K8NHA-grand - board hangs on long "post test"

[no-mbr]

Being a curious geek, I was playing with a box last night and discovered a curious problem.

The mainboard is a K8NHA, nvidia3/250 chipset with two Samsung 3200 matching mem modules. AMD/X64-3200....
http://www.biostar.com.tw/products/mainboard/board.php?name=K8NHA%20Grand
Anyway, the PC "works" fine, but I had been playing SIS sandra and Everest, checking to see if there was any way to boost performance.

I change the "HT" clock multiplier from 4x to 5x and booted. I think, this is supposed to increase the FSB to 1000Mhz from 800.....

When I went into everest to do a memory benchmark, I noticed that HD drive light was staying on, and although the screen showed 1000 on the bus, the mem benchmarks looked slower than the normal 800 setting.

I wanted to see what the current temps, and boot screen said, so i rebooted and, for some reason, I reset the BIOS post to "normal" instead of quick boot.

This is when I discovered that the board will not "post" in normal test mode.
I switched back to "quick boot" and it would post. Later, on, I cleared the BIOS and took out a stick of RAM and discovered it will post normally with one stick of RAM, but never with two sticks installed.

I reversed, and exchanged sticks to verify the both sticks will post by themselves..... so what's the deal?

Anyone with similar boards, could you go try a post, using the mem-count test[non-quick-boot-mode], and see if it works?

Right now, the box continues to work fine, never hangs, or BSODs, but I'm curious as to why this board will not count RAM with two identical sticks installed. And yes this is with BIOS defaults, and optimal defaults.

I can't remember if the board ever posted using "long" post test.

[ski]

IMHO, the normal memory test by the BIOS is a waste of time, since it not only takes a long time, but it normally does not catch any RAM problems.
Suggest just leaving the setting on 'Quick Post', and then run memtest86 if you suspect problems with the RAM. It's a much more reliable RAM test.

[no-mbr]

Thanks, I agree the post mem test does little to verfiy the RAM's true integrity.

However, I hate things that go "bump" in the night, and the fact that this box wouldn't post correctly was stuck in my pants.

As it turns out - it wasn't RAM test that was holding up the post routine afterall. It was actually, some kind of USB/serial bus device "ready/discovery" routine. {no screen output- but going on just prior to mem count}

I have an wireless USB Network adaptor, that turns out to be the offending, I mean "hanging device". The SMC 22662W trips up the "long post" somehow.

In addition, I've discovered that if another USB device is added to the same
bus/header as the the SMC adaptor, then the post proceeds correctly, too.

Meanwhile, back at the ranch, I still haven't gotten a handle on "HT" technology and whether or not this box is going as fast as it could.......

But at least, it still does post correctly, I'll have work harder to break it.

Thanks for reading.....

[ski]

That's nice detective work. Thanks for posting the results.