Which version of memtest86+ ? What motherboard ?

or even no fan at all

What ? you tried with no fan ? 🤣

Horun wrote on 2022-09-19, 03:39:

What ? you tried with no fan ? 🤣

These days if things don't work you try with a smaller fan, or even no fan at all and increase voltage 🤣

I guess one of the high address lines is failing, that's why there are lot of errors above certain threshold. I would check for socket and trace damage if the problem persists with other memory module.

My version of MemTest is 2.11, and I just recently got a Chaintech 6BTM.

I can't increase voltage to the CPU or memory, because it's locked.

Any change testing one stick at a time? In case your sticks aren't matched, some boards are picky about the order of the sticks. When using non-matched sticks, I've had Memtest fail with one order and pass with a different order.

Nope. I did that already. I thought it was the memory finally giving out too. But I was wrong.

The thing is, it does just fine at 840mhz, but I'd like to be able to use the full 1GHz. (I was able to use an 866MHz CPU with the last 6BTM I had, before it failed.)

The Chaintech 6BTM has a 440BX chipset which is rated at 100MHz FSB and RAM. Though this board supports 133MHz (being an "enthusiast" overclocking capable board), I would start testing with overclocking anything (ie. run RAM and FSB at 100MHz) . Incidentally, running at 133MHz will run the AGP base clock at 89MHz instead of 66MHz and not all AGP cards like that (not saying this is the issue here, but it is another thing to keep in mind .

Additionally, the 440BX is more finicking about SDRAM density and organization than newer chipset, so the RAM you are trying to use, while working in another (possibly newer ?) board may not be compatible in this case . If the RAM tests fine in another 440BX based board, it is likely fine .
Look here : https://web.archive.org/web/20180103032407/ht … ram_bx_faq.html

EDIT: Corrected typo

There was a well known critical bug in Memtest86 until recently and sometimes it gave errors in the block move test. So it was pretty much useless.

I seem to remember reading that the block move test consists of moving some known data around the memory, and only then checking it matches at the end. If it doesn't match then it throws a lot of errors, but can't tell you exactly where the error crept in.

I had a memory stick that would show up errors fairly consistently using block move, but would only error about once in 24 hours for the other tests (except I think the Mod20 test was also useful in this case for finding errors). So block move was useful for testing various possible causes. In the end it turned out to be an overheating DRAM chip; when I pointed a fan at it then it would pass block move. I then left it running all the tests for 24 hours and it passed.

If the board turns out to be stable at 100Mhz FSB you can use a PIII 900E just like me. 440BX at 100Mhz FSB is nearly as fast as VIA 694X at 133Mhz FSB.

https://theretroweb.com/motherboards/s/chaintech-ct-6btm lists it as 100Mhz FSB board and manual states minimum CPU voltage is 1.8V and max PII (PIII is unofficial). 440BX boards with 3x ISA were early ones, not fully stable at 133Mhz FSB. There seems to be also 2x and 1x ISA version of this board which probably supports lower voltage and 133Mhz FSB setting (probably unstable though).

It was the power supply, tentatively. Doing some research, I found that it was manufactured in 2006, and I am assuming that its capacitors were involved in the Great Capacitor Plague of the early-to-mid 2000s.

The previous power supply I used when I first got my computer case (which supports Pentium 4 and AMD Athlon) doesn't throw up any errors. I'm shocked.

That memtest86+ version is ancient and buggy, why don't you use a recent one?