I'm begining to think Memtest is totaly useless. I let it run for 11 hours last night to test some tighter timings. Got up this morning and had no errors. However, once I started stability testing in windows I had lots of problems. I just could not get it stable. So I went back to some more relaxed timings and I've had much better stability. Is there any particular reason why memtest didn't show any errors?
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Is it just me or is Memtest useless?
- Thread starter Vulcan
- Start date
- Joined
- Jan 5, 2004
- Location
- Denver, CO
Vulcan said:
I've been saying that right along. Memtest stable does not mean Windows stable. Same as being stable in one test does not make you stable in every other test.
I've seen this happen several times before. I use Memtest as a tool to test quick stability. Typically what i do is run 5-20 passes of test #5 only, and if it passes that i take my testing to Windows where it counts.
- Joined
- Mar 7, 2005
I've wondered if you have to go into BIOS and turn off your L1 and L2 cache when running Memtest or does it do it for you? If it automatically turns it off for you, I don't know what the problem is. But if it doesn't and you don't turn off the cache yourself, then you'd really just be testing the cache memory because the cache is write-through, which means that memory reads and writes are cached and it's just reading from your on-die memory. At least that's the way I understood it. I could be wrong. I Probably am. I don't know.
- Joined
- May 30, 2004
Vulcan said:
I see a possible problem here. It's recommended that you run Memtest86 for at least 24 hours!!!
So far, when I ran a 24-hour run of Memtest86 and when it passed, never had Windows crash because of RAM. Thus that seems strange for Windows to crash because of the RAM if Memtest86 passed!!!
- Joined
- Jan 5, 2004
- Location
- Denver, CO
RJARRRPCGP said:
You can't compare a non-Windows based application to Windows stable. There are to many variables in Windows. 24 Hours "Memtest Stable" is just that, 24 Hours of Memtest, doesn't mean anything else.
- Joined
- Nov 27, 2004
- Location
- Port Richey/Tampa, FL
yeah, memtest isn't all that great. I had my bh-5 running up at 250mhz 2-2-2-5 @ 3.2v, and it was memtest stable. I go to play CS:S, and it crashes after about 10 minutes, giving me an "unable to read memory 932u9085798uhj3849f8 etc...." error. I figured out it takes 3.4v to actually be stable enough to game on.
- Joined
- Feb 24, 2005
- Location
- Buffalo, NY
I used memtest to test a crashing box I had and left it run for 3 days straight (in-between classes, MWF). I got back and it was still running, was like sweet, no errors. Boot into Linux, crash, with hex memory error. yea, its useless imho....
- Joined
- May 30, 2004
SlipViper said:
It may be the processor if it still crashes with a memory error even after Memtest86 passed. Have you tried giving it more Vcore? In that case, then it means that it requires a LOT of Vcore to get it stable!
- Joined
- Dec 17, 2004
- Location
- Calgary, Alberta
Clone said:I run
1.memtest
2.pi1M
3.pi32M
4.pcmark04-memory
5.3dm01,03,05
6.prime-memory
7.and last the most sensitive > battlefield2
Ditto to that.
- Joined
- Feb 24, 2005
- Location
- Buffalo, NY
Just a story on that pc btw. It was an old Gateway Dual Processor 350mhz server box in our lab class. Me and my buddy got stuck with it.RJARRRPCGP said:
The assignment over the semester was to set up a network of computers using WinServer 2003, a hardware firewall, router, and switch, with some linux boxes and some linux boxes running VMWare for XP Pro SP2.
We tried to install Linux about 10 times on that box. Had it working the first time, but a vcard error crashed the whole thing and after that we couldn't getting working again. Tried different cards too. It always produced a different problem and we could never make it reproduce the same error twice.
After about halfway through the semester, I carved into the top :LOST CAUSE: lol. Our teacher kept telling us that b4 we formatted he had linux working perfectly on it. Through the course of hours of troubleshooting, we noticed 4 green "wire jumps" on the mobo, since we were looking and troubleshooting so much and so closely. They were almost camoflauged!! We didn't notice them when recording serial #s cuz who really looks for that?! Only after we graduated (same semester) that our teacher told us that Linux was never working on it!!!! He was testing us to see what troubleshooting steps we could do. He told us it was an RMA'd board that he bought back telling the techs to make it work with random hardware errors, lol, hence the 4 "wire jumps" re-routing some of the circuit paths.
SlipViper said:Just a story on that pc btw. It was an old Gateway Dual Processor 350mhz server box in our lab class. Me and my buddy got stuck with it.
The assignment over the semester was to set up a network of computers using WinServer 2003, a hardware firewall, router, and switch, with some linux boxes and some linux boxes running VMWare for XP Pro SP2.
We tried to install Linux about 10 times on that box. Had it working the first time, but a vcard error crashed the whole thing and after that we couldn't getting working again. Tried different cards too. It always produced a different problem and we could never make it reproduce the same error twice.
After about halfway through the semester, I carved into the top :LOST CAUSE: lol. Our teacher kept telling us that b4 we formatted he had linux working perfectly on it. Through the course of hours of troubleshooting, we noticed 4 green "wire jumps" on the mobo, since we were looking and troubleshooting so much and so closely. They were almost camoflauged!! We didn't notice them when recording serial #s cuz who really looks for that?! Only after we graduated (same semester) that our teacher told us that Linux was never working on it!!!! He was testing us to see what troubleshooting steps we could do. He told us it was an RMA'd board that he bought back telling the techs to make it work with random hardware errors, lol, hence the 4 "wire jumps" re-routing some of the circuit paths.
wow, your teacher sure is nice to you
but seriously, thats a pretty creative way of trying to gauge your ability. shows that at least you are pretty determined....
- Joined
- Jul 24, 2002
- Location
- Seattle, WA
Exactly. Memtest is hardly useless...it's a great way to find a rough ceiling for your memory and avoid trashing Windows. If I had to run Prime every time I wanted to try a new memory configuration, I would've had to reinstall windows a bunch of times by now.
- Joined
- Mar 8, 2004
I just do prime95 and memtest. Super Pi is useless since I get errors in memtest and in prime95 but super pi is able to complete calculations.
Memtest as the title says is a test of memory, so if your memory is capable of running 600mhz then it will be able to pass memtest. On the other hand cpu might not be able to run 5ghz.
Memtest as the title says is a test of memory, so if your memory is capable of running 600mhz then it will be able to pass memtest. On the other hand cpu might not be able to run 5ghz.
- Joined
- Feb 24, 2005
- Location
- Buffalo, NY
Haha...our teacher is amazing. After we got really upset for a minute, we thought about it and were like, wow....that was an awsome way to test us! He is an Old Skewl guy who wears hoodies to class and who has trouble (on a quickness level) of navigating a GUI, but give him a command line and he flies through everything. "bah, just gimme the command line!!"el_leumas said:wow, your teacher sure is nice to you
but seriously, thats a pretty creative way of trying to gauge your ability. shows that at least you are pretty determined....
Greatest quote from him:
Me: You drinking tonight?
Partner: Of course! You know what man, I want a beer right now.
Me: Haha, sorry, only pop here now...
Partner: Ya, I wish the schools vending machines had beer in them.
Me: Damn straight.
TEACHER: Naw, I wish the vending machines had the reefer.
Me, Partner:
I don't know what floored us the most, the fact that he said that, or the fact that he called it "the reefer."
edit: Sorry about the thread jack....
Last edited:
- Joined
- Aug 12, 2005
- Location
- Montreal, QC
I like memtest. Yeah, most of its tests won't find out anything... But if they do find errors, you know for sure that something's wrong. ;p Not to mention pinning down bad RAM blocks has let me smoothly use faulty RAM on my server for 2 years now!
- Joined
- Jul 24, 2002
- Location
- Seattle, WA
^ You can have a system not address faulty blocks? How does that work?
- Joined
- Aug 12, 2005
- Location
- Montreal, QC
With a badram, a Linux kernel patch for which you can generate a bad memory block report with memtest86... I don't know if it has an equivalent outside of Linux, but it works surprisingly well on my server. It boots smoothly using old faulty SDRAM sticks... One of which had a terribly mangled area right at the start which wouldn't even let it boot without a kernel panic.
But then again, the only memory intensive things run on it are folding and some postgreSQL playgrounds.
But then again, the only memory intensive things run on it are folding and some postgreSQL playgrounds.
Similar threads
