HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

[momaka]

We have about 40 or 50 of these HP DC5000 SFF desktops at work and a big portion of them have developed bad caps. As with many other SFF “mini-oven” desktops, these PCs run quite warm so they eventually cook their caps and then start to give random intermittent problems (including BSODs) or simply refuse to boot at all. They use a grab-bag of caps. So far, I have seen Sanyo, United Chemicon, Fujitsu, Rubycon, Nichicon, Panasonic, Teapo, and OST for the larger caps, and G-Luxon, GSC, and OST for the smaller ones. While there are a lot of good caps on this list, the majority of these boards employ Teapo caps in hot areas so eventually they will all bulge. Also, some of the Nichicon caps are the HM series with a bad date code so those fail as well. And lastly, even Rubycon aren't a safe bet – I've seen the MCZ around the CPU regularly bulge and leak due to the heat.

The motherboard used in these PCs is the ASUS P4SD. Generally, I am not a fan of ASUS and AsRock motherboards. Given that these DC5000 PCs are Pentium 4 -based, I thought they may not be worth recapping when I first saw them. The P4SD, however, is actually engineered quite well IMO - almost every major rail on the motherboard is buck-regulated and there's very few linear / non-buck rails (unlike the cheap consumer-grade ASUS boards). In fact, the reason why many of our DC5000 PCs still work even with bulging capacitors is precisely because of the robust design of the P4SD. Out of those 40 or 50 PCs we have, only 7 were labeled as “dead” because they didn't want to boot or gave blue screens. The rest work okay and we have about 10 or 20 that have been in service for 2 years now – some even with bulging caps! This convinced me that maybe it wasn't such a bad idea to fix these computers. After all, they have 2.4 GHz or faster Pentium 4 CPUs with HT. They can even do 720p YouTube in fullscreen – that's more than enough power IMO for office work.

So, because the P4SD motherboard has a lot of buck-regulated rails, this allows the majority of the caps to be changed with polymers without causing issues. After fiddling with the motherboard for about half a day and determining all of the voltage rails, I was able to come up with a capacitor list consisting mostly of polymers.

The first of these (the "prototype") that I did actually used polymers almost exclusively. It has been over 5 weeks now that I have been using it as my work desktop computer, and it hasn't had any problems. However, I also tried a second, “tweaked” version of the polymod – mostly to bring down the cost per machine and the work I have to do since I have 30 that still need a recap. The second version also works well, but I haven't tested it as much as the first version (only ran a bunch of bench marks and brief stress tests on each machine). So far I have only done 8 of these PCs, 7 of which have been revived and work fine (1 still won't look for boot devices, but it may not be a cap-related problem).

Attached below (in order) are:
- a list of the original capacitors on the P4SD
- a “cap map” showing where all of the original capacitors are
- a detailed list of the voltage rails
- 2 pictures of a board in its original condition that needs a recap (it had a note with “No Internet” on it).

I'll post both polymod versions in separate posts in a moment. Stay tuned.

[momaka]

HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

This is version 1 of the polymod (i.e. the "prototype" computer I have been using).

Attached below (in order) are:
- a list of the replacement caps I bought
- a table showing which caps were replaced with the caps I bought above
- 2 pictures of the recapped board

Note that I didn't replace some of the small caps. I didn't *think* they are that important (but they may be?). If you would like to do a full recap, add 4 more 6.3 V, 470 uF Panasonic FR or other equivalent caps. These will replace the 2x 16 V, 100 uF caps on the D rail by the NB, the other 6.3 V, 470 uF cap on the J rail, and the 6.3 V, 470 uF cap between the two lower PCI slots. That leaves 1 more 16 V, 100 uF cap by the NB to be replaced. Going by the G-Luxon LZ data sheet, this is a "low ESR" cap. Funny thing is that even Panasonic FC (which is an entry-level low ESR cap series) surpasses the specs of G-Luxon LZ - so pretty much any low ESR cap from a reputable brand will work here. Finally, there's also 2 more caps on the lower left of the board by the Front Audio (?) connector (not shown in my cap map). Not sure what they are either (perhaps 16 V, 100 uF as well) but I doubt they are that important.

Polymod Version 2 coming right up.

[momaka]

HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

And here's Polymod Version 2...

Order of files is same as above.

Note that in this version, I took some "shortcuts" - pretty much like the bean counters do . There is, however, a single good revision in this version - the 6.3 V, 1000 uF cap on rail H is fed from a linear regulator so replacing it with a the 6.3 V, 1200 uF Chemicon KZE rather than a polymer won't compromise reliability. Not only that, but the capacitance on rail H is not lowered that way, and the two small 6.3 V, 330 uF GSCs (also on the H rail) can be left alone rather than changed with the 470 uF Panasonic FRs. This saves both time and caps.

Another place where I got lazy was with the CPU VRM high side input caps (the 16 V, 1200 uF ones). There I simply replaced only 1 of the original ones with a Chemicon KZE and filled the free spot with another KZE. That's probably a risky move there and I think the Teapo SM caps will give me troubles again... but with 20-something machines to go and not a lot of time left, I'm trying to save every minute I have. I actually bought a complete replacement for the caps on this rail for each PC, but since budget allowed caps for only 32 computers, I'm planning on using the leftover caps on other computers for which we did not buy replacements. And if a computer needs a full recap on this rail, I will still have the caps necessary to do it.
Also would like to point out: be careful if filling that free spot on the CPU VRM high side. The hole for one of the cap's leads is very close to a MOSFET pin. Make sure to double-check your work that the cap's lead is NOT shorted to the MOSFET's pin.

Lastly, the 6.3 V, 1000 uF cap between the two lower PCI slots on the D rail... I changed it from a 4 V, 560 uF poly in Version 1 to a 6.3 V, 1200 uF Chemicon KZE. Why? Because it's easier to put in since the KZE is taller. There's not much heat near those slots, so going from the poly to the Chemicon KZE should not compromise reliability. I also didn't fill the empty 6.3 mm spot on the C rail with the 470 uF Panasonic FR like I did in Version 1 - again, both a time and cost-cutting move.

With this second revision, I was able to bring down the cost to under $10 per computer. Of course, that's the cost of buying caps in bulk for 30 computers.
For the prototype in version 1, the cost was about $20 for the parts (including shipping) - but here I did not buy parts in bulk (only bought for that single computer, and the shipping alone was $6).
Overall, though, there isn't much difference in cost between Version 1 and 2. I just wanted to try different versions to see what the P4SD motherboard can take.

If fixing one of these DC5000/D530 SFF computers for yourself, I would also recommend adding some kind of support under the CPU socket area on the motherboard to relief some of the stress from the CPU heat sink clips. All of the DC5000 motherboards I have seen so far have a warp under the CPU socket area, and I'm sure that's not good for the BGA on the CPU socket.

I would also recommend adding a small fan on the Northbridge heat sink - that thing runs so hot I actually burned myself a few times. Either that or a large fan on the top part of the case. With these mods, I think this computer can joint the other indestructible polymods.

Anyways, enough chat. I really need to learn how to make my posts shorter .

[Wester547]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

I don't mind that your posts are long. If anything I prefer that as more information is relayed in such a way, especially on this forum, but that's coming from someone who loves to post long books. :P Of course, by that I'm not demanding people post epic rants and posts here ad nauseam. As for me, I'm (no offense) kinda happy I don't have a HP DC5000/D530 SFF.... I don't like the idea of having a motherboard or PSU run that hot, especially with Taiwanese capacitors.... I think those are a huge no-no on motherboards since (if I'm not mistaken), motherboards are susceptible to the hugest surge currents of any kind of hardware (by comparison to PSUs, video cards, sound cards, monitors, speakers, etc, if I'm not mistaken) which means the most stress (hence the need for low-ESR/motherboard grade capacitors but people here know much more about that than I do ). This is especially so with aqueous capacitors (to my understanding) and of course with electrolytic capacitors (which via heat dry out quicker, obviously, even Japanese capacitors), along with Class II and below Ceramics, I believe (as in their capacitance may go down very quickly especially with added heat).

But to think these motherboards have GSC capacitors.... OUCH! MAJOR ouch. Hard to get worse than that, especially on a motherboard. These things aren't the only systems with hot northbridges, though - to my knowledge - some of those GeForce 6100 chipset motherboards have notoriously hot northbridges and maybe even southbridges. I guess also that momaka's posts kinda explain why they cut costs in choice of capacitors in PSUs (IE using general purpose capacitors). Even then I think they go too low in quality. But another problem with these SFFs is of course the use of BGA chipsets.

[mockingbird]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Momaka - Are you at all looking inside the power supplies? What did HP use for these SFF models? I think their towers had Hipros... Is it economical to repair Pentium 4 systems?

[momaka]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Momaka - Are you at all looking inside the power supplies? What did HP use for these SFF models?

Yes. It's part of the routine after the motherboard is recapped. Only saw one with visibly bulging caps - but that's because the fan in that one had stopped for some reason (the fan was okay, but I think the controller has kicked the bucket).
I have counted 3 different manufacturers so far:

E143709 - Chicony Power / HiPro. These have lots of tan/brown conductive glue and use mostly OST RLS caps. In fact, the glue has become so well cooked in these that I was actually able to measure resistance across it with my multimeter when it was out of the power supply! And no - I didn't measure the resistance across my fingers. Got a good few hundred KOhms.

E132068 - Lite-ON. Some of these have United Chemicon KZE and also Taicon IIRC. My notes don't tell me if I saw brown glue or not so I assume there was probably small amounts in places.

E141190 - couldn't find it in UL's database. The only info I could find was on Olilio.com, which says that the power supply company that makes these is INTERTEK CORP. Anyone heard of it before? These are actually not bad either. Caps are mostly Samxon GF (ok, not good I know ) or Rubycon ZL. There's not a trace of tan/brown conductive glue in these, though.

Speaking of their power supplies, I do also have a list of the caps that these PSUs use. If anyone cares, let me know. Despite the fact that these PSUs are made by different manufacturers, they all share the same design and layout on the PCB.

Is it economical to repair Pentium 4 systems?

Well, that all depends on your needs. Like I mentioned, these are more than powerful enough for office needs. Certainly cheaper than buying new computers. Probably less work too since we'd have to set up all of the new computers, whereas these are set up already.

In terms of the repair - I got it down to under $10 per computer (I think it was actually close to $8.50 IIRC). That's for 32 computers. Do more and the cost per computer will fall even further due to economies of scale.

It takes me about 2 hours to do one of these computers - that is, take it apart, take out the motherboard and recap it, check the PSU caps, put it all back together and test for about 20 minutes. I can actually do these in less than 2 hours, but I didn't notice that some of the caps I bought had pre-formed leads and were in ammo packs rather than bulk - this takes me time to remove the caps from the tape and straighten then cut the leads.

I'm also using only a 35W iron. This doesn't slow me as much though. Only around the CPU VRM low caps - have to wait about a minute between cap removal/installation to let the iron recover.
I do actually have a digital soldering station at home, but I don't feel like lugging that around with me (although I might start to).

I don't like the idea of having a motherboard or PSU run that hot, especially with Taiwanese capacitors....

Hehe. I wish you'd have seen the DC5000 in one of our employee's office - it was flipped on it's side so that both the intake and exhaust vents on it were blocked by the floor (so the CPU was blowing hot air inside the case). When I touched that case, it felt like hot bread in my hands! I asked the employee why it was on the floor flipped like that and she said because it was blocking part of a walking isle. I then offered to relocate the PC so it wouldn't bother her and she agreed. I really don't know for how long it's been running like that, but this is not one of the recapped PCs, and all of the caps in it were still good. Go figure.

I think those are a huge no-no on motherboards since (if I'm not mistaken), motherboards are susceptible to the hugest surge currents of any kind of hardware

Not really. When I recap these, I never use an anti-static wrist wrap or anything like that. I just touch a metal surface to discharge myself, then take out the motherboard and put it on my legs to work on it. Been doing all of my electronics like that.

But to think these motherboards have GSC capacitors.... OUCH! MAJOR ouch.

There's only 2 of them and they are tiny. No stress on them whatsoever either. Mostly there for bulk capacitance - hence why I didn't bother to replace them in version 2 of the polymod. They are in parallel with a 1200 uF KZE cap that does all of the filtering.

[c_hegge]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

I'd stick 470uF 16V polies on the CPU VRM High side too.

Looks good otherwise, though.

[jsog]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

The power supply in the D530 has problems, but can also be repaired: http://webdevsys.com/d530psuRepairNotes.htm

The cooling in these systems really does not need to be as bad as it is. One problem is that the inside fan on the power supply is blocked by a bundle of cables. It helps to zip-tie them and get them out of the way as much as possible.

[momaka]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

I'd stick 470uF 16V polies on the CPU VRM High side too.
Looks good otherwise, though.

Thanks.
Yeah, that would be a very good idea if someone wanted to do a full polymod.

The power supply in the D530 has problems, but can also be repaired: http://webdevsys.com/d530psuRepairNotes.htm

Wow, very useful article. Thank you for the link!

Luckily, none of our DC5000 had PSUs go bad (even though the PSUs are the same as those for the D530). 31 of our computers have sat in storage for close to 2 years now, so we will see what happens as I (slowly) recap them. IIRC, there was a YouTube video showing the power supplies blowing in these after sitting in storage for a while. Some people thought it was fake, but given the brown conductive glue, I don't think it was fake. Every time I plug in a "new" DC5000 that I haven't worked on before, I sort of anticipate a bang.

Yesterday, I had a though one. The Panasonic FL caps around the CPU on one computer had really soft rubber bungs and their leads would easily pull out of the body as I tried removing them. I have removed other FLs before, but they were much tougher than than. Not sure why these were so delicate. After removing only 4 Panny FLs, I said "the hell with it" and didn't bother with the rest.

I think I'll bring my soldering station one of those days. Need to make the repairs go faster. The 35W iron I'm using works absolutely fine for the rest of the board. Even the CPU VRM low side isn't much of a problem but I have to wait about 20 seconds for the iron to recover before removing/installing another cap.

BTW, a very important thing I forgot to mention (funny I say it at the end of everything now, lol):
THIS IS AN ASUS MOTHERBOARD AND CAP MARKS ARE BACKWARDS! THAT IS, THE WHITE SHADED SEMI-CIRCLE FOR EACH CAP SPOT INDICATES WHERE THE POSITIVE SIDE OF THE CAP GOES AND NOT THE NEGATIVE SIDE! Make sure to double-check your work, or expect fireworks otherwise.

Since I haven't done that many motherboard recaps prior to this, it's almost natural to me now put the cap's negative lead on the non-shaded hole. I guess I have to be careful not to put the caps "backwards" on a "normal" board now .

[mockingbird]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

momaka -

I actually had to repair a P4SD last week... I couldn't find a datasheet for the three Teapo SM caps on the VRM high... Are these Teapo's version of a Rubycon MBZ? I noticed you used KZE... The best equivalent I had were 10mm Nichicon HM 16V 1500uF caps. They seemed to work fine. Those weren't the caps that were bloated though, it was one of those two Teapo SZ caps (6.3V 1500uF) near the memory slots that bloated.

Everything else on the board was Nichicon HE IIRC so I left them alone. I also left the VRM low alone because like yours, mine were populated with Panasonic. Were yours out of spec?

[mockingbird]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

To answer my own question...

I found the datasheet for Teapo SM. I don't know why Teapo Cheapo eradicated every trace of it from their site, but attached is a bona fide datasheet from them.

According to the datasheet, there are two versions of the 16V 1200uF cap. It comes in 10x19 and 10x22. Ours is 10x22mm.

Teapo "SM" 16V 1200uF 10x22mm:
2150 0.022

UCC "KZE" 16V 2200uF 10x25mm:
2150 0.022

Yup. KZE is a perfect match here...

For curiosity's sake, let's have a look at the Nichicon I used:
Nichicon "HM" 16V 1500uF 10x20mm:
2550 0.013

So Teapo "SM" is not the same as G-Luxon "SM", it needs to be replaced with UCC KZE / Nichicon HD / Rubycon ZL class caps or better.

[momaka]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

I actually had to repair a P4SD last week... I couldn't find a datasheet for the three Teapo SM caps on the VRM high...

I couldn't find a data sheet before either. Thanks for attaching it. Luckily, one of the 31 DC5000 PCs we have had Nichicon HDs on the VRM high, so I just assumed Teapo SM was of equivalent class. Moreover, many P4 motherboards of that era had 3 to 4x 16 V 1000-1200 uF United Chemicon KZE caps on VRM high so I figured Chemicon KZE would work fine.

Everything else on the board was Nichicon HE IIRC so I left them alone.

I only have 2 or 3 boards with those so it's not enough of a sample to say if they will be a problem or not. But they all seem okay and I *think* they should be okay. Also have one board with ZLs.

I also left the VRM low alone because like yours, mine were populated with Panasonic. Were yours out of spec?

Unfortunately, I don't have an ESR meter to check them, but out of the 31 computers I checked, only 12 had Panasonic FL on the VRM low side and none of them appeared bloated or leaking.

The other 19 computers had Rubycon MCZ. Here, the story is different, though - 14 out of 19 systems had at least 2 or more bloated/leaking MCZs. So if you have MCZs, recap it.

I also started replacing some of the Panasonics in those PCs that had them - mainly the 2 lower ones below the CPU socket (since those don't get enough airfrow from the CPU fan) and the uppermost one along with the 3rd up from bottom. If using polys for the replacements, you'll end up with 8 polys total - which is more than enough to sustain a stable CPU voltage even if the rest (3) of the electrolytics completely failed.

[tmiha71]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Nice thread, I'll put some pepper on it :-))...

Proper way to check, if recap is successful is – putting scope on CPU-vrm-out capacitors (or/and CPU voltage feedback resistor to switching regulator), take a shoot (trace) in idle (very important), and on full load (or any heavy load), also on +5V rail, +3.3V rail and 12V rail in same conditions (probably in dual setting CPU and +3.3V, CPU and 5V rail, CPU and 12V rail)...

What's next?

Looking in time-domain (traces per se), very questionable, scope should have FFT analysis – and all results (traces) should be analyzed in frequency domain (and correlated if possible, that I'm guessing). Or exported to computer and analyzed with some software in frequency domain.

It should be looked for dominant spectral component in CPU voltage (and other voltages)...

And now some guess-work (data not specified in data sheets of capacitors, at least not by majority of manufacturers), if we have spectral components below resonant frequency of each used capacitor – we have potential problem -> electrolytic capacitors will work overtime (poly or not poly – same thing, poly will last longer because of capability of higher ripple current)...

For example: spectral component of 10 kHz is identified, and Sanyo SEPC is being used (they published needed information), we look at "Frequency coefficient for ripple current" and we'll see that we have only 0,7 of declared ripple current, if it's below 10 kHz we have ONLY 0,3 of declared ripple current (yaaa, it hurts :-()...

If we identified spectral components above 500kHz (upper limit of specified value for Sanyo SEPC, or near to), we are on VERY VERY GOOD track, that means that only ceramic capacitors (on CPU, or on MOBO) are working – as it should be, and electrolytic are only holding output voltage...

If we are in range 100 kHz to 500 kHz, we have coefficient of 1 (max ripple to use), and we can hope that we have enough ripple current, for capacitors, to live long.

So, next few things to be done, put scope to 10 mega samples per second (theory, Nyquist, says that we will capture frequencies up to 5 MHz), record traces as long as possible (but no more than few second), and analyze them in frequency domain.

As noted above, properly recapped (designed :-)) system should have majority of spectral components AT OR ABOVE RESONANT FREQUENCY OF EACH USED CAPACITOR. Or at least, should not lower dominant frequencies -> capacitors will last longer.

Additional note (pure ques, can not find any proper info): For electrolytic's "seems to be truth" -> higher capacitance (or bigger case?) -> lower resonant frequency, so in some cases increasing electrolytic capacitance can help (partly removing hard work from electrolytic capacitors to ceramics, or regulator), (probably ?) another example of lowering resonant frequency of capacitor https://www.badcaps.net/forum/showpo...8&postcount=24 ...

Actually I'm pretty curious, on frequency images of CPU voltages on NEW motherboards, and GPU voltages on graphics cards too...

Additional note: Low spectral components in CPU voltage could be result of too much ceramic capacitors (on CPU, or on mobo, failure at design level)...

[Behemot]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

I've swapped all the bloated Rubycons for OS-CON SEPC, other lytics for same lytics. Haven't changed other caps though but they measure fine. Wanted to try if it works, board is frozen at 0809 (or maybe 0908?) on POST card. Any idea? I may changed it for some other uATX board, it is more or less compatible, unlike bloody Dell systems.

[momaka]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Wanted to try if it works, board is frozen at 0809 (or maybe 0908?) on POST card. Any idea?

Nope, I've never seen that before.
Out of all of the boards I fixed, only one didn't work right - it failed to detect IDE and floppy drives. Couldn't fix that one even after changing all of the caps.

Normally, I change most (but not all) of the small caps. The small G-Luxon and Ost caps around NB always prime suspect as there is lots of head in that area and they can dry out.

[mockingbird]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Sorry to bring up an old thread.

I'm having a problem with one of these boards I re-capped. I'm now getting the following message:

"The following configuration options were automatically updated"

Then it tells me that what was updated is that there's no fixed IDE device or that there's a CD-ROM (Depending on what's connected), but then it sends me on an F1 save configuration loop, where it reboots and throws the same message.

I've tried a total re-cap, I even swapped the 270uF Tantalum cap underneath the CPU, I put in a fresh Energizer CR2032, I've tried everything. The only thing I haven't tried is replacing the two tiny 47uF caps near the white IDE connector or replacing the purple Sanyo OSCON near the CMOS battery.

Any suggestions?

[stj]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

if you power it off for a while - completely.
does it remember the date & time?

[mockingbird]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Yes.

But if I disable the IDE and SATA controller in the BIOS completely, that's not retained...

The memory slots are also finicky and the letters on the screen become garbled and the computer completely freezes after that happens, which is why I re-capped in the first place. But I didn't get the settings saved loop before I re-capped. The garbled letters crash would happen while Windows was loading. Maybe the BIOS is corrupt?

I wonder if I can run the HP BIOS recovery if the machine posts.... I'll have to look at the documentation for this thing.

[Behemot]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Is there a chance you have damaged some paths during recap?

[Th3_uN1Qu3]

Re: HP DC5000 SFF and D530 SFF (ASUS P4SD motherboard) - polymod

Either scratched/damaged traces or it has a problem with the southbridge.