Man.. I feel like a newbie. (music by S.Twain)

Ok, so I got the right stuff I think.
But... I can't overclock it seriously.

I got the Abit IC7-MAX3 M.board.
Paired it with:
Intel 3.0 HT (512kb) 800fsb SL6WK
Corsair Twinx 4000 2 x 512Meg
Ati X800XT gfx, (also tried Ati9600XT)
Antec tru-power 480

Problem 1:
When raising the fsb over 223-225 I get errors in windows/benchmarks.

Problem 2:
When raising the cpu voltage to 1.650(in bios) and above my pc wont boot.
Real voltage is probably around 1.57-1.6V


History of the rig:
I have gone through 3 pcs of Abit IC-7 MAX3. First MAX3 showed ram instabilities at 200fsb running mem at 1:1. Lowering the cpu-fsb/mem-fsb ratio helped stabilize problem but needless to say I wasn't satisfied.

The second MAX3 board was recieved as a replacement was already a reject by another customer and (as I was told) repaired at a facility nearby the MB vendor.
That MB had a burnt traces between PCI slo1 and 2, hell-bent optical ports(since it wasnt packed properly), mem-pins had snagged the egde of memory and was pushed down to a little crumble at bottom of memory slot.

At that point I kinda lost confidence in that vendor and took my money somewhere else. Needless to say I didnt plug my parts into that MB.

And now MB number three: Seems to function but I cant overclock anyway.

The raise-the-voltage-no-boot problem was present at both MB one and three. That makes me think I got a bad 3.0 chip.
Damnest thing is, the chip works fine at 200 fsb, it functions as it was intended so I can't RMA it.
Also, MB works fine at 200 fsb, cant RMA that one either.
MEM works fine at 200 FSB, tested that in 2 different MB.

So far I've tested it all running at relaxed timings for mem, generous voltage for CPU and such.
Raising the voltage from stock to whatever doesnt help my overclocking.
I.E. I get to 220 fsb stabile on stock voltage, running prime95(although I remember now, I run only one instance, DOH!)
It also runs hot according to Abits report tool, but judging from searches I've done thats just abit. I get at stock idle 46 / load 56, overclocked 1.6V(in bios) to 220fsb I get 50 idle / load 62.

I have borrowed a completely different board from a pal to test with, an Aopen AX4SPE-N (a I865 chipped MB with USB problems instead). I'm gonna run some tests and see what happens.

In the mean time, any ideas would be highly appreciated.
Also, to spare me the search and for you to be creative: If the chip is indeed dud, what should I replace it with? (something resonably priced)

What timings are you using? I suspect they are 2.3-3-3-X and not the correct 3-4-4-8

I've tried SPD default, and also manually set it to 3-4-4-8.

What powersupply? Its sounds like a power issue now.

Lol.... thanks for your additional questions, but, all your questions are actually answered in my first post.

I run it all off a Antec True-power 480. Thats 480W from a pretty descent manufacturer. According to Abits own "MB monitor tool" the voltage are pretty stabile at 11.75-11.83V.
I have also monitored the cpu voltage, that one didn't sway alot either. It moved about around 0.04V (idle/load)

I'd first run memtest on that ram using all stock settings (timings and mhz). If your memory is in fact PC 4000 then you should have little problem hitting 250 using default timings. And while I am not a big fan of Antec units I don't think it is at fault here. My suspicions lie in either your ram or the motherboard itself.

deception``

I second the power supply as a possible culpret. Also, are you using a divider or is the ram 1:1? Try to islolate the ram, and use a divider so you can test the cpu's max boot up. Also what do you have your ddr voltage at? Have you locked the agp/pci? I'm not familiar with the Abit board Bios but there are probably a few other settings that should be changed to ensure stability.

EDIT: use a multimeter to test the PS's rails

Thanks for all good suggestions.
I'll try to measure the voltage rails vith my Fluke multimeter.

I have been monitoring my +12V rail.
I bet that the hardware monitor doesnt pick up fast dips, but neither will my multimeter.

The powersupply has been used now for a little over 1 year serving as power for my old rig: 1.8NW @ 2.4GHz. It also featured a GF4 Ti 4600.
I havent had any problems with that rig.

The 3.0NW probably chews up a little more power than my 1.8, but still.. I get the feeling that the PSU is ok.

I have just put together a new test rig.
Aopen AX4SPE-N Intel 865 chipped.
The same 3.0C NW
ATI 9600XT
Corsair 3200 C2 2 x 256Meg
(and my SCSI components, windows XP installed + prime95)

I raised the cpu voltage and noticed that the Aopen SERIOUSLY undervolts by around 0.075V. So at stock my 3.0 only got 1.4-1.41V in IDLE.

So, I set it to 1.750V in Bios and got 1.68V out in bios HW monitor.
Ok, so the NW runs fine in Aopen at 1.68V.

In the "old" rig with MAX3 board I only touched about 1.6V and the whole thing locked up. I'm beginning to suspect BAD mothboard... again... third time now.... this is getting too close to a habit.

I'm going to raise the FSB on this I865 Aopen and see where I land. Oh, and the temps on this board in bios reads 43C @ 3.3G (1.68V) bios idle. That is alot healthier and a little what I expect from the eqiupment.

Dang, I totally forgot to mention the aircooler I'm usin, I got a thermalright SLK 947U + 80x25mm casefan doin about 2500 rpm. I got a quick and dirty application of AS ceramique.

Double check to make sure your AGP/PCI is set to fix and 66/33. Also, disable the last 2 GAT settings.

finally some OC'ing...

Ok, here we go live! this update is broadcasted running at fsb 240 = 3.6GHz.
I set the mem/cpu to 0.33:1 so mem fsb is rather slow. Mem timing is set to 2.5-3-3-8 on my Corsair 3200 volted to 2.7V. Cpu voltage set to 1.725(in bios) given the undervoltage problem cpu sees around 1.65-1.68V.

I cant use Prime95 for more than 1 minute, then it gives me a error.
This time I used two instances of prime set to high heat, inplace FFTs, loading was 100%.

I tried booting at FSB 250, I got to windows but I got constant but recoverable errors, IE no running prime

Some weird stuff at the moment is:
In MBM I see cpu "core voltage 0" and "cup core voltage 1"
One is at expected 1.65V the other is at 1.74 !!!!!!! at idle... Yikes!

Does other members see that big difference between core voltages?!?

When I load cpu core 0 + 1 I get core 0 =1.57V and core 1 =1.66 as per MBM.

Another little "weird" thing is the -5V which is floating at -1.22 !!!

I measured the +12V lines with multimeter. At P4-plug near cpu I got 11.99 +-0.03V, at the big MB connector I got 12.01V +- 0.01V. Measuring time about 10 sec.

Also temps are 35C idle, and I reached 53C when prime95 gave up on me.
I'm going to install memtest and see whats cooking there.
I'm also going to look into GAT thingy.. never thought or known about it... hmm thanks for the "heads up" !

EDIT:
I'm now trying to find a stabile overclock. I havent found the "the last two GAT" thingys yet... Hmm can anyone guide me further?
At the moment I'm doin 230 fsb 3.44GHz at Core0 =1.55V
So far (about 40 mins of prime95) no errors but the temps are up to 60C.

batboy said:

Double check to make sure your AGP/PCI is set to fix and 66/33. Also, disable the last 2 GAT settings.

Click to expand...

Ok done, AGP/PCI locked... actually not a problem but, I cant budge them, they seem to be auto-fixed to 66/33.

However, I cant find any "GAT" settings.
Is GAT settings only supported in I875 chipset?

My overclock has now reached 240FSB with 40 mins of prime stability, ram set to MEM 1:1.30 CPU FSB .
If I set mem ratio to 1:1.60 (meaning ram will run at DDR384) I get errors in prime95, indicating my PC3200 corsair isn't cutting it for dual channel mode.
They are not twinx, they are even of different versions. One is 1.1(BH-6) other is 1.2 (CH-5 DOH!!!!). Well... in goes the Twinx 4000 now. Let's see how that worx out.

Thanks for watching, and if anyone knows, tell me more about the "GAT" thingy.

Memtest good? or bad?

Hehe.. wierd stuff.
I can do Prime 95 with 240fsb mem ratio set to 1:1.60 (155 something fsb)
I cant do prime 95 with 240 fsb mem ratio set to 1:1.33 (190 something fsb)

ok, so memory is to blame.

So, I tried memtest-86 at 190 mem fsb, and it passed.

Mem flys through memtest, but not prime, even though known to me mem-fsb is the only difference.
I'm beginning to put more and more faith in Prime95 for detecting unhealthy overclocks.

Well, to memtests defence I only made one pass, I've read that 4-8 passes gives more certainty.