Memtest+ slacking......

[Poci]

Is it just me or does anyone else think that newer versions of memtest+ arent as strict as old ones when it comes to detecting errors?For example my memory can pass memtest fine @250 mhz 2.5-3-3-6 timings but in windows I get memory errors upon closing IE for example,if I run at rated 3-4-4-8 timings its rock solid stable though.
So basically my question is are there better options then memtest+ for testing RAM?

On UBCD (ultimate boot cd)im sure many of you heard of this thing.
Well its has more then one memory tool:
memtest+ and its predecessor are of course there but theres also "Windows Memory Diagnostic,DocMem RAM Diagnostic and TestMem4
Would any of these be a better option?
I want the most strick test posible

[oralpain]

Memtest86+ is not perfect, nor is any test.

Ones that run at boot are generally not multi-threaded, and that's a problem in of itself.

To ensure memory stability in my system, I run three loops of S&M's memory stress test, each one immediatly after the prior one, run 500-1,000 loops of memtest86+ test #5 (this test produces the most heat), then do 24 hours of prime95.

Generally S&Ms test is the quickest way to find errors, in my experience. Still, you need multiple tests as nothing is perfect.

I haven't found docmem or windows memory diagnostic to be any better than memtest86+, but it's conceivable that some problems may be found sooner with them, on occasion.

[Brandito]

memtest doesn't put the whole system under stress the way an operating system will.

memtest is really best for finding faulting memory, it does work well as a stress tester but since it doesn't stress everything in your system it's not an be all end all stress tester.

orthos, stress prime, and prime95 do a good job of memory testing as well. i would run memtest86+ as well as in windows based stress tester to really make sure i have a stable oc. 12+ hours is what i do, everyone likes to test for different amounts of time, do whatever you're comfortble with. running overnight is the best time.

the bad thing about windows based stress testers is they don't have access to the majority of your memory the way memtest86+ does. but since you probably ran memtest86 for a number of hours you should already know whether you have bad dimms or not, so really what priming is doing is testing the actual stability of the ram and/or cpu while in a windows environment, which is more real world than what metest86 does.

btw, when finding a max overclock i usually speed things up by running memtest86+ test 5 a few times, it runs relatively quickly and usually is the test that finds errors first if at all. it's a good way to know what your chances of getting into windows safely are going to be without risking your filesystem.

[alex666]

When I build a new system (in process of doing so now), the first test I run is memtest86 at various settings/speeds to test the limits of my memory and see if the memory is faulty. Memtest is good for that. I'll run a couple of loops of the whole test, then 10 - 15 rounds of test #5, because that's where I always tend to get errors. Then from windows I run orthos. Like brandito said, it also stresses your memory along with your cpu of course.

Personally, I only run these tests for an hour or so. I've never found it necessary to run them longer, i.e., my systems always have run stable so long as they've passed an hour of memtest86 with plenty of runs of #5, and an hour or so of orthos. Others may jump on me for that statement, but that's been my experience.

marty

[cwolv]

i didnt get mmeory errors closing ie until after i updated to ie7. ive always blamed microsoft coding, not the memory

[Poci]

Yeah well MS coding hasnt given me any errors with memory at correct timings

[_dangtx_]

when u oc, its good to isolate mem but once u stress test it u may see errors, and its not the ram but the chipset that craps under whatever u put it tru..or maybe teh cpu? soemthign warms up? many a thing

[Poci]

Nah trust me dan it aint mobo or cpu,it was RAM causing the error.Timings were too tight.

[madhatter256]

Well if u get errors in windows when u oc ram, but not when you run the ram at stock speed, then it is your oc! You'll need to downclock some.

Also, for testing/stressing I run orthos + super Pi in windows, and memtest outside of OS.

[Poci]

Everythings stable now.
The RAM simply errored when I tried running it @ 250 with tight timings.
Loosening timings back to rated timings solved it.
But my whole point of this thread is that memtest+ did NOT catch the error

[Brandito]

Everythings stable now.
The RAM simply errored when I tried running it @ 250 with tight timings.
Loosening timings back to rated timings solved it.
But my whole point of this thread is that memtest+ did NOT catch the error

and it likely shouldn't catch the error.

the dimms could work just fine when they're the only thing being stressed, start stressing the rest of the system as well and you can get problems not found otherwise.

you also have to take into consideration how long you ran memtest and what tests were run. one pass of all the tests may not always find an error, you might not come up with errors till the 10th run.

[alex666]

The other test I have found strangely but consistently useful in ferreting out system problems is 3dmark2001se. Yeah, I know, it's rather dated now, but if there is a problem with your system, 2001se will freeze in a heartbeat. More so than the other 3dmarks. At least that has been my experience, so who knows, maybe it's somehow been unique to my experience. Also, I have found this with XP. I don't know about Vista.

marty

[Poci]

and it likely shouldn't catch the error.

the dimms could work just fine when they're the only thing being stressed, start stressing the rest of the system as well and you can get problems not found otherwise.

you also have to take into consideration how long you ran memtest and what tests were run. one pass of all the tests may not always find an error, you might not come up with errors till the 10th run.

I usually leave it running for 8-12 hours

[oralpain]

My most recent hardware problem was confirmed with memtest86+. I had been getting some unusual behavior and the occasional error/crash-to-desktop since building my system, but none of the stress tests I ran could isolate the problem for sure. S&M would sometimes show a memory error if I forced it to run a second or third pass right after the first, so it at least pointed me somewhere.

Running memtest86+, I would very rarely get 1 or 2 random errors a few passes in, but then 100 more passes would go by with no problem. The only way I could be sure the error would appear was to do at least 165 passes at a time. So, I tested each stick of ram, and then each ram slot, at verying settings, with 1,000+ passes each and confirmed that dimm slot #4 on my motherboard was defective.

I then teook the boardout of my system, looked over every visible trace going to DIMM 4, and I found a small area ont he back of the mobo that had been scratched, along with an attempted, but apparently failed, repair.

I RMAed the board. Problem solved.

Also, allowing the whole set of tests to run is a waste of time. Had I let every standard test run on memtest86+, it would have taken weeks, not hours, to find the problem. I'll run one loop of the standard tests, then I will set it to run test #5 and leave it. If it passes #5 for a sufficent ammount of time, I'll run another loop of the standard tests while the RAM is still scorching from test #5. The if I feel like it, I'll run the bit fade test.

The other test I have found strangely but consistently useful in ferreting out system problems is 3dmark2001se. Yeah, I know, it's rather dated now, but if there is a problem with your system, 2001se will freeze in a heartbeat. More so than the other 3dmarks. At least that has been my experience, so who knows, maybe it's somehow been unique to my experience. Also, I have found this with XP. I don't know about Vista.

I have found this to be true on some occasions, untrue on others. Back when the 6800 was a new card, I bought a gainward 6800GT golden sample and after upping the voltage a notch it seemed like the card was stable at 440MHz (a rather impressive clock at the time on stock cooling), everything usually ran fine, but I had this naging instability. I ran 3dmark 2001SE at the highest settings possible, no artifacting, no crash. Then I tried what I thought would be the much less stressful standard settings -- hard-lock 3 seconds into the first test.

After this puzzling and very repeatable result, I went back and tried different clocks and voltages on the card. 1.45v on the GPU was stable to 405MHz in the 3dmark 2001 standard test, but 1.35v was stable to 420MHz on the standard test. Temps were never an issue and no artifacts ever showed up in artifact testers, even at much higher clocks at 1.45v.

I came to the conclusion that at 1.45v the card was demanding more power than the single molex connector on the PCB could provide and that the much higher frame rate of the standard test was actually more demanding when it came to power consumption than the greater feature/memory use, but lower fps, of the test on higher settings.

In other situations I've had 3DMark 2001 loop for hours with no problem when Opera would crash, or SuperPI would crap out during the first 3 second test. Nothing is perfect.