Few questions about capacitors.

[Mod]

Hello. I'm curious about some of these ratings I've been seeing in the various tech docs. Is a higher ripple better? Also, how about ESR?
This is to replace 5 GSC caps that went to absolute hell after I put in my Radeon 9800XT and a new A+GPB 510w psu. The cpu is an Athlon XP barton core overclocked to 2.65GHz (not PR, true MHz). Put a bit of strain on the caps I believe and now my +5 is unstable and drooping to 4.75v even with an 800MHz Athlon installed sitting idle. The board is the original Epox 8RDA.

My system experienced instability and blue screens out of nowhere.. I checked my water-cooler but the temps were fine, everything looked ok until I saw the leaking/buldging caps.

I also added a resistor to a mosfet near the chipset and increase the VDD (chipset) voltage to allow higher FSB speeds. As it turns out, a GSC 1500uf/6.3v capacitor near the chipset was buldging as well.

I've already replaced the single 1500uf 6.3v with a Sanyo 1500uf/10v. Now all I have left is the four 2200/10's near the cpu and 4 other 1500uf/6.3's I'm going to replace anyway. The Sanyos near the cpu are fine, I trust.


Here is what I see in the documentation for Nichicon caps:

HE Series 2200uf 10v 10x31.5 "Low Impedance"
Impedance(20C/100kHz)0.031
Ripple (mA rms 105c/100kHz)=1910

PL Series 2200uf 10v 12.5x25 "Extremely Low Impedance, High Reliability"
Impedance(20C/100kHz) 0.034
Ripple (mA rms 105c/10kHz-200kHz) = 1710

I've already ordered the HE series 2200uf/10v. Should I have bought the PL Series instead? The numbers seem pretty close (but I'm a newbie when it comes to capacitors, so...). Also, can I use a higher uf rated capacitor (say, 2200uf) in place of the 1500uf that failed that was closest to the chipset? How about using four 3300ufs in place of the dead GSC 2200's?

If the PL is that much better I'll just order those and put the HE's into another board I have here. I really wanted the Rubycon MCZ 2200's but uh.. I'm sure I'd have better luck finding gold in my back yard.

Any suggestions on would be greatly appreciated. Thanks.

[Mod]

Got everything modified.. left the Sanyos in. Well, guess what? Upon inspecting the close up pictures (I didn't pay much attention to the Sanyos other than getting pics), I notice *3* of them are slightly leaking on the top. No wonder my +5 rail on the board was still ridiculously low! With my Athlon XP underclocked to 600MHz it was 4.81v (should be 5.00 of course), and at 2.65GHz it was 4.4v.

They aren't bulged in any way, but it looks like they're leaking. They are the Sanyo WG series, 3300uf 6.3v.

Here is a pic of two of the Sanyo caps, 6 in all, 2 not shown. 3 supposed bad and 1 good in these pics. Please let me know if you think these are leaking or possibly just dirty...



Going to replace those with some Nichicon 3300uf 6.3v 12.5x20mm (UPW0J332MHD) caps. Hopefully then my board will be good to go.

Btw, I also wonder why the Sanyos went to hell. I had an 80mm and 120mm fans blowing directly onto the mosfets(heatsinked) and caps. Neither of which ever got above 30C (tested via digital temp sensor). Had the board about a year and a half. Any ideas what happened?

[kc8adu]

looks like dirt embedded by your hsf.
bet those sanyo's are ok.

[Mod]

looks like dirt embedded by your hsf.
bet those sanyo's are ok.

That HSF was just my quick "test" setup (and Athlon 800), so the water-block is on there normally. Any possible way I could test the caps with a standard multimeter to make sure they aren't completely dead? I know I need some special probe to check it properly but unfortunately a simple multimeter is all that is available for now.

[Mod]

Alright. Replaced the four 2200uf low ESR Nichicons with four ultra low ESR. Also replace two of the 3 suspected Sanyos (I would have done all of them, but there is no room w/ 12.5mm caps). End result was a marked stability increase. Now 2.6GHz is stable again, finally. Still power problems however. +5 rail is sagging all the way down to 4.4v(!). Blah :\

[ncdave]

5v comes straight from the power supply. Though it is hard to believe that a new 510w psu could be incapable of supplying enough current at 5v, that must be it. Try a different psu.

Better yet, if you know someone who has a DC "amp clamp" meter, then borrow it and measure the current drawn at 5v (red wires), 3.3v (orange wires), and 12v (yellow wires). (Example: simply gather all the red wires together where they come from the psu and clamp the meter around them to measure the summed current drawn on all the 5v wires.) That'll tell you what current capability your machine needs on each supply rail. Try to get a psu with at least 50% more rated capacity than the PC actually draws.

Wow, looking on eBay, I see that DC amp-clamps have come way down in price since I got mine a few years ago! Do a search for:
DC (amp,current,meter,ammeter) Clamp
in "all catagories"
Forget borrowing -- do yourself a favor and buy one!

Warning: read the fine print in the auction descriptions and look up the specs elsewhere before you bid! Don't get stuck with a meter that reads AD & DC voltage, but only AC current, like the Craftsman 82180 or the Extech 38387.

-Dave

[bushytails]

Check to make sure the ATX connector isn't getting warm... I've seen a few melted ATX plugs. A high-resistance connection to the motherboard would cause even a perfectly regulated supply to look low... test by sticking some meter probes in the back of the atx connector, and see if it also reads 4.4v (indicating a bad power supply), or something a lot closer to 5v.

--Randy

[Mod]

I appreciate the replies. Yes, I've had problems with burned ATX connectors before. In all cases the PSUs were bad. This time isn't the case however.

For now all I have is a multimeter, the voltages coming out of the psu were solid, and I have since replaced the PSU with a 600+ watt server psu. Same deal. Ended up getting an Abit NF7-S v2.0 with all Rubycon (woohoo!) caps. All is great now.. back up to 2.6GHz on the XP.. works like a charm. No idea what was up with that Epox 8RDA. Replaced ALL caps on it except a few of the sanyos. Maybe the +5 mosfets are going bad? Not sure what the deal is..

[bushytails]

There are no "+5 mosfets", at least not on any motherboard I've ever seen... the 5v is provided (and regulated) by the power supply. A low 5v line can only be a bad PSU or a high-resistance connection to the board.

Most of the time it's a bad ATX connector, although I guess it's possible that a very, very poorly designed and tested motherboard could use too thin of traces for the 5v supply.

Fixes I've heard are people soldering the psu directly to pins of the atx connector, or running a separate wire spliced onto the psu leads over to the filter caps on the input of the vcore regulator.

If you don't want to fix it, feel free to send it (and a suitable cpu for full-load testing) over my way. I could use another box.


I didn't notice how long ago you'd posted it... my mistake. saw the response today and thought it was current. (ugh, bad pun. bad randy! bad!)


--Randy

[ncdave]

saw the response today and thought it was current. (ugh, bad pun. bad randy! bad!)

Ouch!

[Mod]

There are no "+5 mosfets", at least not on any motherboard I've ever seen... the 5v is provided (and regulated) by the power supply. A low 5v line can only be a bad PSU or a high-resistance connection to the board.

Most of the time it's a bad ATX connector, although I guess it's possible that a very, very poorly designed and tested motherboard could use too thin of traces for the 5v supply.

Fixes I've heard are people soldering the psu directly to pins of the atx connector, or running a separate wire spliced onto the psu leads over to the filter caps on the input of the vcore regulator.

If you don't want to fix it, feel free to send it (and a suitable cpu for full-load testing) over my way. I could use another box.


I didn't notice how long ago you'd posted it... my mistake. saw the response today and thought it was current. (ugh, bad pun. bad randy! bad!)


--Randy

I did test the middle pin on 3 of the mosfets and it showed +4.4v (same as my +5 rail in the bios). I did a quick mod (I did the same thing back in the day on a KT7-a that had a 2.2v Tbird 1.0 @ 1.7GHz, WAY more than the board could handle) that involved running the PSUs +5 directly to the mosfets middle pin. It did raise it 0.1v per mosfet, so it got it up to 4.7v total. Trying to get 2.6GHz on it was better, a little more stable, but I gave up on that mod a while back.. as I didn't want the chance of a wire breaking off of that middle mosfet pin and going loose with a live +5 line dragging all over the place.

Not sure what to make of it.. the board still works fine, tested it just the other day. Works like a charm, just can't handle the extremely high overclocks anymore (due to the +5 sagging way down I think). Guess since it has all Nichicon/Sanyo caps now and doesn't appear to be getting worse.. maybe throw it in my server.

The thing is.. even with low wattage cpus, the +5v is 4.4v. On every other board I've seen, once I start pushing a cpu up over 120 watts or so the +5v rail will sag. On the newer boards with the +12v boost, the cpu gets it power from the +12 on those (like my NF7-s), the +12v will sag... but it isn't much. Then, if you kick the voltage/cpu clock back to reduce the CPU wattage the +5 will go back up to where it should be. Not on this 8RDA however. I know so little about mosfets, transformers, etc I don't know what to make of it. Any ideas?

[ncdave]

When you measure from the red wire to a black wire on a drive connector, with a DVM, does it measure 5.0v? Or does it measure 4.4v (or 4.7v with your mod)? There's no way it should drop 0.3v across the wire between PSU and the motherboard!

[Mod]

When you measure from the red wire to a black wire on a drive connector, with a DVM, does it measure 5.0v? Or does it measure 4.4v (or 4.7v with your mod)? There's no way it should drop 0.3v across the wire between PSU and the motherboard!

All 3 psus show stable voltages on all lines (+5 = 5.00, +12 = 12.00, plus or minus a little bit). During operation the +5 molex connector line (that is open, unused) will sag a little bit, +4.7-4.8 or so with a 2.4-2.5GHz Athlon XP at 2 volts.. but upon testing the middle mosfet pin & ground will show around 4.4v, which is also what the bios shows on the +5. When I connect the 3 mosfets directly to the +5 line on the psu then I get a 0.3v boost on the +5. Seems like maybe there is something being lost between the ATX connector (supplying the power fine) and the mosfet itself.. I can always get a high resolution picture of that area of the board if you guys would like.

[ncdave]

All 3 psus show stable voltages on all lines (+5 = 5.00, +12 = 12.00, plus or minus a little bit). During operation the +5 molex connector line (that is open, unused) will sag a little bit, +4.7-4.8 or so...

By the molex connector, do you mean an unused drive connector, coming from the PSU? If so, then where are you measuring the stable 5.0?

BTW, IMO 5.00v is a bit low for the 5v line from the PSU. I like to see about 5.10-5.15v, to allow for a bit of voltage to be dropped en route to the components. The ATX spec allows for +/- 5%, so 4.75-5.25v is within spec, but being toward the high end of that range is best.

-Dave

[Mod]

All 3 psus show stable voltages on all lines (+5 = 5.00, +12 = 12.00, plus or minus a little bit). During operation the +5 molex connector line (that is open, unused) will sag a little bit, +4.7-4.8 or so...

By the molex connector, do you mean an unused drive connector, coming from the PSU? If so, then where are you measuring the stable 5.0?

Yes, measured the voltage in the 4 pin molex connector (type that you plug into hard drives), and also measured the +5 coming out of the ATX connector itself with and without it connected to the motherboard. To test it non-connected you can turn the ATX PSU on by shorting the green wire with any ground wire.

[ncdave]

If, measured from an unused drive connector, 5v is sagging down to less than 4.75, then your PSU is inadaquate. You need a PSU with more moxie on the 5v output. Look specifically at the current rating for the 5v rail (rather than the wattage rating of the PSU), and get a PSU that is rated for higher current output at 5v.

[Mod]

If, measured from an unused drive connector, 5v is sagging down to less than 4.75, then your PSU is inadaquate. You need a PSU with more moxie on the 5v output. Look specifically at the current rating for the 5v rail (rather than the wattage rating of the PSU), and get a PSU that is rated for higher current output at 5v.

I was thinking the same, as this has happened to me in the past with the KT7a, however the PSU I have now (and was using at the time) is rated at up to 61 amps on the +5 (most psus are 30-40) and 28 amps on the +12. I tried another PSU (Allied Truepower 450w) that had 50 amps on the +5 and 18 amps on the +12. Doesn't explain why though even with the CPU not sucking a lot of power (clocked it down to 500MHz, 1.3v or so) and the +5 was still low on the board. Usually you back off the wattage, +5 will go back to normal.. in this case it stays low, constantly. Weird stuff..

[ncdave]

61A! Wow! That has GOT to be more than plenty.

I'm stumped. The only thing I can think of to suggest is to get
yourself a DC amp-clamp and find out what it is really drawing.