@#$%*! Capacitors

[phaedrus]

Thrusday night, around 10:30, my box flatlined. I was running my computer as I usually do, and I was listening to the Detroit Industrial Underground 'net radio station, and my box had a hard crash. The screen froze without Xwindows running and the music ground to a halt, leaving the soundcard locked repeating the last few samples and making a sound disturbingly like the life support monitor of a dead patient (I swear, it was almost exactly the right tone).

I stared at my screen listening to the tone in my headphones realizing what must have just happened. Slackware releases never grind to a halt. Ever. Particularly not when only command line programs are running.

I shut my machine down, and tried to get her to boot. She wouldn't consistently POST with both processors in and would only recognize one processor. I tested each processor by itself, and she would boot up running uniprocessor. She was just stable enough to fsck my harddrive and shut her off. The symptoms looked a lot like the problems I was having with my dual celerons with only a 250W PSU, only much more severe.

I had decided earlier on Thursday to re-cap my board, and had ordered a replacement set of caps from Digikey already. I prayed that all that had happened was that the caps had finally gone.

I ran some errands this morning and when I got home for lunch, the capacitors were sitting on the porch. I brought them inside, checked the box to make sure everything came (all pieces were there). I made lunch, ate and packed up my BP6 to take in to work. Redoing the caps took longer than I thought it would, by many hours (remember, I do point-to-point construction normally, not multilayer PCB, and the solder on my board was old and didn't want to melt for anything).

I'm typing this from my BP6. The operation seems to be successful. I'll post pictures later. I increased most of my capacitor values. I'm also using 10V instead of 6.3V caps, so they're all a lot bigger instead of just a little bigger. I put just about everything about 1000MFD at 1500MFD. I boosted the 12 caps around the CPUs to 2200MFD as per Yoichiro and I boosted EC12, EC16, EC26 and EC27 to 1500MFD from 100MFD as per the Friendtech suggestions. It now looks like there is a forest of caps around the CPUs.

133MHz FSB still eludes me. When I set the FSB to 110, 124 or 133, keyboard LEDs flash, but the monitor blanks and turns off, and nothing happens. This makes me think that either the GTL+ bus or the AGP bus doesn't like the higher settings. However, I get the same symptoms when I run my AGP bus at 2/3 of the FSB, which makes me think it is not the AGP bus that's causing problems. There's only one mod I've seen out there that might have an effect on the GTL+ bus, and that's the "short the diodes" mentioned by Friendtech, but left unexplained (they never replied to my email inquiring about it, bastards).

The quest continues.

Jeff

[Dave Rave]

nice write up.
bloody hard to get an indignant board to post.
must buy a special hammer to intimidate them with next time.

I've had my boards re-capped, and two of them were still flaky, turns out it was the also flaky psu. (sigh)

[phaedrus]

nice write up.
bloody hard to get an indignant board to post.
must buy a special hammer to intimidate them with next time.

No. No hammer needed. We've got an understanding. I treat her with care, she runs, but she is being indignant about running any faster than 100MHz.

I'd been gambling on the caps, I had some indication they were going for quite a while. It's kind of amazing though, they went the day I ordered the new set--the caps went late at night, and I'd order the new set earlier that afternoon.

I've had my boards re-capped, and two of them were still flaky, turns out it was the also flaky psu. (sigh)

She's feeling nice and stable, so I think it did the trick. The PSU is new, so I knew that wasn't the problem (well, by "knew," I really mean, I could reasonably rule it out as the likely cause, particularly when I knew I had a timebomb sitting on the board in the form of the caps).

It feels so good to have rigel back up and running.

Jeff

[jaybird]

I am presently running a Ver. 1 board with ONLY the C10 and BASIC voltage regulator mods plus the basic 370 adapter mods, but the one point I want to make with people is the NEED to cool the NorthBridge chip set!!!!!

IE: "greenie", what ever you want to call it, it MUST get EXTRA cooling!

GOTO ebay, "mini-peltier" (costs $4.95 US plus shipping & handling), remove greenie, add "Arctic Silver" paste, take a plastic wire-tie or wrap and secure "mini-peltier" to "greenie".

I have seen a 20 to 30 deg F drop on initial startup on NorhtBridge!

Bottom line, you want 110 mhz fsb with PIII's, MUST COOL!

Dual PIII's, 1100E's, @ 110 mhz, 768 meg PC133 SDRAM
Dual boot 98SE/2000 Pro
To much to list! want screen shots or equipment list email me!

Regards,

jaybird

[phaedrus]

Bottom line, you want 110 mhz fsb with PIII's, MUST COOL!

I know. Cooling isn't my problem right now. After posting the above, I got curious, and pulled my PCI cards out and tested to see what my maximum FSB is right now. Turns out, I've got a hard limit between 104 and 106 MHz--in other words, she'll POST and run just fine at 104, but when I crank it up one more notch to 106, I get the "nothing displays" problem. Everything above that won't POST. So, she doesn't seem to be running (or if she is, something is broken such that nothing displays, and it isn't usable at the moment). I'm still running at 100, just because 104 hardly seems worth it (the gain isn't enough to warrant OC'ing the PCI bus, even just a little).

An overheating problem would likely manifest itself as stability issues once booted--but the board would POST ok, because it takes at least a little time under load to overheat. So, it isn't overheating at the moment.

I am planning on cooling the BX once I get the rest of the board to play along. The question is how far I can push the board. Hopefully, I'm not at the limit yet. (I feel like I should be able to hit 110 at least by messing with components, because we know the PCB can handle it, and I've seen higher in screenshots, Friendtech has a shot of a board POST'ing at 124. I recant seeing 133, I thought Yoichiro had one, but, in fact, he doesn't, he only made it to 110, in fact, at 124MHz, he runs into same problem I'm having... Hmm, looking over his data one more time, well, it's giving me bad ideas ).

Jeff

[Dave Rave]

are you also playing with the pci/agp ratio in bios ?
might be 104 is the limit for whichever you're using, and you need to use the other ?

[purrkur]

Redoing the caps took longer than I thought it would, by many hours (remember, I do point-to-point construction normally, not multilayer PCB, and the solder on my board was old and didn't want to melt for anything).

I can also say that the first bunch of caps took me quite a while to remove. I have now found a good method of removing them. I use a solder "sucker" (one of those "gun" type of things), but my problem has always been trying to heat the old solder and getting it out with the sucker. So what I do before trying is to resolder the points that I want to remove, adding a generous amount of tin on the point. This way it is a lot easier to reheat and remove.

But I agree. Removing components from a multilayer PCB is not easy, especially if they have been soldered on there with a wave machine.

I'm typing this from my BP6. The operation seems to be successful. I'll post pictures later. I increased most of my capacitor values. I'm also using 10V instead of 6.3V caps, so they're all a lot bigger instead of just a little bigger. I put just about everything about 1000MFD at 1500MFD. I boosted the 12 caps around the CPUs to 2200MFD as per Yoichiro and I boosted EC12, EC16, EC26 and EC27 to 1500MFD from 100MFD as per the Friendtech suggestions. It now looks like there is a forest of caps around the CPUs.

You have used pretty much the same replacements I did on my ver 1.1 board that I have recapped.

Btw, I have found out what is ailing my BP6 that I have recapped. It seems as if one of the coolers isn't sitting correctly on the CPU so that CPU gets really hot and eventually the board freezes up (of course). I am going to London today to work but I intend to take a good look at it once I am back on Thursday.

Oh, and thanks for the net radio station link. This should be right up my alley...

[phaedrus]

are you also playing with the pci/agp ratio in bios ?
might be 104 is the limit for whichever you're using, and you need to use the other ?

Didn't have any effect. At either 1/1 or 2/3, the highest I can get a display with the CPU voltages at their default (1.7V) is 104MHz.

If I bump the voltage to 1.8V, I can get a display at 106, and this seems to be the final limit (without doing any more soldering--I could push the processors harder, but that would require more soldering. Yoichiro has directions on how to do this, but I think the higher road might be to figure out what is blocking up the GTL+ bus, maybe it's a component I can change out). At 1.9V, 106 is the highest, but I checked more carefully, and she'll POST at 108 and 110 (I can tell by watching the mouse and keyboard LEDs, they flash in a particular order, and the BIOS turns numlock on right before it loads the boot loader). At 124, I get absolutely nothing. No LED flashes beyond the first (which just happens at power-on, no matter what).

My reasoning for why I think it is the GTL+ bus is that boosting the CPU voltage gives me a little boost to get over the "obstacle" (if only a little bit). However, changing the AGP bus speed doesn't affect whether or not I get any display when it POSTs, which I take to mean that my problem isn't the AGP bus to the card (and the card itself is designed to handle 266MHz @ 1.5V and 133 @ 3.3V, more than enough for what I'm trying).

Boosting the CPU voltage was the bad idea that I got after rereading some of Yoichiro's later mods (where he was trying to OC his PIIIs. I didn't read that area carefully before, because I just want to run the PIIIs at spec, however, for me, that means OC'ing the BX, and for him it didn't (he used 800/100 PIIIs for his mod)).

The question, then, is what is the culprit? Some component between the BX and the CPUs? The BX itself? (I hope not the latter... I just checked Intel's spec sheet on the BGA 492 package, and I have no way to replace it... It's designed for largescale industrial settings, like a motherboard manufacturing plant. Damn.)

Jeff

[hyperspace]

I think the BX can handle 133 if running uni-processor. Dual is just too much overhead for the BX to handle when overclocking higher than 110Mhz. I wonder if the BX was designed with Dual in-mind.

[InactiveX]

Holodeck2 had his BP6 running dual at 133MHz FSB, but that was with Celeron 366s. If I remember correctly, he used a lot of ducting with watercooling and peltiers.

GOTO ebay, "mini-peltier" (costs $4.95 US plus shipping & handling), remove greenie, add "Arctic Silver" paste, take a plastic wire-tie or wrap and secure "mini-peltier" to "greenie".

FYI, I couldn't get any results with "mini-peltier", but "mini peltier" (without the hyphen) threw up the right page. Looks like a nice cooler!

[phaedrus]

I think the BX can handle 133 if running uni-processor. Dual is just too much overhead for the BX to handle when overclocking higher than 110Mhz. I wonder if the BX was designed with Dual in-mind.

Very interesting. This implies a new test. I'm going to have to try pulling one of my CPUs and seeing how high I can clock the bus.

On an unrelated note, I've finally uploaded the post operation pictures of my BP6.

I can also say that the first bunch of caps took me quite a while to remove. I have now found a good method of removing them. I use a solder "sucker" (one of those "gun" type of things), but my problem has always been trying to heat the old solder and getting it out with the sucker. So what I do before trying is to resolder the points that I want to remove, adding a generous amount of tin on the point. This way it is a lot easier to reheat and remove.

I love those things. I've got one sitting at my workstation. I use it all the time on projects for work. I read over at http://www.badcaps.net that they can damage the PCB if used to many times. That and the solder didn't look like it was liquifying, I just barely got the solder to let go of the leads and was then able to pull the caps out.

But I agree. Removing components from a multilayer PCB is not easy, especially if they have been soldered on there with a wave machine.

Yeah. I ended up running out to a local hardware store and picking up a pin vice, it's a little hand held chuck for small drill bits. I also got some #65 bits (if I do this again, I'm going to get a slightly larger selection of bits between #60 and #70). It worked wonders for drilling out the holes. Yoichiro's Han-Getsu trick is what inspired this (I actually went in asking after one, but they didn't have one and suggested a pin vice instead, it worked great).

You have used pretty much the same replacements I did on my ver 1.1 board that I have recapped.

Btw, I have found out what is ailing my BP6 that I have recapped. It seems as if one of the coolers isn't sitting correctly on the CPU so that CPU gets really hot and eventually the board freezes up (of course). I am going to London today to work but I intend to take a good look at it once I am back on Thursday.

Cool. Good to hear you tracked that down.

Oh, and thanks for the net radio station link. This should be right up my alley...

Sweet. Glad to be of service.

Jeff

[phaedrus]

I got a small boost out of that. 124 fails as usual, no visual, doesn't POST. However, I can get 110MHz FSB without boosting Vcore.

I'm going to try shorting D1 and D2 this weekend.

Jeff

[purrkur]

I'm going to try shorting D1 and D2 this weekend.

Can't wait to hear how that turns out!

[davd_bob]

...solder on my board was old and didn't want to melt for anything.

133MHz FSB still eludes me. When I set the FSB to 110, 124 or 133...
This makes me think that either the GTL+ bus or the AGP bus doesn't like the higher settings. However, I get the same symptoms when I run my AGP bus at 2/3 of the FSB...
the GTL+ bus...

Jeff

OK, whats the GTL+ bus?

This whole thread with all the other posts brought back memories of my various experiences.

My board would run like a scalded cat at 1.9v at 100, boot W2K and run OK 2.0v at 104, but not even complete posting 2.3v at 110.(dual 366s)

Before replacing the 2-4packs of (8 caps) with Jackson free replacements, the system temp was like 10deg hotter then the cpus...after replacement it was lower then the cpus.

I only replaced 8 of the 24caps because I had difficulty getting the solder to melt on the GROUND traces and one of the new caps is stuck half an inch from being fully seated. I scared myself out of repalcing any more.

OH NOOOOOO. The board started acting pretty flaky so I did the Abit RMA thing. Guess what. They replied that they are currently OUT OF STOCK of replacement BP6s and offered a store credit instead. I am still waiting and hoping...

I agree with Hyperspace that BX MAY go 133 in UNI but I believe only if all onboard componants are in pristine condition. As far as cards
PCI is UNDERCLOCKED at 92 and 124
If AGP runs 1/1 at 100, it would be(comparativly speaking)underclocked 2/3 at 124.

I hope to get my board back from Abit RMA

OH, im a grandpa again since 3/5/05.

[hyperspace]

... OH, im a grandpa again since 3/5/05.

Beautiful! Boy or Girl !?

[davd_bob]

... OH, im a grandpa again since 3/5/05.

Beautiful! Boy or Girl !?

BOY
8lb 8oz. Mother and child are doing fine.
before the delivery the mother screamed her head off for 2 hours.
after the delivery the child screamed his head off for 2 hours.
then they both went to sleep.

[phaedrus]

...solder on my board was old and didn't want to melt for anything.

133MHz FSB still eludes me. When I set the FSB to 110, 124 or 133...
This makes me think that either the GTL+ bus or the AGP bus doesn't like the higher settings. However, I get the same symptoms when I run my AGP bus at 2/3 of the FSB...
the GTL+ bus...

OK, whats the GTL+ bus?

It's Intel's proprietary bus that goes between the BX and the CPUs. I'm enough of a nutbag that I've been digging around in Intel's technical documents.

This whole thread with all the other posts brought back memories of my various experiences.

My board would run like a scalded cat at 1.9v at 100, boot W2K and run OK 2.0v at 104, but not even complete posting 2.3v at 110.(dual 366s)

Before replacing the 2-4packs of (8 caps) with Jackson free replacements, the system temp was like 10deg hotter then the cpus...after replacement it was lower then the cpus.

I only replaced 8 of the 24caps because I had difficulty getting the solder to melt on the GROUND traces and one of the new caps is stuck half an inch from being fully seated. I scared myself out of repalcing any more.

Understandable. As I said earlier, I had a hard time myself. I'm just insistent and I have a backup board that I can throw in my box if I really hose things (I don't want to, it's a uniprocessor PII board, but it will let me get my school work done and it should still play Quake pretty well).

OH NOOOOOO. The board started acting pretty flaky so I did the Abit RMA thing. Guess what. They replied that they are currently OUT OF STOCK of replacement BP6s and offered a store credit instead. I am still waiting and hoping...

And since they're out of production, I don't know that they'll ever have any more, because it doesn't sound like they are re-capping the boards.

I'd get in contact with them and see if they're going to get more in stock.

I agree with Hyperspace that BX MAY go 133 in UNI but I believe only if all onboard componants are in pristine condition. As far as cardsPCI is UNDERCLOCKED at 92 and 124 If AGP runs 1/1 at 100, it would be(comparativly speaking)underclocked 2/3 at 124.

Yeah, 133 is rough on the BX for just Uniprocessor, and my board won't do it even now. Intel's docs say put the design spec at 2 Watts for the BX. I'm guessing it will probably won't put out more than 10 Watts if I can OC it to 133 with dual processors, but even at 10W, it's going to require something pretty heavy duty to keep it from cooking itself. The greenie is definitely too wimpy for that task.

My goal is to see if I can push it so it will just POST higher, and see how far I can raise that bar. Once I've done that, I'm going to go back to the drawing board and design a crazy Peltier system, which I probably won't have money for for a while. I'm about to graduate (June), and I don't know whether I've gotten in to grad school or not. If I don't get in, I'm not going to sink any more money into my box until I've gotten a job (I want to be able to eat for as long as possible in the event that getting a job takes too long).

OH, im a grandpa again since 3/5/05.

Congratulations!

Jeff

[phaedrus]

I'm going to try shorting D1 and D2 this weekend.

Can't wait to hear how that turns out!

is how it turned out. I'm running on my backup motherboard (PII400 system, uniproc, ugh). Shorting the diodes didn't help, in fact, it might have hurt. My board worked under load for about 45 minutes, and after a hard crash, the kernel would consistently oops before running init. I haven't had time to figure out if the board is just being wonky or if I managed to burn out a processor or two.

I clipped the shorting wires, but still haven't had time to throw her back in my box and test her. It might be a week before I get a chance to do so.

Jeff

[purrkur]

Sorry to hear that phaedrus. I do hope that clipping the wires helps you get it up and running again.

I haven't seen the site at Friendtech for a while but wasn't the shorting of diodes a part of a larger picture?? I also didn't trust their guides because they are not stating why they are doing what they are doing.

In any case, I am caps modifying my version 1.0 BP6. There I will only replace all the large capacitors, the 1000µF 10 Volts with the same and the 1500µF 6.3 Volts will be replaced by 2200µF 6.3 Volt jobs. I will also replace the EC10 cap. I bought everything yesterday so I might get around to doing this over the weekend. I intend to run my 366@550MHz Celerons and give up on the P3 jobs until I can find a good fix for my version 1.1 board. That effectively means finding a heatsink that I can use with the broken ZIF socket . My messing about has not only taken too much time but rendered me without a second computer that I really need in my daily work. So 2x550 it is and thats all there is to it. I'll let you know how it turns out with just a change of caps.

[purrkur]

Phaedrus: Any update on your BP6?

I have recapped my BP6. I replaced 25 caps on the board and installed it with my 366MHz Celerons today.

At first bootup I used the 366MHz setup but things were really weird. My CD-RW drive was making strange noises but both BIOS and the OS both saw it. But I couldn't read from it. Even more weird, I had no network either and when trying to revive it, I saw that it complained about not finding the module for the NIC.

I recompiled the kernel with some minor changes, made the necessary adjustments and fired up my little machine again. This time around things went smoother and my CD-RW stopped giving me grief and my network was back without issues.

I returned to BIOS and changed to 550/100 + the turbo setting (for something like 560 MHz) at normal voltage settings. Bad news. The board started up but before it really started booting I got a black screen. I returned to BIOS and changed the voltage to 2.1 volts. This time the board booted and I got things running. I can say that I am really impressed by the fact that Vtt seems very stable. For the most part it stays at 1.49 volts and occasionally dips down to 1.47 but never for long.

After playing around a bit the computer froze. Back to BIOS and turn off the turbo function. Reboot and play around some more. This time it felt more stable so I decided to do a CPU load test. After about 3 minutes the computer froze again.

Hmmm. Not good! New caps and all but the CPU's are not that willing! It might be my 366's though. I have read here that not all will behave at 100 FSB. I wonder what the differences can be?

Back to BIOS, change voltage to 2.2 volts. I really do think that this is a bit extreme but I am willing to try granted that the temperature doesn't shoot through the roof. This time around I ran a CPU load test that took 7 minutes at full load. The system cleared it without any issues.

The box is open still and the CPU's are showing 27.5 and 28 degrees C when idle. After the load test they were at 36 and 37 degrees respectively. I don't find that too offensive so I decided to keep the CPU voltage at 2.2 volts.

One question to people here: I got all memory slots filled. I have read that many have had problems with this. What sort of problems can appear? What is is that happens?

[BCN]

instability happens, but I guess it is because it does not get all the 3.3V it needs. might be wrong

[BCN]

Oh, i have just noticed in your sig, that you are using GF2MX... well, be sure to put spread spectrum at 0.50 in BIOS - it helped me with that card on BP6. though when spread spectrum enabled bios won't report correctly your CPU speed as I recall when FSB range=66-100. if it does not very well, try 1X AGP for that card.

GOOD LUCK!

[purrkur]

Oh, i have just noticed in your sig, that you are using GF2MX... well, be sure to put spread spectrum at 0.50 in BIOS - it helped me with that card on BP6. though when spread spectrum enabled bios won't report correctly your CPU speed as I recall when FSB range=66-100. if it does not very well, try 1X AGP for that card.

Thanks for the tip BCN. I actually have it disabled because this parameter is only used for reducing EMI. Spread spectrum is known to give problems when enabled on overclocked systems.

I haven't had any issues at all with my GF2MX. It has been rock stable both with and without the nVidia drivers so I can't complain. Did you have some sort of a no name card? I bought myself a brand card in order to get better quality (and not the boards that just mimic the reference design from nVidia).

GOOD LUCK!

Thanks!

[Dave Rave]

my main setiqueue server ran on my sq-2 machine, bp6, 768mb ram.
never had instability probs with that.
unless you're trying overclocking with pc66 ram

[BCN]

No, I had leadtek vidcard. oh yeah, It is notorious that Gf2MX causes instability on BP6 yea, I know about that spread spectrum, but that is the only thing that helped me with gf2Mx on BP6, it did not influence omn overclocking at all - it just showed wrong the CPU speed in bios (more than it was, but at the end it just showed wrong - I have assured that with realtime tests like winrar time etc and not sandra or 3dmark