Why is it impossible to get 512MB of RAM to work?

ARGH! This is the second time this has happened to me:

I'm building a new computer for my family. I have 512MB of PC2700 DDR-RAM in the machine (one stick of 512, not 2 of 256). When I disable "Quick POST" in the motherboard's BIOS, the POST runs, checks all 512MB of memory, then it starts over from 0 and loops through the same test several more times! My first machine I built did this too when it had 512MB of RAM! And Google is absolutely worthless for trying to find information on this (naturally) undocumented error.

On my first machine, I had to remove one of the 256MB sticks to get the machine down to 256MB of RAM. After I did that, the POST ran only once, and Windows could be installed just fine.

On this machine, I keep getting errors during the install of Windows 2000 Pro. It varies between KMODE_EXCEPTION_NOT_HANDLED and PAGE_FAULT_IN_NONPAGED_AREA. The latter error says it can be caused by faulty RAM, but this is brand-spanking-new PC2700 memory, running at its designated stock speed. FSB is set to 166MHz, as it should be.

In both cases, the computers had an Award BIOS. I think the only award that Award deserves is the "****y BIOS that can't friggin' handle 512MB of memory" award. What is it about Award BIOS's and being absolutely incompetent at handling that magical, mystical number of 512MB?

Has anyone had any experience with this problem, and is there any solution?

BTW, these are the specs of the two machines, for comparison:

The first machine:

Biostar M7MIA motherboard
AMD Athlon 1.33GHz (Thunderbird model)
512MB PC2100 DDR-RAM (when the error was caused)
nVidia Ti4200 64MB graphics card
Aureal Vortex2 SQ2500 sound card
ATI TV-Wonder VE TV tuner
generic LAN adapter
ADS RDX-1187 AM/FM radio tuner (ISA bus)
Sparkle 300W PSU

The new machine:

AOpen AK77-400N motherboard
AMD Athlon XP 3000+/333FSB processor
512MB PC2700 DDR-RAM
ATI Radeon 9600 SE 128MB graphics card
Turtle Beach Santa Cruz sound card
WinTV-GO TV tuner
Sparkle 350W PSU

It's most likely bad ram. I had that happen with half my 1GB of RAM in my new computer. First try raising the VDIMM in the BIOS (if you can) a little. If that lets you boot run memtest86. You should also try the ram in the other slot(s) on the board. And if that fails try another board to see if the ram is bad.

Well, something's wrong. Memtest86 reported over 120,000 errors before I finally shut it off. All at about the same general area in memory (right around the 360-380MB mark), and all were single-bit errors in the third-least-significant bit. That is, the test pattern being written was FFFFFFFF, and the result coming back was always FFFFFEFF.

Could such a regular error be a physical memory fault? One would think that defective memory would produce more sporadic or varied errors. Or perhaps it's a bad chip.

Perhaps such regular errors could be caused by incorrect memory timings? That seems to me to be about the only possibility, outside of Award BIOSs just being cheapie chips that are incapable of handling >256MB of unregistered DDR-RAM.

I'll have to do more testing to see which memory module it was; I put in the spare 256MB PC2100 I had lying around and set the host FSB back to 133MHz to run the slower memory. (BTW, having 768MB of memory doesn't change anything - the POST still loops several times before the BIOS decides to boot the system)

Crap, that's happened to me before, but I enabled "Quick boot" or something like that so it skips the memory test on POST. I think I need to get a floppy drive so I can run Memtest.

could be bad ram, could be the slot that you put the ram in on the motherboard.
Some sticks of 512MB ram are dual bank sticks. This will cause issues if placed in a single bank slot. There is also the issue of power draw by the ram, some sticks are just incompatiable wit some MB's. Move the stick around, try new ram...


kiyoshilionz Memtest can be burned to a CD, just download the ISO...

is the ram compatable w/ the mobo?

Compatible? The motherboard accepts up to DDR400-speed memory; I am using DDR333-speed. A memory stick is a memory stick, there shouldn't be any incompatibilities between brands or any crap like that.

I tried lowering the new machine's memory to 256MB, and I still got the looping error at the POST (it cycles through POST several times before booting) -- so it doesn't seem to be a capacity issue. Additionally, as I ran memtest86 on the 256MB stick alone, I got far fewer errors (meaning all those 120,000+ errors were all on the 512MB stick -- how is that possible, as it is brand new memory?). BUT, is memtest86 supposed to re-run the tests again on the same memory? Because it ran through all the tests, and then I noticed it was back on test #4, going to test #5; I think memtest86 is also looping again!

Might such looping be caused by having double-banked memory in a single-bank slot, or vice versa? I also found out that the 512MB has a CL timing of 2.5, so I'm going to check to see if maybe the timings were ratty.

BTW, the memory is by Centon, it appears to be single-sided 64x8.

Admiral Falcon said:

Well, something's wrong. Memtest86 reported over 120,000 errors before I finally shut it off. All at about the same general area in memory (right around the 360-380MB mark), and all were single-bit errors in the third-least-significant bit. That is, the test pattern being written was FFFFFFFF, and the result coming back was always FFFFFEFF.

Could such a regular error be a physical memory fault? One would think that defective memory would produce more sporadic or varied errors. Or perhaps it's a bad chip.

Perhaps such regular errors could be caused by incorrect memory timings? That seems to me to be about the only possibility, outside of Award BIOSs just being cheapie chips that are incapable of handling >256MB of unregistered DDR-RAM.

I'll have to do more testing to see which memory module it was; I put in the spare 256MB PC2100 I had lying around and set the host FSB back to 133MHz to run the slower memory. (BTW, having 768MB of memory doesn't change anything - the POST still loops several times before the BIOS decides to boot the system)

Click to expand...

Never underestimate crappy memory that you can buy from the store. I bought a stick of 512MB PC2700 from BestBuy (had to get it quick, they had a rebate) to replace what I thought was a bad stick of Corsair memory. I had the same problems with the new stick so I assumed it was a CPU or motherboard issue.

Eventually, I ran memtest86, which I should've done in the first place. I found that both the Corsair stick and the new one had errors. Another run to BestBuy gets me a new stick of memory. Just out of curiosity, I tested the 2nd stick from Bestbuy. It had four errors. Gimme a break. A third trip to BestBuy gets another stick. This time, it passed two loops with no errors.

Lesson learned-

-Rebate RAM might be 75% known defective repackaged crap (my theory)
-At the first sign of wierdness in a computer system, I check RAM with a memtester

And the errors you type of are exactly the kind that bad memory has. Even one bit can be wrong and it messes everything up. Randomness isn't necessarily a condition. Take the stick back and test the new one too.

-ben

I had 1 stick of a pair of good Mushkin ram go bad. And I mean that it was bad from the moment I first put it into my computer. RMA with New Egg and it's all good now. You have to expect this to happen every once in a while.

I was once playing hl/TFC when the maps start stretching out into space. So I ask if anyone eles can see this and they all ask if im on drugs. Shut the game down and reinstall and same thing. Uninstall delete the folder reinstall and still have the same problem. The thing is the comp is running fine. At some point I burned a data cd of some program and gave it to a friend and they told me it was corrupt. It took about three weeks for me to figure out that it was the second stick of 512 in slot 2 that caused all the problems.

Ah. Both the 256MB stick (which reports about 20 errors) and the 512MB stick (which got to 120,000 errors before I just turned the computer off) were bought on rebate deals . The 256MB was Kingston-brand, however, so you'd think that it'd have been fine :\ .

Lesson learned, then. Does this sort of thing happen less often with brand-name memory like Crucial?

Admiral Falcon said:

Ah. Both the 256MB stick (which reports about 20 errors) and the 512MB stick (which got to 120,000 errors before I just turned the computer off) were bought on rebate deals . The 256MB was Kingston-brand, however, so you'd think that it'd have been fine :\ .

Lesson learned, then. Does this sort of thing happen less often with brand-name memory like Crucial?

Click to expand...

I would assume that this sort of thing happens more often with generic memory, or with stuff that is pretty cheap at the store. Companies like Crucial, Corsair, Kingston... I'm pretty sure they have pretty high standards for the stuff they sell, so I'd trust them more than I'd trust el-cheapo Fast Eddie's special memory.

-ben

Personaly, companies I trust are Mushkin, Corsair, Buffalo, Kingston and Crucial. Maybe a few others, but I cant think of any. Anyway I've had 3 rebate chips go bad on me. I learned my lesson, and in my main pc I use buffalo, which serves me very well.

I just ran a few more tests tonight:

I think I can ignore the BIOS POST looping -- it appears that it loops three times no matter what memory I have in. I checked this on my own machine by disabling Quick-POST, and then ran Memtest86 on my own 256MB of memory. The memory passed the test without a single error.

It's probably bad RAM on the other system after all.

However, there is one thing about running Memtest86 on the new system (the one I'm building) that worries me. When I ran it on my own machine, the program correctly listed the L1 and L2 caches, and the northbridge of the chipset. When I had run it on the newer machine, it listed the L2 cache as "unknown" and the chipset field was blank.

Does this indicate a possible bad motherboard?