Bad caps, but machine is going strong.

[PTH]

Hi everyone,

I have the opposite problem of most people here. I was cleaning out my PC (dust and all) and found 2 pretty bad caps. What looks like rust entirely covers the top of one capacitor, and covers 80% of the surface of another one, just beside the processor. I'm guessing this damage has to be 6 months old at most, since I should've noticed this the last time I cleaned it, if it was there. The caps are directly beside the processor socket.

This is an old celeron 1.7 with a 400FSB, so it's not worth having it fixed, but I wonder for how long it'll keep on working. It's been turned on 24/7 for essentially the last 18 months and has been solid as a rock.

How long do you guys think it'll keep on working? Is there any danger (other than to the motherboard and perhaps other components in the case) to letting it keep on acting as a server until it dies on me? (if it ever does!)

Thanks for your thoughts.

[linuxguru]

Re: Bad caps, but machine is going strong.

There are usually a bunch of caps in parallel on most power rails, including Vcore. The system may continue to function as long as there is at least one good cap on each rail. However, as caps keep failing, the stress on the remaining good ones keeps increasing, causing them to fail in succession.

The last few good ones may fail fairly quickly - within a few days. It's hard to predict how long your board will last - but failure is a certainty, it's only a matter of time. If the caps are replaced before the board completely dies, there's a good chance that the board will last a long time. OTOH, when all the caps on a power rail fail, they can take a bunch of components with them - including the CPU, Northbridge, Southbridge, Memory, AGP cards, PCI cards, etc.

[PTH]

Re: Bad caps, but machine is going strong.

Thanks for the advice.

I guess this solves my problem, I won't be selling off my Sempron 2200, I'll use it to replace the Celeron, since I know that the Celeron doesn't have much of a lifespan left.

I also have an old Duron with a bad cap beside the AGP slot, and that one works fine as long as you don't use an AGP card. I wonder how long that'll last.

On a completely unrelated note, anyone interested in a GA-SIML motherboard with 2 bad caps (along with countless other potentially bad ones), along with a Celeron 1.7 (400FSB on socket 478), heat sink and fan?

....

....

didn't think so...

[kc8adu]

Re: Bad caps, but machine is going strong.

i am sure someone will recap it and use it for a folder if nothing else.
i sure wont turn it down.
just because its old news dont mean its useless.
hell this is a p3 1.4....duallie!

[Shroomie]

Re: Bad caps, but machine is going strong.

The Celeron might be useful actually. I mean, my mother's on a P3-500, won't be for long since I've gotten a couple P4s, but still...

[larrymoencurly]

Re: Bad caps, but machine is going strong.

If the capacitors go bad enough, the transistors connected to them have to work harder and run hotter, their extra heat makes the capacitors deteriorate even faster, causing the transistors to work even harder and hotter, and eventually...

At the very least, I'd turn off that computer at the power strip or rear switch whenever it wasn't in use.

Not worth fixing? My newest stuff is slower than your "old" Celeron.

[PTH]

Re: Bad caps, but machine is going strong.

i am sure someone will recap it and use it for a folder if nothing else.
i sure wont turn it down.
just because its old news dont mean its useless.
hell this is a p3 1.4....duallie!

Well, not useless, but I did the math:
55$ + tax for a new socket 478 motherboard (I'm not going to get into fixing these caps, I used to work with some saudering equipment and, well, I was pretty lethal for the boards...) to keep on using an old celeron, or 80$ and I bought a Sempron 2200 with motherboard, used.

I actually managed to sell the motherboard with processor to an unsuspecting twit for 65$ (he was warned explicitly about the caps, what he chooses to do about it now is his problem).

Just for the heck of it, I put up an add for my Duron 900 on another mobo with dying caps, and I've gotten interest even with a sticker price of 100$... I don't get it - you can get a mobo+processor for 140$, and they'll be 2-4 times as powerful as the Duron.

(all prices in Canadian dollars)

[gonzo0815]

Re: Bad caps, but machine is going strong.

Well, 65$ is ok for it, assuming that caps are handy and soldering isn`t a problem this combo will made a in any way good office system. But certainly, at the moment with those realy cheap soket 754 boards & processors i won`t buy any soket a or similar intel stuff any more. Those semperons on socket 754 are that amzingly fast, cheap, stable and energy efficient that IMHO nothing else could bet beter suited for a silent office system.
i can`t understand why people paying more than $100 (in Germany they are some XP-M sold well above 120€) for some soket a proc`s. But mey be people are getting used tho that stuff as it has been availiable over several years.

[linuxguru]

Re: Bad caps, but machine is going strong.

If you have an existing very stable and reliable Socket-A/SDRAM system, but merely want a faster CPU, it pays to upgrade just the CPU - you don't have to pay for a whole lot of other stuff like DDR, PSU, etc. However, due to motherboard FSB and core constraints, you may have to pay a premium - for instance, for high-clock Thoroughbred parts at 266 FSB.

That said, I'd second the recommendation to upgrade to S-754 or even better, S-939. The only issue is that the supply of S-939 Semprons is very tight, thus forcing most people to use an Athlon-64 instead. Even so, I was able to get a S-939 Athlon-64 3000+ Venice + mobo for about $140 - it's probably even lower now because of the AM2 transition.

[gonzo0815]

Re: Bad caps, but machine is going strong.

Well going a little further ot
I my experience you can use any ddr ram on soket 754, at least i have used ddr 266 and 333 foe testing purposes without any problems. I doubth that the rig would loose that mutch performance with that. And with SD-ram i think it is time to upgrade, as it could be dificult (due to chipset limitations and price)to obtain that large ammount of ram i probably want to have even in a office system today. I think people arent aware, that those soket 754 systems can deliver roughly the very same performance than those soket 939. And therefore they ignore them. I have done it a long time too until i realized that any chepo soket a system will cost a lot more an be therefore a wast of money. But ok, hardcore oc like trodas pushing the system near the 3ghz and further could be a reson for those XP-M with barton core.

[linuxguru]

Re: Bad caps, but machine is going strong.

those soket 754 systems can deliver roughly the very same performance than those soket 939.

1. Socket-754 is Single-channel DDR, so it's the same as Socket-939 when there's only one DIMM installed. But with two DIMMs, they're very different beasts - Socket-939 dominates on any memory-intensive task.

2. There are fewer CPU upgrade options on Socket-754 - in particular, no dual-cores. S-939 supports some superb dual-cores, including the Opteron 165, 170 and 175.

3. A Socket-754 Athlon-64 costs the same as a Socket-939 Athlon-64 at equal clocks, so there isn't any savings there. Socket-754 is really useful only with Semprons.

4. Socket-939 Semprons are in (deliberate?) short supply, but S-939 mobos are easily available in the gray market. That seems to a strategy by AMD to move the higher-priced Athlon-64s in preference to Semprons on S-939.

[toastygoodness]

Re: Bad caps, but machine is going strong.

Socket AM2 also has come out, using DDR-2 RAM. All desktop processor lines from AMD are available with AM2.

[gonzo0815]

Re: Bad caps, but machine is going strong.

1. Socket-754 is Single-channel DDR, so it's the same as Socket-939 when there's only one DIMM installed. But with two DIMMs, they're very different beasts - Socket-939 dominates on any memory-intensive task.

2. There are fewer CPU upgrade options on Socket-754 - in particular, no dual-cores. S-939 supports some superb dual-cores, including the Opteron 165, 170 and 175.

3. A Socket-754 Athlon-64 costs the same as a Socket-939 Athlon-64 at equal clocks, so there isn't any savings there. Socket-754 is really useful only with Semprons.

4. Socket-939 Semprons are in (deliberate?) short supply, but S-939 mobos are easily available in the gray market. That seems to a strategy by AMD to move the higher-priced Athlon-64s in preference to Semprons on S-939.

1. but what will do this for an gamer in real world? may be 2 to 6 fps.
Thus it is no point for me.

2. Dualcore is at the moment 170 -280€ entry level.
Not my price range and for most people too. For that money il better go with a better gfx.

3. yes, i am refering to semperons, which cost roughly half the price of any s939 cpu.
And i think they are all faster than most soket a cpu`s, may be faster than all without oc.

Fore shure, if you want to buy any Amd Athlon 64, i won`t go the s754 route. At the moment i probably went straight to AM2, as it only makes about 30€ difference.

[linuxguru]

Re: Bad caps, but machine is going strong.

... i am refering to semperons, which cost roughly half the price of any s939 cpu.
And i think they are all faster than most soket a cpu`s, may be faster than all without oc.

Yes, I agree - Semprons are indeed the highest-performing value CPUs out there and will easily dominate just about any Socket-A CPU, and cost about the same for a CPU+mobo.

However, they're cache-constrained - so you'll be tempted to upgrade somewhere down the line - when you'll find that single-channel Athlon-64s are the only option. This makes the LGA775+Celeron D/64 a viable alternative. If you choose the right mobo, the upgrade options include Smithfield/Cedar Mill dual-cores, or maybe even a Kentsfield quad-core on an i975X board.

(If I have a working and stable mobo+RAM, I dislike throwing it out when I only want to upgrade the CPU for higher performance. The time and effort to recompile/reconfigure/revalidate an existing, working, custom GNU/Linux installation on a new mobo is too high. OTOH, a new CPU alone can breath a lot of life into an older, validated box with little or no effort to reconfigure the software. That's why I went for S-939 - it allows an inexpensive future dual-core upgrade. A year from now, Socket AM2 might also be a viable option - it supports everything from Semprons to dual-cores.)

[emboss]

Re: Bad caps, but machine is going strong.

I have the opposite problem of most people here. I was cleaning out my PC (dust and all) and found 2 pretty bad caps. What looks like rust entirely covers the top of one capacitor, and covers 80% of the surface of another one, just beside the processor. I'm guessing this damage has to be 6 months old at most, since I should've noticed this the last time I cleaned it, if it was there. The caps are directly beside the processor socket.

Heh, had the same thing happen here with a Soltek 75DRV5 (the one I'm slowly moving towards recapping). All four CPU VRM caps leaked out the top, and most other big caps on the board are bulging.

It had been running about 6 months straight as a gameserver (and the board had already lived through 3 years of relatively hard use), during which I never had a reason to open it up as it was perfectly stable. I then upgraded the ram, at which point I noticed the caps in this state. The machine was still rock-solid stable. I didn't believe it could be stable in this state, so I had it run prime95 overnight. Passed the test with no problem. Since I wasn't too concerned about anything in the server, I just chucked everything back together and set it running again. It ran without any problems for another two months 24/7 before I had to shut down the server because I was moving.

Would I recommend you do the same? No, unless you don't care about anything else in the server and have a smoke detector mounted close by!

[Per Hansson]

Re: Bad caps, but machine is going strong.

This is pretty interesting, I guess it must be because it has a nice VRM design (3 phase by the pics on Soltek) and also I'm assuming that you use a CPU which draws relativley little current therefore the ripple filtering done by the PSU is enough

Still very very impressive

Just curious; how hot does the Mosfets get when running Prime95?

[Rainbow]

Re: Bad caps, but machine is going strong.

I've recapped 3 or 4 PCPartner Tualatin boards at one site about a year ago. There is one more remaining there where the caps are bad too. The boards were bought together and that one of them still works even with bad caps. I'm going to recap it probably next week. These boards have 4x 4700uF Licons on Vcore (IIRC).