Intel D945GCLF Bloated Samxon GC due to overheating

[linuxguru]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

... I'd say Taicon is worlds ahead in quality excepting their HD and HI series which seem to be notorious for failing on motherboards every which way...

Taicon is an Taiwanese associate/subsidiary of Nichicon, Japan. Their series are generally a subset (and similarly named) of Nichicon's. I haven't actually seen a failed Taicon cap yet, including the HD, though it's entirely possible that they fail in hot and high ripple-current locations like VRM outputs (where MCZs, KZGs, KZJs, etc. also die like flies). I have also used 16V-rated Taicon HD in SMPS recaps without a single failure in about 9 years - they seem to be similar in size and specs to NCC KZE, another series in which I haven't encountered a single failure.

[Wester547]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

Taicon is an Taiwanese associate/subsidiary of Nichicon, Japan. Their series are generally a subset (and similarly named) of Nichicon's. I haven't actually seen a failed Taicon cap yet, including the HD, though it's entirely possible that they fail in hot and high ripple-current locations like VRM outputs (where MCZs, KZGs, KZJs, etc. also die like flies). I have also used 16V-rated Taicon HD in SMPS recaps without a single failure in about 9 years - they seem to be similar in size and specs to NCC KZE, another series in which I haven't encountered a single failure.

There are a few threads on this forum that detail many Taicon HD (and HI) related failures on motherboards. Heat is a possible cause for failure. KZG and KZJ fail on the shelf more than we'd like to see due to a defect in the electrolyte, and MCZ seems to be too sensitive to heat (but understandably since the lower ESR an electrolytic is, the more aqueous it is). One member (UraBahn I believe) also observed a failed 1200uF 16V Taicon HD in a 400W Lite-on, a long time ago, but that could have been a heat related failure as well (which is possible - I believe that particular model had a bottom mounted fan and a forest of wires covering the output capacitors) if Lite-on power supplies really do run that hot on the secondary side (though they seem well built to me and I've never seen discoloration from heat in them).

9 years of daily use or 24/7 use is impressive, but it would be understandable if the power supply is cooled and ventilated well (even crap brands last a long time in power supplies that run very cool). The fact that they survive on the +12V output is also understandable. I believe you correctly pointed out a while back that the higher voltage retards or at least prolongs the electrolytic dissolution (the failed Taicons on those motherboards were in the VRM output). The other series from Taicon seem to be fine and only seem to fail under discolorating levels of heat. I'd guess they're about the same in quality as the good series from Samxon.

[linuxguru]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

So I can take a 220uF 16v 6mm ZLH Series as replacement for my 6,3V, 220uF. Right?

*Bump* - Yup, you can use a ZLH. I used a 120uF/16V 6mm ZLH at C1UB and C10UB (both originally Samxon GD 220uF/6.3V/6 mm).

Other replacements (for future reference):

C86BV, C162BV: Samxon GC 1000uF/10V replaced with Rubycon ZL 1500uF/6.3V/10mm
C47BV: Rubycon MCZ 820uF/6.3V/8mm replaced with Panasonic FJ 1500uF/6.3V/8mm
C20BV: Samxon GD 3300uF/6.3V/10mm replaced with Nichicon HM 3300uF/6.3V/10mm

Everything else (mostly 470uF/10V Nichicon VR and 100uF/25V Samxon GS) is stock - though it might not hurt to replace C110BV and C46BU next to the Southbridge and PCI slot also. I didn't have 470uF in 6mm diameter, so I left them alone.

Replacing a Rubycon MCZ at C47BV with a Panasonic FJ might look pointless, but MCZ is rated only for 2000h and they often fail in high-heat, high-ripple locations. FJ is far more durable in such applications, and I got double the capacitance in the same 8mm diameter. We'll sign how it goes, but I anticipate no problems. Also available in 8mm are Panasonic FJS, FL and Rubycon ZLH.

I ran Memtest86 on it to check stability, and it's rock stable and everything is cool as a cucumber. BIOS HW monitor reports CPU temperature as 47c, which is fine. This is a great platform that has been ruined by bean-counters looking to save a few cents - it's an EM64T Atom that destroys everything else (except maybe a Dothan Pentium M) on single-threaded performance at the same clock.

[stj]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

why not just 100% polymod the whole board? there arent many caps to change.

[linuxguru]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

Some of the smaller electrolytics appear to be bypasses for linear regulators - there's a risk of instability if the ESR is too low at those locations. For those that are used in switching converters, polymers would be a good choice - but the values have to be determined by trial and error, e.g. is 820uF/6.3 sufficient to replace the 3300uF/6.3, 1000uF/10, 820/6.3 electrolytics, and so on? Since I had only one board to play with, I didn't want to risk too much by trial and error.

(I also forgot to mention that C20BV, the 3300/6.3 at the corner of the board is a *bear* to replace. It's sitting on one or more copper planes, with a thick copper inductor nearby conducting heat from the planes. Add to that the higher melting point of Pb-free solder, and it's pretty hard to get the cap out, clear the through holes of solder without damaging them, and install the new capacitor without shorting the pad(s) to one of the ground planes - the isolate ring between the pads and the ground plane seems to be only 10 mil or so, so it's rather easy to create bridged shorts to ground. If it doesn't look bulged, it may be prudent to leave it alone. The only reason I replaced it, is because it was a Samxon.)

[stj]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

i intend to put some poly's on my one soon,

i was going to replace the 3300uf with this:
http://uk.farnell.com/wurth-elektron...ial/dp/2466521

[Per Hansson]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

The Samxon ULR polymer caps I used in my original post here have hold up well for over 4 years now.
I never replaced the original Samxon GD 3300uF cap or the Rubycon MCZ 820uF cap next to it by the memory and it's still working fine like this over 7 years later.
System is still in use 24/7 as my firewall with only a tiny silent fan on the chipset heatsink.

The fan I have had to replace though, those small spinners are not very reliable...
Now I've been using a Noctua NF-A4x10 FLX 40mm fan for a little over 2 years and it started making noises.
It has a 6 year warranty though so I should be ok, right now I just flipped it over which made it silent.
I've learnt the hard way that sending in things for RMA that are not 100% dead is useless

[linuxguru]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

i was going to replace the 3300uf with this:
http://uk.farnell.com/wurth-elektron...ial/dp/2466521

Thanks for the link - I had no idea that such high capacitances existed in polymer. It's moderately expensive, but probably well worth the money for certain applications, like Vcore/Vdimm bypass.

I left the single polymer next to the CPU alone, since it probably doesn't see much ripple from the 2W Atom. However, I might investigate whether it's worth populating some of the empty locations next to it in a future recap.

[linuxguru]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

The fan I have had to replace though, those small spinners are not very reliable...
Now I've been using a Noctua NF-A4x10 FLX 40mm fan for a little over 2 years and it started making noises.

Yup, it runs at ~4500 rpm so the bearing probably wears out quicker. Mine wasn't used 24/7, so the original fan is still intact, but I'd probably go with an NMB or similar ball-bearing fan when I replace it.

[Per Hansson]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

Replacing the original fan was one of the first things I did to my board, along with the heatsink.
So it's now running with it's third fan
I've always run the fans with a Zalman power adapter at such a low RPM that I can't hear it...

[Wester547]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

Replacing a Rubycon MCZ at C47BV with a Panasonic FJ might look pointless, but MCZ is rated only for 2000h and they often fail in high-heat, high-ripple locations. FJ is far more durable in such applications, and I got double the capacitance in the same 8mm diameter. We'll sign how it goes, but I anticipate no problems. Also available in 8mm are Panasonic FJS, FL and Rubycon ZLH.

FJ is rated up to 3,000 hours but 12.5mm and shorter FJs are only rated for 2,000 hours. Suncon/Sanyo WG is rated up to 4,000 hours on the other hand (2,000-4,000 hours depending on case size) but doesn't seem to fare better than MBZ (still quite good, even being rated for 2,000 hours @ 105*C only). PCBONEZ once said that due to differences in the chemistry that it's not fair to compare capacitors of different grades and I agree with that much. FL I'd only compare to MCZ and GC, FJ/FJS/FJZ I'd only compare to MBZ and GD, etc... FL seems less likey to bulge and leak than MCZ, which denotes that MCZ may be slightly unstable (IE somewhat low heat tolerance, but nowhere near as bad as KZJs or early HNs).

Cases like this make me nervous about using ultra-low ESR Samxon electrolytics, especially given the fact that their entry level low ESR electrolytics are very unreliable. I suppose it's possible that their QC testing isn't the best, which is why good batches of Samxon and bad batches of Samxon exist.

[stj]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

the intel fan has a lifetime warranty - just so you know.

[mockingbird]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

Replacing a Rubycon MCZ at C47BV with a Panasonic FJ might look pointless, but MCZ is rated only for 2000h and they often fail in high-heat, high-ripple locations. FJ is far more durable in such applications, and I got double the capacitance in the same 8mm diameter. We'll sign how it goes, but I anticipate no problems. Also available in 8mm are Panasonic FJS, FL and Rubycon ZLH.

You're totally right about MCZ not taking abuse well at all.

The electrolyte in very old Panasonic FJ (like pre-2005) starts to congeal towards the bottom. I might post pics of this some day. They still might work perfectly fine, but this is just something to consider.

[Wester547]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

You're totally right about MCZ not taking abuse well at all.

The electrolyte in very old Panasonic FJ (like pre-2005) starts to congeal towards the bottom. I might post pics of this some day. They still might work perfectly fine, but this is just something to consider.

Are you saying that they leak from the bung? Panasonic capacitors possibly have softer bungs than the rest of the good brands (at least after the early to mid 90s). When they leak, it almost always seems to be from the negative lead.

[stj]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

i'v only seen FC series leak.

[mockingbird]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

Are you saying that they leak from the bung? Panasonic capacitors possibly have softer bungs than the rest of the good brands (at least after the early to mid 90s). When they leak, it almost always seems to be from the negative lead.

No. I'm saying that when you open them up, towards the bottom of the rolled up sheets, you see dark-brown congealed electrolyte. They might still be OK, I don't have an ESR meter to test them with.

I've used Rubycon YXG with 2000 datecodes though, and they were perfectly fine. Maybe I should cut one open and see how it looks inside.

[Wester547]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

No. I'm saying that when you open them up, towards the bottom of the rolled up sheets, you see dark-brown congealed electrolyte. They might still be OK, I don't have an ESR meter to test them with.

Sounds like they're dried up inside. Maybe they aren't sealed all that well or had a rough life. Also, capacitors can still measure in spec when the actual leakage current is high. So I would guess that those FJs are leaky if the electrolyte has darkened and congealed (but they may be partially open as well).

I've used Rubycon YXG with 2000 datecodes though, and they were perfectly fine. Maybe I should cut one open and see how it looks inside.

Well, if you want to autopsy it. But that would be a destructive process. Maybe it's just safer to leave it alone if it's still working. Then again, it couldn't hurt to supersede them with NOS capacitors since they're 15 years old.

I agree that MCZs (and MFZs) don't handle abuse well. Still, they're nowhere near as bad as KZG/KZJ, as those like to bloat in storage (which speaks volumes about their quality).

[stj]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

did anybody ever find a spec sheet for MFZ's?

[Wester547]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

https://www.badcaps.net/forum/showpo...20&postcount=1

No datasheet exists for MFZ. MFZ was a custom order discontinued in 2010 along with MCZ (MBZ was discontinued the year before). Presumably, they (MFZ) are close to MCZ in specs, possibly equal to HZ and GA going by this (see page 8).

[linuxguru]

Re: Intel D945GCLF Bloated Samxon GC due to overheating

I should add that the 820uF/6.3V MCZ pulled from the near the DIMM slot looks fake - the bottom bung is the one with concentric circles, which I've never seen on an authentic Rubycon on ammo tape or bag before; and the ESR measures about 40 mohm, which seems to be a bit high, but that's after 6 years of intermittent use. OTOH, the 3300uF/6.3V Samxon GD seems to be within spec at 10 mohm.

Did Intel China knowingly use fake Rubycons, or did they also get conned by a flaky vendor? Or alternatively, did Rubycon outsource production of (some) MCZs to a volume OEM producer in China?