Announcement

**linuxguru** · 01-20-2006, 01:10 PM

Re: ooh, this isn't good...

If the caps on the PSU +12v rail went bad, the ripple current that they were filtering will divert through the motherboard caps on +12v. This can increase the current through the connector sufficiently to cause it to overheat and oxidize. That will increase the contact resistance, cause further heating, and thermal runaway to burnout.

**Newbie2** · 01-20-2006, 01:18 PM

Re: ooh, this isn't good...

You're lucky you got warranty. What is the new PSU?

What is Gateway going to do?

**willawake** · 01-20-2006, 01:30 PM

Re: ooh, this isn't good...

4 months for the psu to start going bad is kinda early (probably is that if we take the post problems into account) what about a poorly rated connector or just bad metal oxidising on its own or losing its coating? is that possible?

I have not seen that happen on a 12v 4 pin but have on atx connectors particularly on epox boards. what is the cpu? must be pulling a lot of amps....305w is low unless properly rated imho.

check the fets standing up. never seen that before...

**Newbie2** · 01-20-2006, 02:42 PM

Re: ooh, this isn't good...

By the looks of it that looks like Socket 775. 305W Hipro power supply, that is just not enough to power a system like that. Hipro makes standard average power supplies, in my opinion they are not as reliable as an Enermax or Antec power supply.

Now, what is the manufacturer and model of the new PSU?

(You're lucky dood, PCI-X slot)

**dood** · 01-20-2006, 03:03 PM

Re: ooh, this isn't good...

It's a P4... 3.4ghz, I think? One of the goofy sockets with the pins in it, and the bottom of the processor just has pads...

According to the invoice, they're sending a Hipro PSU to replace it. I don't make the choices... I also don't know what they're going to do with the dead parts... nor do I care. I am assuming they'll want them back.

**gg1978** · 01-20-2006, 03:49 PM

Re: ooh, this isn't good...

Um, this is a 3.4Ghz P4 Prescott we're talking about then. Given that they have a >100W TDP, this honestly isn't that surprising.

**dood** · 01-20-2006, 10:01 PM

Re: ooh, this isn't good...

You'd think that Gateway would spec out a PSU that would be adequate for the base system. We didn't customize these at all except for upgrading from 512 to 1gb of RAM. I mean, I only say that because you'd think they wouldn't want to have to replace motherboards and PSUs constantly...

**Chris1992** · 01-21-2006, 06:50 AM

Re: ooh, this isn't good...

Originally posted by dood

You'd think that Gateway would spec out a PSU that would be adequate for the base system. We didn't customize these at all except for upgrading from 512 to 1gb of RAM. I mean, I only say that because you'd think they wouldn't want to have to replace motherboards and PSUs constantly...

Avsolutely right. I have as my primary box a Dell Dimension 8300 with 250w PSU. I've upgraded it with a DVD burner and additional HDD and it runs fine. Must be a great 250w. Some stability probs tho when making 1:1 DVD copy.

**Rainbow** · 01-21-2006, 09:54 AM

Re: ooh, this isn't good...

I tested a board with P4 2.4GHz (IIRC) using the 200W DTK PSU (photos in other thread) and it worked fine. However, with SkyHawk 250W, it was unable to even power up (maybe just too weak +5VSB).

**jonnyGURU** · 03-23-2007, 01:39 PM

Re: ooh, this isn't good...

Check it!

Same problem! Same pin! Also an Intel board!

Like willawake said, I've seen Epox boards do this on the main ATX and those didn't seem to care what PSU you were using, so from experience with that board, I would be more inclined to say it's a motherboard problem and not a PSU problem.

I'm wondering if we have a similar issue here, but I have to wonder what it is about the motherboard that would cause this.

In my case too, the motherboard and PSU is only 4 months old. The PSU is still fine and has correct voltages on all leads.

Any ideas?

**jonnyGURU** · 03-23-2007, 01:50 PM

Re: ooh, this isn't good...

Here's a thread I made in my forums: http://www.jonnyguru.com/forums/showthread.php?t=1743

**jonnyGURU** · 03-23-2007, 03:32 PM

Re: ooh, this isn't good...

Ah! I found the "attach" button. I'm used to an attach icon in vBull.

Ok.... pics....

Attached Files

**Oklahoma Wolf** · 03-24-2007, 07:53 AM

Re: ooh, this isn't good...

Interesting that the board's silkscreened for an 8 pin EPS connector

**jonnyGURU** · 03-24-2007, 11:09 AM

Re: ooh, this isn't good...

Originally posted by Oklahoma Wolf

Interesting that the board's silkscreened for an 8 pin EPS connector

Yeah. The full ATX version has an 8-pin. I guess Intel figured if you were to put this in a MicroATX with a typical microATX PSU, you wouldn't have the 8-pin and they didn't want to confuse the customer.

**linuxguru** · 03-25-2007, 09:19 AM

Re: ooh, this isn't good...

Neither of the two boards has an AC current-limiting inductor on +12v between the 4-pin connector and the VRM input caps. The VRM input caps seem to be ultra-low ESR KZGs on the upper board. If the secondary caps on +12v in the PSU are comparatively higher ESR, the PSU ripple current will divert through the motherboard caps (in the absence of the inductor). Depending on the CPU and other installed loads on the +12v rail, you're looking at 3 to 6 Amps ripple through the VRM input caps. This will quickly overwhelm the contact plating on the 4-pin +12v connector - these are just stock tin-plated contacts. There's a good reason that an EPS 8-pin connector was specified on the Intel board (and later removed by a bean-counter?).

**bgavin** · 03-25-2007, 12:55 PM

Re: ooh, this isn't good...

Originally posted by linuxguru

There's a good reason that an EPS 8-pin connector was specified on the Intel board (and later removed by a bean-counter?).

It was probably decided on one of those never ending Bridge (conference) calls, where a bunch of faceless names get together and squabble over the design. I suspect, as is often the case, the lowest common denominator won out.

I can't understand the bean counter mentality... they scream over a 10-cent premium cap vs. a 5-cent cheapie, because it "doubles their cost". What gets lost is "doubling the reliability". Add a dollar to the cost of the board, just for the quality.

**jonnyGURU** · 03-26-2007, 01:26 PM

Re: ooh, this isn't good...

Originally posted by linuxguru

...the PSU ripple current will divert through the motherboard caps (in the absence of the inductor). Depending on the CPU and other installed loads on the +12v rail, you're looking at 3 to 6 Amps ripple through the VRM input caps.

Would this still be the case if I'm not measuring more than 50mV of ripple on that connector with a 15A load on it?

If the PSU had inherently bad ripple, like the FSP Epsilon, I'd buy that. Or are there some other variables at play?

**linuxguru** · 03-27-2007, 01:39 AM

Re: ooh, this isn't good...

The ripple *current* can still be fairly high, even with a peak-to-peak ripple voltage of just 50 mV - it depends on the capacitance and the frequency: I = jwCV in the (fourier) frequency domain. Here, w = 2*pi*f ~ 6*10^5 radians/sec (corresponding to a 100 kHz switching frequency of the PSU), and C ~ 6600 uF (2 x 3300uF typically found at VRM inputs of Intel boards). The RMS value of ripple voltage, Vrms ~ 50*10^-3/1.4 V ~ 35 mV. (j is a notational variable that mathematically accounts for the phase lag due to a reactive load - it does not alter this computation, so just set it to unity).

Plugging in the values: Irms = 140 Amps, which corresponds to about 200 A peak.

However, I'm ignoring ESR of the capacitors, contact resistance, lead resistance, etc. in this calculation. In practice, those resistances will limit the ripple current, but at the cost of ohmic (resistive) heating. That ohmic heating at the contact is the main cause of oxidation and premature failure of the contact. If you have 100A (RMS) and even 1 milliohm contact resistance, the thermal dissipation at the contact = 10W, which is enough to melt the tin plate and oxidize it over time, eventually causing thermal runaway leading to burnout.

Older Intel motherboards from the PIII/PIV era had an inductor between the +12v 4-pin connector and the VRM input caps, which limited the 100 kHz ripple current from the PSU diverting into the motherboard caps. That inductor is missing in the boards shown in the pictures above - perhaps another case of a bean-counter intervention.

**Tom41** · 03-27-2007, 11:50 PM

Re: ooh, this isn't good...

Burnt connectors like that can be caused by bad caps, but far more likely is that a bad connection developed. Poor connections have a high resistance, which in turn causes it to heat up, and eventually burns out the connector.

What happens is that one of the contacts develops a poor contact due to oxidization or a loose-fitting contact spring. When that happens, it starts to heat up somewhat, causing the connection to get worse. When that happens, that pin is out of action. The load is carried by the other pins only. Then those other pins quickly overheat with the extra load, and they too start failing. Then any remaining pins quickly snowball into complete failure, with lots of smoke and sometimes fire.

Note that also, excessive current draw can cause it. But the previous scenario is common, and the latter is quite rare.

Announcement

ooh, this isn't good...

ooh, this isn't good...

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment