Memtest86+ Guide/How To's

[KillrBuckeye]

I'm having some problems with my system that I *think* are motherboard-related. However, I ran memtest86 (at stock voltage, speed, and latencies) for 24+ hours, and through 37 passes (all tests) it reported 2 errors, both in Test 7. My system was Prime95, SuperPi, and gaming stable, but was getting frequent BSODs while just idling. Could 1 or 2 errors every 12+ hours be considered "normal"?

 

[johan851]

Any number of errors is generally considered unstable. Your system isn't stable if you're getting errors in Memtest and BSOD's in idle. You may need to look into RMA'ing your memory or giving it some voltage to get stable again.

 

[Super Nade]

In this case, I believe it could very well be a problem with the mainboard or RAM. The Fortron is almost 100% reliable and your load is not heavy enough to warrent a bigger PSU. Does this happen with a different set of RAM? If so, the mainboard is the culprit.

 

[KillrBuckeye]

In this case, I believe it could very well be a problem with the mainboard or RAM. The Fortron is almost 100% reliable and your load is not heavy enough to warrent a bigger PSU. Does this happen with a different set of RAM? If so, the mainboard is the culprit.

See this thread for a description of what I have done. My leading theory at this point is that there might be something wrong with one of the DIMM slots on the motherboard, since both 512MB sticks, when used independently in the same slot don't give memtest errors or BSODs (with limited testing).

 

[ochungry]

See this thread for a description of what I have done. My leading theory at this point is that there might be something wrong with one of the DIMM slots on the motherboard, since both 512MB sticks, when used independently in the same slot don't give memtest errors or BSODs (with limited testing).

I think your rams have compatibility issue. both are good independently but errors when together. I had same problem and the manufacture rep said mem sticks are not 100% compatible in dual channel. Told me to RMA both sticks for a set of brand new kit. Both rams run fine up to DDR500, but from there on start failing test #8. I can run each up to DDR620 one at the time without any error.

 

[shadowdr]

A side note on the newer versions of Memtest.The third line where it says Memory it also shows the installed amount and then the bandwidth.Note the MB/s as your memory will fail at the same number (within 10 points), regardless of FSB or timings.Even changing the command rate will not affect failure at the same MB/s number.This is important because it shows you the max memory score you can achieve while changing the settings in the bios,or even flashing to a modded bios.This number shows the max bandwidth that the ram will run without Memtest errors.Only voltage can make a difference as well as cache L1 and L2 failures.This can also apply to Windows stability and is confirmed buy Sandra.
I found this out trying to run modded bioses to achieve higher FSB and tighter timings on my NF7S.The latest bios would achieve the same memory score in Sandra with lower FSB and looser timings then any modded bios.This might save someone a lot of time testing different bioses to achieve more bandwidth.

 

[RollingThunder]

Memtest8 Fail

I am a little new to this so please bear with me. I have read most of this thread before posting this. Details of machine in my sig.

Had a random reboot with machine idle so immediately ran a boot disk Memtest8 (v1.65). Many failures, actually to the point of "unexpected interrupt-halted." I am at a 10% over clock with these Crucial Ballistix PC3200 400DDR and never tried higher. No mods to the board, strictly a BIOS over clock.

RAMs and machine have been stable for 2 months under Superpi and Memtest. Memtest in Windows runs fine, no errors as does Superpi and it's always run at 32mb. Never a problem. These RAMs were checked for compatibility on the Crucial site before ordering them so that doesn’t appear to be an issue.

I switched memory slots to make sure it is the RAMs and not the MB. They failed with the same address errors there also so thankfully it doesn’t appear to be my MB.

My question is:

1. I am aware that the DOS boot disk test is more reliable but why doesn’t Windows Memtest show any errors? It’s always perfect. Is the Windows Memtest that unreliable?

I have run Memtest8 in Windows for as many as 20 iterations and it’s always perfect. Same with Superpi @ 32MB. This is a little scary for those who are very inexperienced and only use a Windows Memtest. Thankfully I have been reading and learning on this site to know enough to always run the DOS Memtest first.

 

[{FKR}Loki]

Very helpful indeed

 

[Hellac]

Nice Post - helped me with a problem I have had for months, thanks.

 

[alphagojo]

Thanks alot for this

 

[Rickster]

Help! Im only getting errors on test 8. Theres about 80 errors when I ran test 8. I ran it 4 times; all with errors.

I have G.Skill F1-4000USU2-2GBHZ 2GB PC4000. I've overclocked it to 264Mhz, stock is 250Mhz. All timings is default at 3-4-4-8. Im going to run it at stock 250Mhz when I get back, but is there anything else I can do to rectify the problem?

 

[johan851]

Have you increased the voltage at all?

 

[brz]

i've set in bios timings to 2.5-3-3-6 but memtest86 says settings:cas:2-2-2-5, does mem86 override my bios settings?

 

[Rickster]

Im using the G.Skills Samsung UCCC, they dont like voltage increases. Infact I ran the system under an air cond and a hi speed panaflo and I downed the voltage from 2.6V to 2.5V. I looped test 8 about 22 times with 21 errors. Most of the errors were from times that I was lazy to hold the fan. An occasional error came from times when the fan was there although it was very rare. But this just means Im still getting 1 error rarely. Do you guys try looping test 8 for more than 20 times to see if it's rock solid?

 

[johan851]

If it's rock solid, you won't see any errors ever. Do you start getting more errors when you give it 2.6v?

 

[Rickster]

If it's rock solid, you won't see any errors ever. Do you start getting more errors when you give it 2.6v?

I did notice more errors on 2.6V. But Im going to give it another go with the air cond on and the fan suspended over the drives.

 

[crossed_wires]

I just ran Memtest86+ last night and have a few questions:
ECC: disabled (in regards to cpu -- should this be disabled?)
RAM timing is 3-3-3-8 (good, or no?)

I ran the bootdisk and was surprised to see it start running on its own, when I noticed several steps that someone posted in getting it started...
Anyway, I started at about 9pm last night and it kept going, one test after another. I eventually went to bed and left it running and was surprised to see it STILL running this morning (over 10 hrs at that point). Should it take this long, or will it just keep looping until you stop it? I was under the impression that this was about a 2 hr test.
Good news is that, even after all that testing time, it encountered NO errors.

 

[RollingThunder]

I just ran Memtest86+ last night and have a few questions:
ECC: disabled (in regards to cpu -- should this be disabled?)
RAM timing is 3-3-3-8 (good, or no?)

I ran the bootdisk and was surprised to see it start running on its own, when I noticed several steps that someone posted in getting it started...
Anyway, I started at about 9pm last night and it kept going, one test after another. I eventually went to bed and left it running and was surprised to see it STILL running this morning (over 10 hrs at that point). Should it take this long, or will it just keep looping until you stop it? I was under the impression that this was about a 2 hr test.
Good news is that, even after all that testing time, it encountered NO errors.

You RAMs did ok. It will repeat all the tests until you stop it and it won't hurt anything to keep it running but it's not necessary for more than 6-8 hours if that. ECC should be "disabled." On occasion later tests will find an error which makes your RAMs borderline. Any error is not good but the longer it runs without errors is always better.

If you have value RAMs, the default timings are about right depending on the brand.

Were you having a problem with them or just curious?

 

[shadowdr]

The steps you saw were likely to loop test #5 or #8 as they are the ones that most AMD users test. I normally only loop test 5 for 10-15 passes while testing for highest fsb. If failure is present test 5 will find it.

 

[crossed_wires]

I was just curious. No problems that I've encountered yet, but I've had the computer for a year and just opened it up yesterday and used a vacuum and compressed air can to get rid of all the dust inside.
the RAM is probably "value" stuff, or whatever Gateway puts in their machines. I'll likely upgrade the mb and cpu to AM2 before long, which will enable me to get some good DDR2 RAM at the same time.

You RAMs did ok. It will repeat all the tests until you stop it and it won't hurt anything to keep it running but it's not necessary for more than 6-8 hours if that. ECC should be "disabled." On occasion later tests will find an error which makes your RAMs borderline. Any error is not good but the longer it runs without errors is always better.

If you have value RAMs, the default timings are about right depending on the brand.

Were you having a problem with them or just curious?