D865GBF with oddball KZG

[quiggs]

Re: D865GBF with oddball KZG

yyonline - just to be clear, you say I need different replacements for the ones near the CPU vs the ones near the ram, even though the originals are all identical?

[yyonline]

Re: D865GBF with oddball KZG

BTW, i also need to replace the 2200uF 10V (10mmx25mm) near the SATA jack.

What brand/series of capacitor is that? My photographic memory is usually pretty good, but that one wasn't bad on my board, so I don't recall what it was. It's possible my board uses a different brand of capacitor there. If it's another KZG, you can use this:
https://www.badcaps.net/store/produc...roducts_id=137

[yyonline]

Re: D865GBF with oddball KZG

yyonline - just to be clear, you say I need different replacements for the ones near the CPU vs the ones near the ram, even though the originals are all identical?

Yes. They're on different circuits, so the applications are different. Exact replacements aren't available, so we have to pick the "next best" thing. Given that the circuits are different, the next best things will be different for the two applications.

Take CPU VRM. The original caps are rated for 6.3V. One of the proposed replacements is a 2.5V cap. P4 CPUs run at "about" 1.5V, and that's the voltage on that circuit. Therefore, it's safe to reduce the voltage rating in this case. The 2nd proposed replacement changes the capacitance. This is low risk, as having extra capacitance in the CPU VRM circuit isn't an issue.

Now, let's take the ones near the ram. If my memory is correct, these were at DDR voltage, which is 2.5V. Therefore, the 2.5V rated capacitors that I suggested for the CPU VRM are inadequate for use here (barely adequate, but it's good to have a safety margin). Therefore, we can't reduce the voltage. It's not a VRM circuit, so I don't think changing the capacitance is a good idea here. Keeping the capacitance the same, the only things that meet the specs of the original capacitor (exceed in this case) are polymer capacitors. Polymer capacitors are smaller, but have superior specs and are constructed out of different materials.

I hope this makes sense - I'm pretty tired so my explaining skills might not be what they usually are :-)

[PCBONEZ]

Re: D865GBF with oddball KZG

You want the total uF in Vcore to be ~5500uF or more.
[That's based on what is found on production all-poly boards.]
Only cheap POS boards like AsRock [or boards for very low power CPU's] use less.
All poly server boards tend to have 6000-7500uF in Vcore - for each CPU.
.
10x 680uF Poly would do fantastic.
Just make sure their [individual] ESR is less than the original KZGs and ESR will be fine..
.
The total uF is to keep voltage in spec during transients like CPU from 10% to 100%.
If you don't have enough and you go to a heavy load suddenly you can cause BSOD or reboot.
.

[quiggs]

Re: D865GBF with oddball KZG

Sorry guys -- I missed your responses last night because I didn't notice that this thread had grown a second page! (You know, the internet continues to astound me: total strangers from across the country are helping me with an obscure problem in the middle of the night. I really appreciate it.)

1) The 2200uF 10V is a Nichicon HM(M) which is silver & black (vs. the copper & black HN(M)'s near the CPU).

2) Other than cost difference (which is trivial in light of the amount of work I'm about to do), is there any reason not to use polymers everywhere I can (even if an exact replacement lytic is available)? And was there originally some technical (not financial) reason that Intel mixed 3 lytics in with 7 polys in the VRM?

[yyonline]

Re: D865GBF with oddball KZG

1) The 2200uF 10V is a Nichicon HM(M) which is silver & black (vs. the copper & black HN(M)'s near the CPU).

HM and KZG have identical specifications, so the suggested replacement will work fine.

2) Other than cost difference (which is trivial in light of the amount of work I'm about to do), is there any reason not to use polymers everywhere I can (even if an exact replacement lytic is available)? And was there originally some technical (not financial) reason that Intel mixed 3 lytics in with 7 polys in the VRM?

Cost, most likely. A few cents savings over millions of boards adds up to a lot. I've never had any issues changing the CPU VRM capacitors to polymer.

Elsewhere you have to be a bit careful differing too much from the original specifications. It might work, it might not. Changing this around too much gets into experimentation territory. Furthermore, polymer capacitors tend not to be available in capacitance above 1500uF. There are a few, but not many series that go higher.

[quiggs]

Re: D865GBF with oddball KZG

OK, then, I think I'm good to go. Thanks again for all your help!

[dpdemay]

Re: D865GBF with oddball KZG

Got 2 of these D865GBF boards that I'm going to attempt to revive ~ found some Rubycon MBZ 820uf capacitors on eBay. One of the boards has 5 bulging Nichicon HN(M), the other has 3 bulging 820uf KZG, the two near the RAM haven't swollen, but I might as well swap those out as well. Anything wrong with using the Rubicon caps in all 5 locations, or should the ones near the RAM be the ones from Digikey mentioned above?

[c_hegge]

Re: D865GBF with oddball KZG

Ruby MBZ should be fine for all of those locations, but I wouldn't encourage using fleabay for them, as most on there are fakes. Just get yourself some newer Hichicon HNs from here (https://www.badcaps.net/store/produc...roducts_id=163).

[dpdemay]

Re: D865GBF with oddball KZG

With any luck, the Rubycons that are scheduled to arrive on Monday won't be bogus. The eBay seller has 11K+ Positive Feedbacks, and sells scads of capacitors, so I'll give 'em a shot. The D865GBFs are obviously old boards at this point, so if my attempts at repairing them fail, it's no big deal. They came out of a pair of machines I built for a friend back when the D865GBF & D865GLC boards were relatively new. I've since upgraded just about everyone else's machine that I built with those boards back then, and I'll likely be replacing this pair soon, even if the Rubycons fix works. They've been doing a fair job of running Windows 7, but it's really time to retire 'em...