Can't fill 2 mem slots without crash.

[ROBAXET]

Hi

First, I mounted my computer entirely from new pieces gotten as cheap as possible on ebay.

I did this as a kind of challenge to mount a decent gaming machine on the cheap entirely from my judgment.

I got:

PSU 650W "no name" 40$
I5 2500k 192$
GA-Z68X-UD4-B3 120$
GA-GTX 460 1GB 100$
500GB DRIVE 42$
3TB DRIVE 155$
8 GB KINGSTON 99924 1333 (4X2GB) 50$
TOTAL=700$

As you can see, that's quite cheap for the performance and everything is brand new.

I was ready to risk paying for one defective piece and it seems it will be the motherboard.

Upon mounting everything together, I updated the BIOS to F10 and passed memtest86+ for 5 cycles without a mistake except that the program somehow didn't update the number of cycles (stuck at 100% even if the test continued).

I installed windows 7 64 and was greeted with random reboots, I played with the software for 2 days and ended-up installing windows XP out of frustration.

At least, the random reboot stopped, I noticed that windows 7 used over 2 GB of memory right from the start comparatively to 80 for XP. That would prove to be a coinciding point later.

Upon starting XP, I immediately installed a few drivers and OCCT. Running it normally was fine for 12 hours so I upped the ante and started linpack, a minute later it rebooted the machine, temp stayed under 70.

At that point I threw myself in a 3 days burnout quest of voltages tweaking to pass that test for half an hour and I never succeeded at lasting more than 5 minutes.

Along my testing I came to the conclusion that I had a memory problem.

Two factors gave me hints; windows 7 may have regurgitated every time it gobbled more than 2 GB and OCCT never went beyond showing 1800 MB in testing which I saw as odd, but maybe that was a windows XP problem.

So I went ahead and removed all my memory leaving only one stick in slot one.

I passed OCCT linpack for my half hour quick test.

Then I tested every slot with a different stick to test both the slots and the sticks.

They all passed the test! and that gave me 2 hours of stability testing the machine with one stick anywhere.

Next I planted 2 sticks for channel A

Failed.

And here is my incomprehensible problem; as soon as I plant 2 sticks in that board, OCCT fails.

Otherwise it works fine, so I don't want the classic "Probably a wild PSU" answer.

I'll study any other suggestion before blowing-up that board.

thanks in advance

 

[nikorr]

Well, always more RAM sticks have a higher chance of having 1 faulty one in there.

Did u tried to run these 2 only?

 

[ROBAXET]

Well, always more RAM sticks have a higher chance of having 1 faulty one in there.

Did u tried to run these 2 only?

Thanks for the help

Well, if you read well what i wrote; you'll get your answer.

"and passed memtest86+ for 5 cycles without a mistake except that the program somehow didn't update the number of cycles (stuck at 100% even if the test continued)."

I did that with the 4 sticks in.

later in my post i say:

"So I went ahead and removed all my memory leaving only one stick in slot one.

I passed OCCT linpack for my half hour quick test.

Then I tested every slot with a different stick to test both the slots and the sticks.

They all passed the test! and that gave me 2 hours of stability testing the machine with one stick anywhere.

******************Next I planted 2 sticks for channel A***********************

Failed."


As you can see, every stick and every slot have been tested once for half an hour on OCCT linpack=sucess

Next, all 4 sticks have been tested together in memtest86+=sucess

but plugging more than one stick get an immediate fail in OCCT lindpak.

that would either mean that the motherboard have a problem with more than one stick, or it got a problem with more than 2 GB of memory, or there is an obscure software or BIOS failure somewhere unexpected.

 

[ROBAXET]

Well, I tried more voltage on the RAM, 1.6 volts to be exact, and it did tough a big 28 min at lindpak.

Almost there, but now that it failed so near, i want it to last a whole hour since that may entirely be a lucky lotto win.

Is 1.6 volts normal for a stock speed machine?

The memory is rated at 1.5 volts, the BIOS want to put it at 1.5 volts, if I have to run it at 1.65 volts right out of the box, will it last?

And that memory is supposed to take, in overclocking theory, 1600 Mhz; what kind of insane voltage will it need? 2 volts?

I'll wait for some answers before increasing it to 1.65.

 

[nikorr]

Well, I tried more voltage on the RAM, 1.6 volts to be exact, and it did tough a big 28 min at lindpak.

Almost there, but now that it failed so near, i want it to last a whole hour since that may entirely be a lucky lotto win.

Is 1.6 volts normal for a stock speed machine?

The memory is rated at 1.5 volts, the BIOS want to put it at 1.5 volts, if I have to run it at 1.65 volts right out of the box, will it last?

And that memory is supposed to take, in overclocking theory, 1600 Mhz; what kind of insane voltage will it need? 2 volts?

I'll wait for some answers before increasing it to 1.65.

Try @ 1.55v and increase by 0.01

The RAM may be fine @ 1.65, but the CPU doesn't like it.

The CPU wants 1.5v

 

[ROBAXET]

Try @ 1.55v and increase by 0.01

The RAM may be fine @ 1.65, but the CPU doesn't like it.

The CPU wants 1.5v

I tried that previously, I went anywhere between 1.4 to 1.55 and it didn't pass, that's why, out of discouragement, I tried 1.6 last night, which gave the only minimally interesting result I got thus far.

But I'll still do like you say and retry it between 1.55 and 1.6, maybe 1.55 is barely not enough while 1.6 is barely too much, even if I already don't like the notion that it would be so susceptible.

Thanks.

 

[ROBAXET]

By the way, i did the "CTRL+F1" thing in my bios yesterday, and a very strange thing happened; the option "spread spectrum=on" appeared!

I immediately set it to off and it disappeared from the bios!

I can't get it back!

What was that?

Why would they hide it?

If that thing is on it can cause problem so why would they put it at on and then hide it?

I'll stop here, i'm on the verge of throwing insults...

 

[lafontma]

Is your memory from the HyperX series? the 99924 aint really a model number but most HyperX memory from kingston do require 1.65v to function... which if you do have hyperx that require 1.65v will cause an issue for the memory controller in the CPU

 

[ROBAXET]

Is your memory from the HyperX series? the 99924 aint really a model number but most HyperX memory from kingston do require 1.65v to function... which if you do have hyperx that require 1.65v will cause an issue for the memory controller in the CPU

No, 9-9-9-24 is the timing, the memory is kingston kdr...something and it is written 1.5V on it.

For now, I'm at 1.61 volts and still failing.

I say the motherboard got a bad memory controller implementation, everything work fine with one stick, but it seems that each new slot added need .3 volts more for the whole kit, like the more connections, the more voltage needed. The way I see it is that the there is a missing compensation function; the slots are not seen as independent things concerning voltage.

Evidently, that is just a theory from my part, but I have another older board (asus) with the 4 slots filled with 3 different types of memory of different sizes and I never, ever, touched the memory voltage on it, even if some sticks are overclocked to 400, so I say one board got a good memory implementation while the other not, and the way I'm going, I'll scrap that memory because that sh** motherboard can't work well with the 4 slots full. Sorry.

 

[lafontma]

The memory controller is NOT on the motherboard anymore. It's integrated in the CPU. The only thing the motherboard does is the physical connections.

It might still be the motherboard but just wanted to let you know that there is no more a "northbridge" to control the memory.. its all in CPU

 

[ROBAXET]

The memory controller is NOT on the motherboard anymore. It's integrated in the CPU. The only thing the motherboard does is the physical connections.

It might still be the motherboard but just wanted to let you know that there is no more a "northbridge" to control the memory.. its all in CPU

Yes, you're right, i'm just a little out of patience here.

1.62 volts, it lasted 43 min.

Maybe they forgot to plug a function, is this voltage problem known or it's just me?

 

[ROBAXET]

Did it at 1.625 volts.

Lasted an hour without a problem but the memory is quite hot.

I'll let this tread run for a day to have a few opinions and i'll go do some googling about "abnormally high voltage for RAM"

I wonder how "normal" people (sorry, i'm abnormal and I got a computer geek diploma) deal with the matter, aren't we supposed to plug the parts all in, install windows and say hurray!?

Why don't the manual say the computer won't work if we let stock voltages in the BIOS?

"due to poor design, be advised that you to add .04 volts to memory voltage by filled memory slot" or by 2 GB or whatever the problem.

Anyway, I don't want to give a wrong advice, but I've been sweating about this for a week and even if it work now, there is definitively a problem somewhere.

1.625 volts is abnormal for stock speeds, 1.5 volts memory is supposed to work at 1.5 volts, not 1.625 volts. And did I mention all the sticks are fine at 1.5 volts if used alone in any of the slots? As soon as I plug 2 I need to overvolt the memory, creating more heat and increasing breakdown potential.

 

[ROBAXET]

Well, i received an email suggesting i select a best answer, but since i don't like bullsh**, i can't do it 😱

Nobody here got really interested in my problem, and for those that did try to help a little, the answers were completely off target 😉

Not that it was an easy guess, but there are a lot of people buzzing around Gigabyte here so i thought someone already experienced a similar issue, but it seem i was wrong.

My problem is solved thanks to myself but i can't select myself for a best answer 😛

I still got a few previous question nobody talked about; here's your chance to be helpful

Like "is it normal behavior for my motherboard?"

bye.