 2 Due E6600 2.4 GHz 4MB 775
MEM = Corsair Dual Channel TwinX 2048 MB PC8500 DDR2 1066 MHz EPP - total 4096 MB
GPU = 2 Geforce 8800 GTS OC 320 MB GDDR3

I will be re-running the test and making more careful note of the addresses.

Thanks again.

 

[webbdalan]

Failed again on test four after about ten minutes ...

00107124638, 28, 18 and 8 - 4209.2mb
good 1ce8cr5e bad 1ce8cea1 error bits 0000000ff

Main symptom has been the computer shutting itself off. Usually when playing Titan Quest I have from 1 minute of playing time to about five before it shuts itself off. However, did have it shut down one time when browsing the internet. On the average, however, can run less intensive graphics programs for several hours without a problem.

Haven run Nvidias stress test with no problems.

Came set up as sata raid. I changed to sata Seagate and re-installed XP Pro. Still had same problem.

 

[elpresidente2075]

Just to rule it out being in the system ram, try running memtest with only one stick. This will narrow it down to one or the other, if the problem exists there. Run it with each stick, alone in each slot. This will make sure where the problem is. If the problem shows up in the same spot with both sticks, it may be the slot, so change slots. Continue different permutations of the hardware configurations until you figure out exactly where the problem is. You may be running memtest up to 8 times to find the issue. Hopefully you won't, but thats the worst case scenario.

Let me know what you find.

 

[elpresidente2075]

Just saw your last post. Its possible that you're not being able to supply enough juice to the system, even though you have a 650 watt ps. Its also possible that you don't have the heatsink on straight, causing a lot of extra heat buildup when under stress. To check temperature levels, for the videocards, get rivatuner at guru3d.com and to check the cpu temps, get speedfan at almico.com/speedfan. A normal temperature for the cpu should be in the 30-45 degree C range, and for the gpu under load should be at most 88C.

If you have any questions about any of this, i'm here all day.

 

[webbdalan]

Thank some more! You feel like an oasis in the desert to me.

I am interested in the utilities you mentioned and will probably get them too. In the meantime, Nvidia Monitor shows CPU at 30, system at 34, both GPUs at about 64 centigrade under normal load. Can't run this one while playing a graphics intensive game.

Also have PC Probe II which came with Asus MB. It shows CPU and system temps the same and voltages about right. A little hard to track while playing graphics intensive games.

I suspected PS too and will be receiving a Seasonic SS 700HM 700W tomorrow. Selected this one based on Toms Hardware Review. Wanted to at least try to eliminate the PS as a problem and should come in handy with next upgrade anway.

Am programmer, play with hardware, but clearly no technician.

 

[elpresidente2075]

I'm a programmer myself, and an experienced technician as well, and also a decent overclocker. I believe you'll like those utilities a bit more than the equivalent ones you're using currently due to their enhanced graphing capabilities and lower memory/processing footprint. Definitely very basic on the UI part, but they offer very powerful tools.

Both offer overclocking capabilities built in, but I'd recommend against overclocking through speedfan, since its support is a little flaky, and oc-ing is done much better in the bios anyway. However, RivaTuner has a very good overclocking utility, and I have used it greatly to my advantage.

But if OC is not your thing, don't worry about those features at all. Basically they offer much more advanced graphing and monitoring capabilities than anything that comes with any motherboard/gpu vendor.

To get to your problem, it seems that your temps at idle are nominal. While 64C is much higher than I like seeing on any computer component, that is within the normal operating temps of graphics cards :roll:.

Lets hope the PS fixes the shutting down problem, but the memory errors are a bit more troubling. It seems that your system's problems were being caused by two things: not enough juice and (possibly) a bad mem stick. Like I said, try any and every different configuration you can muster to try to isolate the problem. It may be a bad module, something easily replaced under warranty.

A thought I just had: the fact that the computer is shutting down right after you get a game up and running may be due to the position of the error in memory. Since windows loads up in a rather predictable manner, all in the first part of the memory, and loads up programs in a similar manner, it may be that only when you start loading up a game that takes 400megs of ram that you hit that spot, causing errors and possibly a bluescreen/restart.

Keep testing, and you'll find the problem to rectify. Let me know what you find.

 

[webbdalan]

Memtest was having problems with the fourth memory stick, no matter which one it was. New PS had reduced errors, but still this pattern. Turns out that my board P5N32-E SLI could not handle 4GB of memory without some tuning on voltage and frequencies. Got big help from RAM GUY in the Corsair forum.

Thanks for getting me off to a good start.

 

[Mondoman]

Yes, the Corsair support forum is often very helpful.