Intel D850EMV2 Recap project

[Digital Technophile]

Okay I'm getting ready to get back on the RAMBUS. Been putting off taking pictures, but today I finally got the shots I need, a list of caps, and identified a few interesting features between the two hardware revisions I have in my hand.

I first acquired this heap back in 2004 as an upgrade to the Dell Optiplex GX1 I'd been using. After using it for a few years it began having boot issues (~2006). Finally, after it refused to boot up, I sourced an identical (or so I thought) board on fleabay. Tossed the old board back into the box just in case for some reason it might come in handy for something in the future and put it on the shelf. Life was good and in the meantime I also upgraded to a socket 939 Hp a1253w PC and the Intel became a server after I began studying IT as a "hobby" (who does that anyway?). Even though Intel didn't officially support Server 2003 OS on this board, I found it lived there quite happily with XP drivers. Finally, I got an Intel SL6S5 3.06gHz the highest CPU the board would support. From that point I was always greeted by the BIOS at boottime alerting me to the fact that PC800 RDRAM was too slow for this processor and would be defaulted to 400mHz FSB. No biggie (even though RDRAM was still pricey at $220+ for PC1066), still runs okay and now I have an MSI MS-6702E based PC doing DC/DNS duty and the Intel PC was used as Suse Linux/Apache web server for website testing. Finally, after having joined badcaps.net, I began inspecting every piece of electronica I owned looking at all the pretty caps. One day the original board is dug up, and low and behold what do I find but a couple of bloater on the vcore out side (a pair of Nichicon HM series 6.3v 3300uf caps)! Hmm, could this have been the cause of my troubles a few years back???

That's where things have become "interesting". It seems that there have been a few revisions in this board's production. The original board is an AA C18361 revision board which DOES support the SL6S5 3.06gHz CPU. However, the "replacement" board I found is an AA A56420 revision, which is a WAY earlier board revision which supports up to a 2.60gHz. Pictures of both boards and the list of condensitors will be up shortly.

[Digital Technophile]

Re: Intel D850EMV2 Recap project

I tried to get a good set of pictures and here's what I came up with, starting with the AA A56420 Revision Board:

[Digital Technophile]

Re: Intel D850EMV2 Recap project

And this is the AA C18361 revision board, which is the one with the bad caps, and is also the one I'd like to start recapping first:

[Digital Technophile]

Re: Intel D850EMV2 Recap project

....and the list of caps (originals on board in parenthesis):

• 3x 25v 220uf (United Chemicon SME)
• 12x 25v 100uf (United Chemicon SMG)
• 17x 25v 22uf (Nichicon VR)
• 3x 16v 2200uf (United Chemicon LXZ)
• 3x 16v 470uf (Nichicon VR)
• 1x 10v 1500uf (United Chemicon KZE)
• 2x 10v 470uf (Nichicon VR)
• 2x 6.3v 3300uf (Nichicon HM) - These are visibly damaged
• 2x 6.3v 1500uf (United Chemicon KZE)

I would like to order all the ones I can from here (wink wink Topcat!), with the balance from Digi-Key or Mouser. I'm still a few weeks out from having the cash to purchase, but I'd like to have the list ready and get an idea of the cost.

And although the caps on this board look to be in good shape, I think it would be good form to replace them as a set. Not too picky about brand names, but would like to stay with Rubycon, Nichicon, United Chemicon, and Panasonic. I also don't expect to be able to have one unified brand name across the board, but quality counts!

I was hoping to also get opinions regarding the poly caps. I'm operating under the assumption, they probably don't need to be replaced, but should I? And I was curious, in this picture, I noticed that the caps (9v 560uf) have two different markings at the bottom: 2Z1 and 2Z3. What the difference between the two? Thanks to all for your help.

[c_hegge]

Re: Intel D850EMV2 Recap project

You can leave the KZEs, LXZs and polys in place. They are good caps and should be fine, but replace the others

3x 25v 220uf (United Chemicon SME) - https://www.badcaps.net/store/produc...roducts_id=100
12x 25v 100uf (United Chemicon SMG) - https://www.badcaps.net/store/produc...roducts_id=101
17x 25v 22uf (Nichicon VR) - https://www.badcaps.net/store/produc...roducts_id=102
3x 16v 2200uf (United Chemicon LXZ) - These can stay
3x 16v 470uf (Nichicon VR) - https://www.badcaps.net/store/produc...products_id=73
1x 10v 1500uf (United Chemicon KZE) - Can stay
2x 10v 470uf (Nichicon VR) - https://www.badcaps.net/store/produc...products_id=73
2x 6.3v 3300uf (Nichicon HM) - https://www.badcaps.net/store/produc...products_id=58
2x 6.3v 1500uf (United Chemicon KZE) - Can stay

I've suggested samxon because topcat has more sizes in them than rubycon and nichicon.

[Th3_uN1Qu3]

Re: Intel D850EMV2 Recap project

Samxon are decent caps but i'd still go for Rubycon in there... would look better.

[PCBONEZ]

Re: Intel D850EMV2 Recap project

The only ones I see that need to go are the HM and the VR.
HM were defective from '01 thru '04.
VR are 85C and they start going bad due to dry-out sometime after 4 or 5 years old depending on case temps and hours running.

[Th3_uN1Qu3]

Re: Intel D850EMV2 Recap project

Which makes it even more relevant to go for Rubycon.

[PCBONEZ]

Re: Intel D850EMV2 Recap project

Rubycon, Sanyo/Suncon, Panasonic, NIC Components
Same-Same

Most Nichicon and Chemicon are good.

Hitano & Illinios Capacitor are also good when you can find the low ESR series.
.

[mockingbird]

Re: Intel D850EMV2 Recap project

Hitano & Illinios Capacitor are also good when you can find the low ESR series.
.

Hitano, is Taiwanese. Are you sure they are a reliable brand?

[PCBONEZ]

Re: Intel D850EMV2 Recap project

My mistake, but no, I never had a problem with them.
.

[Digital Technophile]

Re: Intel D850EMV2 Recap project

Thank you all for your inputs, especially c_hegge for taking the time to post those links. Do you all really feel the Chemicons should stay? Considering the board is about 10 years old, how much longer will they be truly viable? I just feel like if I'm going to go through replacing the two blown Nichicon HMs and the 85° VRs (can't have much life left in them), why not replace them all?? I considered that the cheapest and most expedient thing would be to desolder the Rubycon MBZs from the A56420 rev. board and put them into the C18361 board. But I'm not in that big of a rush and I certainly feel the board is worth one good recap at the least. At the moment it's looking like 3 weeks until the budget will allow buying the caps.

That being said, I've never had experience with Samxon (doesn't mean they're not a nice cap), but for this board (the C18361 rev.) I want to be able to forget that it even has wet caps. The other A56420 I'm considering doing a recap and selling the board, 1.6gHz cpu, and 512mb PC800 memory as a package. So, I'd really like to stay with Rubycon, Chemicon, and/or Nichicon on the C18361. Tomorrow, I'm going to start getting a list together; within these three companies what series should I be looking for? I mean, I'm well capable of pouring through the data sheets. But are Rubycon MBZs really that different from MCZs for example? I keep reminding myself to remember to measure the can sizes as well. With those thoughts, I'm going to go pass out for the evening. It's been a long week. Thanks again for your help!

[PCBONEZ]

Re: Intel D850EMV2 Recap project

MCZ have 20% to 30% lower ESR depending on which size.
.

[PCBONEZ]

Re: Intel D850EMV2 Recap project

And the part I brain farted 'cause I got interrupted.
If changing the Chemicons makes you feel better then do so.
Probably not a bad idea if it has some hours on it and is gonn'a see 24/7 use.
Doubly so if your business depends on it.
.

[c_hegge]

Re: Intel D850EMV2 Recap project

Avoid chemicon. their motherboard series (KZG) is defective and always has been.

[lti]

Re: Intel D850EMV2 Recap project

There aren't any KZGs on that board.

I would replace the Nichicon HMs and the 85°C caps and leave the rest alone. I have a 12-year-old computer with crappy caps on the motherboard and in the power supply and it still runs fine.

[PCBONEZ]

Re: Intel D850EMV2 Recap project

Avoid chemicon. their motherboard series (KZG) is defective and always has been.

Both their KZG and KZJ have issues but avoiding their other series due to that is foolish.
Their KY series for instance is common in better PSUs and I've never even seen a bad one.
.

[Digital Technophile]

Re: Intel D850EMV2 Recap project

Spending time on the web ferreting out datasheets, suppliers, et cetera. Found most everything here, but I found some things I'd like clarification/advice on. Pretty much considered staying with wet caps, yet I've found a few non-standards.

As for the 25v 22uF caps, are these tantalum caps sensible here? There's a datasheet there, but no mention of ripple current specs. The lead spacing is the same, the only thing sticking out to me (read the datasheet) is that @85c it's rated for 25v BUT @125c it's only rated for 16v?! I know what you all are probably thinking "just stick with wet caps"; just curious on the possibility here. The only thing I see so far is a $35 dollar bill getting them into my mailbox! There also seem to be some poly caps around, but the physical dimensions rule them out (8mm dia. x 10.5mm VS 5mm dia. x 12mm stock size)

And the 16v 470uF caps there's a Nichicon LG poly cap with the correct physical dim.s, but the ripple current is rated @5.4A?! What would that mean on a mainboard? It seems to me that the ripples usually fall within a range (probably ~10x less than 5.4A?) and these caps just have a wider range of smoothing capability?

Finally, I seem to be steering toward Nichicon and it seems that in order of ripple and ESR that HZ > HN > HM < PS (like don't consider these)?

Okay back to work, going to Mouser to see what Chemicons available and how they compare to Nichicon. Might have to bust out spreadsheet so I can better compare; I've got time!

[PCBONEZ]

Re: Intel D850EMV2 Recap project

Tantalum caps don't handle Ripple very well.
Not a good idea to replace lytics with them w/o knowing exactly what the circuit does.
Also there are several kinds of Tantalum caps and the wet lytic ones have an extremely corrosive electrolyte.

Polymers handle butt-tons more ripple than comparable lytics.
The ripple spec doesn't have anything to do with circuit operation.
It tells you how much ripple the cap can handle without overheating internally.
[Much like a watts rating for a resistor.]
As a spec for a capacitor, there is no such thing as too much ripple.

PS are just fine when you are replacing caps with ESR in their range.
-
Choose replacement caps by the specs of the caps you are replacing...
Too much overkill on the ESR 'upgrade' can lead to unstable circuits and an unstable motherboard.
-
What you can get away with in a VRM Vcore is N/A on the the rest of the board.
- The Ripple in the VRM Vcore is created by the VRM MOSFETs.
- The Ripple in the rest of the board is created by the PSU.

- The 'general' mobo caps are in parallel with the PSU output caps.
- To ripple,,, ESR is just a resistor.
So you have:
PSU output 'resistor' ..
.. in parallel [through long wire and mobo traces] with a mobo 'resistor' that is near some IC chip.
....... They are sharing the ripple current to ground.
now you make mobo 'resistor' very small.
- This changes how the 'shares' of ripple are divvied up.
Shit-ton more ripple current goes through long wire and through mobo traces to the now smaller mobo 'resistor'.
[Depending on the layout this added ripple might be 'felt' by that IC chip.]
[There could also be some 'cross-talk' EMI effects in the traces to unrelated circuits.]
- Yeah you are removing more ripple overall, but you are exposing more circuity to ripple because of -where- you are removing it.
[IOW: You are encouraging more PSU ripple to go to and through the mobo.]

~~ Which is one of several reasons I recommend sticking close to 'stock' ESR outside of a VRM.
.

[Digital Technophile]

Re: Intel D850EMV2 Recap project

Tantalum caps don't handle Ripple very well.
Not a good idea to replace lytics with them w/o knowing exactly what the circuit does.
Also there are several kinds of Tantalum caps and the wet lytic ones have an extremely corrosive electrolyte.

Polymers handle butt-tons more ripple than comparable lytics.
The ripple spec doesn't have anything to do with circuit operation.
It tells you how much ripple the cap can handle without overheating internally.
[Much like a watts rating for a resistor.]
As a spec for a capacitor, there is no such thing as too much ripple.

PS are just fine when you are replacing caps with ESR in their range.
-
Choose replacement caps by the specs of the caps you are replacing...
Too much overkill on the ESR 'upgrade' can lead to unstable circuits and an unstable motherboard.
-
What you can get away with in a VRM Vcore is N/A on the the rest of the board.
- The Ripple in the VRM Vcore is created by the VRM MOSFETs.
- The Ripple in the rest of the board is created by the PSU.

- The 'general' mobo caps are in parallel with the PSU output caps.
- To ripple,,, ESR is just a resistor.
So you have:
PSU output 'resistor' ..
.. in parallel [through long wire and mobo traces] with a mobo 'resistor' that is near some IC chip.
....... They are sharing the ripple current to ground.
now you make mobo 'resistor' very small.
- This changes how the 'shares' of ripple are divvied up.
Shit-ton more ripple current goes through long wire and through mobo traces to the now smaller mobo 'resistor'.
[Depending on the layout this added ripple might be 'felt' by that IC chip.]
[There could also be some 'cross-talk' EMI effects in the traces to unrelated circuits.]
- Yeah you are removing more ripple overall, but you are exposing more circuity to ripple because of -where- you are removing it.
[IOW: You are encouraging more PSU ripple to go to and through the mobo.]

~~ Which is one of several reasons I recommend sticking close to 'stock' ESR outside of a VRM.
.

Ahhh, thanks for clearing up those finer points. I guess lack of large ripple current handling the reason why there aren't any specs in that datasheet. Besides that, the price for replacing 17 caps with them ($35), puts them way out of price range anyway...just curious.

The only poly caps I considered that physically fit were the 16v 470uF caps, see this pic which is between northbridge and agp slot and this reference pic the two between cpu and rimm socket just to left of the three polycaps. Should I ditch those??

As far as ESR and replacement of the Nichicon VRs, I'll look for a 105c(?) upgrade with a comparable ESR rating. Intel knows what they're doing, and I don't see a need to go nuts "upgrading" caps. There is a lot of caps to choose from; badcaps.net have most of what I'm looking for(thank god), just trying to find the 25v 100uF caps (12x Nichicon VR) and 25v 22uf caps (17x Chemicon SMG) in a 2000hr @ 105c. Thanks for the explanation on ESR, much better picture of what is happening in the "complete" circuit. And the three caps in picture 4 and picture 5 are the PSU output caps, yes?