D865GBF with oddball KZG

[yyonline]

So, I have an Intel D865GBF here. It has 5 KZG capacitors I want to replace. However, they are an odd size. I came up with a few scenarios for replacing them, I am curious to get other's opinions on them.

The KZG in question are 6.3V 820uF 8mm x 23mm. Now, the KZG data sheet shows nothing in that size, however I can infer the following based on can size (8x20 is the closest in the specs)

ESR: slightly below 0.021
Ripple: slightly above 1870

CPU VRM
VMR input caps are Nichicon HD, I will leave those alone.
VMR output caps:
7x Chemicon PSA polymer 2.5V 680uF
3x Chemicon KZG 6.3V 820uF 8mm x 23mm

So, here I want replace the KZG. I can see a few options.

Option 1 (I think this is best):
Replace KZG with 2.5V 680uF PSA. This will make all the VRM caps 680uF PSA.

Option 2: Stay with wet caps, increase capacitance to get the needed ESR and ripple. Replace with 6.3V 1500uF MCZ or 10V 1000uF MCZ.

Option 3: Keep capacitance the same, go polymer (820uF polymer)

Option 4: Keep capacitance the same, stay with wet caps (6.3V 820uF MCZ). This may have insufficient ripple and ESR, as the KZG are larger than usual. I don't think this would be wise, but thought I'd throw it out there anyways.

Thoughts on my VRM options?

Next area - there are two of these caps near the memory slot. One measures 2.1V, and the other measures 5V when the PC is powered on. I can't reduce voltage here at least on the one, I could reduce voltage on the other.

What I think I am going to do here is replace w/ 6.3V 820 uF polymer - Nichicon LG perhaps, as UCC doesn't make anything in 8mm for those specs. I think I would want to keep the capacitance the same, as it's not on the CPU VRM. However, would a 3-fold increase in ripple current cause any issues? I know "more ripple is better" but I don't know to what extent you can take that.

Other options - reduce capacitance with poly. Increase capacitance with wet caps to get the needed ESR and ripple. Pretty much the same options as outlined above, but I figure what's best for one area might not be the best for the other.

Thoughts on the two near the memory slot?

As always, I appreciate any knowledge you can share!

[linuxguru]

Re: D865GBF with oddball KZG

Option 5: Leave the KZG alone. Those 7 PSA polymers are enough to shunt ~35A of ripple current, which is more than enough for almost any Socket 478 CPU. The KZGs are only going to see a small fraction of the ripple current because they're higher impedance than the PSAs.

[yyonline]

Re: D865GBF with oddball KZG

2 out of the 3 KZG on the VRM are bulging, so I don't think leaving them alone is a good option. The CPU in this case is a 3.2Ghz Prescott core P4, which is just about the hottest thing ever. Even if they got a small portion of the current, I suspect the heat helped kill them. If nothing else, this PC would make an excellent space heater.

The KZG near the memory slots are okay, so I could leave them, but I'd rather only have to pull the board once.

[c_hegge]

Re: D865GBF with oddball KZG

Go for option 1 (using 680uF polies to replace them). Option 4 is probably the next best thing (820uF 6.3v MBZ or MCZ), since the polies on the VRM are taking most of the load

[yyonline]

Re: D865GBF with oddball KZG

VRM caps are probably the easy one to figure out. The non-VRM caps are the ones that have me a bit confused... I've searched, and read mixed information about reducing capacitance of non-VRM caps when switching to poly. If anybody has thoughts on this, it'd be appreciated.

[linuxguru]

Re: D865GBF with oddball KZG

If they're bulging, then Option 1 (replace with polymers) is the best.

[c_hegge]

Re: D865GBF with oddball KZG

I believe you reduce the capacitance for VRM only, but keep it the same elsewhere.

[yyonline]

Re: D865GBF with oddball KZG

Thanks for the tips so far. I've ordered some capacitors from digikey (since Topcat doesn't carry some of the sizes/values). I will report back in a week or so with the results once I've had a chance to solder them on and test. If it doesn't work out, I can always swap them again for something else.

[yyonline]

Re: D865GBF with oddball KZG

Well new caps arrived. Got them installed yesterday. The PSA were a pain to solder in on the VRM... They have thick leads, and the board has some substantial ground planes. Had to turn up the heat, but I got them in ok.

I ended up putting 2.5V 680uF PSA on the VRM and 6.3V 820uF Nichicon LG polymer for the two near the memory slot.

I am testing now, but getting some really weird results. Memtest86 runs for several hours without errors. After about 5 hours, it will crash with several million errors, always in memory slot 0. I've swapped memory, power supply, and cpu, but nothing helps.

So... I haven't seen memtest fail in such a way before. One or two errors for bad ram, sure. But several million from ram that tests okay in another machine? I have a few thoughts...

I'm assuming the PSA is an okay change on the VRM...seems like it's within reasonable limits given what others have done with poly mods.

Perhaps the Nichicon LG polymers (not on CPU VRM) were too big a change from the KZG that were there before? Like I said, I don't know if tripling the ripple current could cause issues.

Maybe everything is okay, but something in intel's bios is screwing with Memtest. I've had this happen before with intel's implementation of legacy USB support, but I'm using PS2 keyboard in this case. Perhaps it's something else similar?

Something else on the board has failed that's not capacitor related?

Anybody have any thoughts on this?

[c_hegge]

Re: D865GBF with oddball KZG

Tripling the ripple current won't make a difference, since it is a limit of the cap, like the voltage. I'm not too sure what is causing the problems though. If it were my board, I would probably see what happens if you use Rubycon MBZ electrolytics (just around the RAM slots, leaving the PSA for the CPU in place). Maybe it doesn't like something else about the poly mod.

[yyonline]

Re: D865GBF with oddball KZG

Well as a proof-of-concept test, I put the original non-failed KZG (tested with ESR meter) back in place for the 820uF caps near the memory slots. Testing now. If the polymers turn out to be at fault, I'll have to look for different replacements or just leave the KZG in place and hope for the best...

[c_hegge]

Re: D865GBF with oddball KZG

Replace them with MBZ as KZG are failure-prone, they just haven't died yet since they weren't under as much stress as the ones in the CPU VRM.

[yyonline]

Re: D865GBF with oddball KZG

I can't find an 820uF Rubycon w/ sufficient ripple... the original caps were larger than the standard 820uF cap, thus lower ESR and higher ripple due to the larger size. I could replace with the smaller 820uF capacitors, but I don't know if they'd be underrated for the circuit they're on.

If it tests okay... I may try 16V 1000uF MCZ which is 8x20, and in the right range for ripple current, though capacitance is about 20% more.

I don't know which would be the better option. I can decrease ripple, or increase capacitance. Grr to custom size caps making things complicated.

[c_hegge]

Re: D865GBF with oddball KZG

I would increase the capacitance and go for the 1000uF MCZ

[yyonline]

Re: D865GBF with oddball KZG

Well, with the original 820uF caps in place, memtest crashed after 14 hours. I wish I had a good working D865GBF to test with and compare results... like I said, I've seen Intel boards that fail memtest before, but never crash in operation. Oh well, I'll probably toss the board into the parts bin for now.

[yyonline]

Re: D865GBF with oddball KZG

Update. I had some time to kill today, so I pulled the board back out of the closet to take another look at it. It works!

Turns out it was a memory issue after all.

I was using 2.6V Kingston memory in a board that only has 2.5V for memory. Apparently 2.5V isn't quite enough to keep the Kingston stable.

Soo...I swapped some 2.5V Corsair. Still failed...and this was known good memory. However, the computer I tested it in to give it its known good status has an Abit board in it, which has default bios settings of 2.65V for DDR. Turns out if I lower the voltage on the abit to 2.5, the Corsair crashes in there too.

So, I borrowed some ram from another PC in the shop. No-name generic PC3200, threw that in the re-capped Intel board, and 48 hours of memtest later, no errors.

So, I will be sending the Corsair RAM back for an RMA.

Also, by happy accident, I found some new old stock of 820uF 8x23mm Rubycon MBZ to replace the KZGs with.

Another board saved from the landfill.

[quiggs]

Re: D865GBF with oddball KZG

I'm about to attempt this exact thing (= replace the 820 uF's on an Intel D865GBF). So what turns out to be the best replacement?

[ratdude747]

Re: D865GBF with oddball KZG

before it was discontinued, rubycon MBZ

where do you live? (we have wordwide members around here...)

[yyonline]

Re: D865GBF with oddball KZG

I'm about to attempt this exact thing (= replace the 820 uF's on an Intel D865GBF). So what turns out to be the best replacement?

Which 820uF capacitors? Are they the 23mm tall ones? If so, they're a custom size, and have lower ESR than the standard size ones.

The ones near the CPU, use these:

https://www.badcaps.net/store/produc...roducts_id=173

or

https://www.badcaps.net/store/produc...products_id=77

The HNs up the capacitance a bit, but keep the ESR near the original specs. This is not a problem in a CPU VRM circuit. The polymers keep the capacitance the same, but lower the ESR. This is also not a problem in CPU VRM.

The ones near the RAM slots, use these:
http://search.digikey.com/us/en/prod...3-1-ND/2347911

We have to go polymer here, as the originals are larger than standard, and we don't know if it's safe to change the capacitance here. Equivalent capacitors worked fine on my board, once I got good memory for it. In fact, the board just came back into the show a few days ago as its owner replaced it with a newer machine.

[quiggs]

Re: D865GBF with oddball KZG

Denver, CO, USA.
Sorry -- the 820's are 20mm (three in the row by the CPU, and two at the ends of the DIMM slot). BTW, i also need to replace the 2200uF 10V (10mmx25mm) near the SATA jack.