I have built 4 machines using identical components.

Antec case with 350W Smartpower power supply
CPU: AMD 64 3500+ (Venice core)
Mainboard: EPOX 9NDA3J
Video: 64 MB GeForce 4 MX440 (Xpertvision)
Memory: 2 GB Corsair ( 2 * CMX1024 - 3200C2)
HDD: 80 GB WDD JB (Parallel ATA)
Zalman Cu7000 CPU fan

Operating System = Win 2K Pro

I think that's the relevant stuff.

The Problem:

Pressing the case power on button will not start 2 of the machines EXCEPT VERY RARELY.
(The other 2 machines function perfectly.)
On SOME occasions, the case and CPU fans will spin as though the PC is about to start.
The onboard debug LED display will also flash indicating power but then it all just "dies".
It's almost like trying to start a car with a battery that doesn't have enough juice.
SOMETIMES, turning off the power supply then back on - via rear rocker switch - and retrying the power on button will get a good start. If the machine does start it runs perfectly.
Pretty much 100% of the time, the machine will start if, while holding in the power on button, the power supply rocker switch is toggled from off to on.

The attempts to fix it:

Shorting the power on pins on the mainboard with a screwdriver.
Result: No joy
Inference: The case power on switch is not faulty.

Test power supply with Antec power tester.
Result: Pass and case fan spins.
Inference: Power supply MAY be OK.

Replace power supply with a new Antec 350W Smartpower.
Result: No joy.
Inference: Power supply is unlikely to be at fault.

Replace mainboard with new EPOX 9NDA3J
Result: No joy.
Inference: ...

Replace CPU with new AMD64 3500+
Result: No joy
Inference: Life sucks.

Replace RAM AND test various slots
Result: No joy
Inference: ...

Replace Video card with identical unit.
Result: No joy
Inference: You have GOT to be kidding.

Reset CMOS via jumper short.
Result: No joy
Inference: ...

Flash BIOS to latest version.
Result: No joy
Inference: Why bother...

I am at my wits end. This just doesn't make sense. I MUST have missed something.

I'd put my money on the case power button any day. If you press and hold it in for a couple seconds longer than normal does it help?

Hi Steve, I did try holding the switch in longer but it makes no difference. Additionally it is happening on 2 machines; 2 cases exhibiting that behaviour would be unlikely at best. I also bypassed the switch by shorting the 2 motherboard power on pins.

From what I can see, this leaves only the power supply- which you tested with a PS tester and replaced as well, right? - and connections. And this sounds like a bad connection somewhere. And this points to the switch:

QUOTE

Pretty much 100% of the time, the machine will start if, while holding in the power on button, the power supply rocker switch is toggled from off to on.

Okay, if you constantly hold a screwdriver down on the two mobo P.Switch contacts and then toggle the power supply, will it start dependably then?

Thank you for the responses; I have found the problem. On a hunch I thought I'd try an AMD 64 3000+ instead of the 3500+ - thinking that maybe the current required to start the system was less for the 3000+. Weird I know but I was getting desperate. Well lo and behold the system words flawlessly. I was able to startup 10/10 - and that's good enough for me. Then I tried a new boxed 3500+ that I had. Lordy, lordy that also works flawlessly. So basically I have narrowed it down to the CPU. Of the 7 Venice core 3500+s that I have, 3 of them are exhibiting this refusal to power up properly. I guess I got a bad batch.

Toppo, in all the years I have been getting split nails mucking around in the guts of computers, this is the first bad proc I have come across. Definitely notify AMD of your experience, it will help them also. Thank you for getting back to us with what the problem was; I was turning it over in my mind last night. Glad you found it!

Hi Ranchhand, Sorry to impart any degree of mental stress but it appears this story is not over. This afternoon (Sydney time) I spoke at some length with Michael the local guy in charge of AMD technical support.

Michael was unbelievably helpful. His spin was that every time he had encountered this issue it was power supply related. During the power up phase the current draw on the power supply can be extreme: based on the configuration I supplied he said 80 amps would not be far off the mark. My counter to this was that I was using quality power supplies (Antec Smartpower 350W), that some of these systems were working perfectly and that I had replaced the power supply - admittedly with another Smartpower 350W.
In support of his claim, I informed him that I had success with an AMD64 3000+. He stated the 3000+ had a significantly lower current draw compared to the 3500+. However, I have also had success with some 3500+s...

So tomorrow I have an audience with AMD where we will test out the defective systems and hopefully get a categorical resolution. Michael is also raising the issue with his Antec contact in Singapore.

I'll relay the outcome of our tomorrow's findings tomorrow.

Hi toppo,

You are not imparting any mental stress at all. I find this extremely interesting; if the prob is, indeed, in the chip, this could be a large help to us all because there will definitely be other folks who will have this problem! At least now we will know what the heck is going on!

As a side thought - how far do you think you would have gotten with Intel, given the same set of circumstances?? This encourages me that AMD wants to find out about this as much as we do!! Go AMD!

Please - keep us up to date; I look forward to your next post.

Well my audience with AMD went very well. Bottom line is the power supply is not providing enough current to start the system.

Michael had previously built a simple adapter where the current drain could easily be observed during power up. Basically, the 12 volt rail just collapsed under load.

Replacing the Antec 350 Watt supply with an Antec 430 Watt supply eliminated the problem 100% of the time. Interestingly, the specs of both supplies were almost identical - according to the labelling on the casing. By far the most significant difference was the wattage.

It was "somewhat" unfortunate for my problem determination that replacing a CPU on a non-working system, converted it to working. Perhaps an electronics engineer might provide an explaination.

I shall return the weak supplies to the manufacturer for replacement.

I must add that AMD's eagerness to assist me in resolving this issue has been most helpful and very impressive. It was brought to their attention yesterday and resolved today. If this had been Intel... who knows...

Although this might be a little off topic, I would like to plug this AMD64 technology. The machines I have built are used for numerical computations and run pretty much 24 hours a day without missing a beat. CPU utilisation is 100%, CPU temperature 35-39 C. That IS impressive.