is nichicon hm compatible with ucc kzg?

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

No need.
I'm still using Opera 10/12 and FireFox 24 for the most part, so the default on those browsers is still Flash Player if the system has it (which I do, typically 10.3 or 11.4). That said, even YouTube with HTML isn't that bad. It's other websites (line news and shopping websites) that are the worst offenders. Was looking for an RC car for my nephew yesterday and visited several websites, including ToysRus, Target, and Walmart. Needless to say, most were slower than watching dough rise on my Pentium 4 PCs. Target even refused to work with anything but the newest browser.

I recall YT used to have an option for switching between HTML5 and Flash but they axed it 1 1/2 years ago. I actually tried going back to Firefox 3.6, only to find that Flash was not accessible for me on YT until I used the add-on.

Anyways, I'm getting way off-topic here. The point is, you DO need at least a dual-core CPU nowadays to browse comfortably anywhere.

Well, I find using older versions of Firefox or Opera helps. The newer versions of Firefox are way too RAM-intensive and too much of a CPU hog. But it's all too easy for today's apps to use 100% of any one core. Hence the necessity of multicore CPUs. And don't forget off-topic is my speciality.

Or good copper layout on the board.
The P5GC-MX I have uses only 2 FETs per phase (yes, just two), 3-phases total. No heatsinks. Yet, that motherboard is actually designed to handle every Netburts, including the hottest of the hottest Pentium D and even Pentium Extreme Edition chips, IIRC. The VRM does run warm, but nothing is scorching hot. Rather just the entire board gets warm. ASUS actually designed the VRM pretty well there.

Guess I should have specified. I meant they added heatsinks to the P5W series. Yes, the whole reason that the P4x VRM FETs ran so hot is inadequate copper pads for the FETs. Having three phases probably helps the efficiency and heat output.

But I DO use "junk" thermal compound ... or at least some no-name eBay stuff. It's very runny, but has worked great on everything I've applied it on so far. Doesn't dry out either.
That's not to say that every cheap no-name compound is good. Early in the days of eBay, I tried this brand called "Spectra Cool" - now that was a JUNK thermal compound. It felt like applying chalk and water, and behaved much the same way after a few days of use. Even the unused compound in the tiny syringe got dried up after about a year. Whereas the compound I am using now is maybe 3-4 years old and still fine.

Okay, when I say junk, I meant dried out compounds or a compound which is no longer making good thermal contact between the heatsink and CPU. I suppose "junk" is a highly subjective statement - one man's trash another man's treasure.

Not really. BGA is actually not that hard to work/rework. In fact, you don't need one of those rework machines at all. Just a temperature probe or two, and something powerful enough to heat the bottom of the board. Gas stove/burner is perfect. If you got hot air for the top, you are all set.

I realize that there are methods of reviving BGA chipsets. But none of them are permanent, to my knowledge. For what it's worth, I hate BGA and RoHS and find them both to be nightmares, although I understand why they are mandated. FWIW, I hate the nVidia bumpgate problem much more.

A thru-hole CPU socket would be much harder to rework, IMO, because after you remove the socket, then you will need to clean all of the holes for the new one. And I think that's where it can get tricky.

Yes... assuming a through-hole CPU socket goes bad. Although I suppose the solder joints could go bad over time.

Actually mostly CapXon and Sacon FZ. Teapo comes in a distant third and probably fighting for that spot with Fuhjyyu and Chemicon KZG.
Then again, everyone's junk bin will be dominated depending on what equipment they get to recap. If you do a lot of power supplies, you're bound to be full of Teapo and CapXon. With motherboards, it would probably be Chemicon KZG.

I realize this too. It all depends on when one was repairing boards as well. Had it been 10+ years ago one's motherboard junk bin would have been dominated by the faulty Nichicon HMs and HNs, no doubt. If one is repairing LCDs on a consistent basis, CapXon, Samxon GF, and Samwha will likely dominate their junk bin. I guess my point is, the fact that Teapo dominates your junk bin, means they are certainly junk. I'm not sure how anyone considers them a decent brand, although I suppose we all have a small sample size to judge them from relative to the amount produced on a per-capita basis (the volume of them that are produced).

Fuhjyyu dominates your junk bin? I thought you had a love affair with them!

...Joking, of course. Don't hurt me.

Well, I've also seen thru-hole socket 775 boards, but they are rare. I think the Jetway 775GT1-LOGE was one of them. Got outbid on the same board on eBay.

Well, again, this was simply Intel's specification. Other manufacturers did whatever they wanted, I suppose.

No free-standing FETs/diodes on the P4VXASD2+. All TO-263 parts are soldered to the board properly.

I was making a general statement. Not talking about your board in particular.

I don't know about the mag-amp typically used on the 3.3V rail, but I do know that in forward converter topologies, one half of each output rectifier is used as a free-wheeling diode for the output inductor.

I realize that in the forward converter, because it's a half-wave rectified output, in order to fully discharge the inductor / complete the current path / prevent reverse voltage spikes created by the inductor, the second diode in a double diode package is used for freewheeling. I recall in an old thread you said you saw diodes going between the ground trace and each rail in some PSUs and only understood why after Pete explained the nature of the forward converter. I take it you meant the output rectifiers when you said that, or did you actually see separate diodes for this exact purpose?

And what are you saying about the +3.3V magamp? That there is no freewheeling diode required for it in a forward converter?

You're right, it is much more eventful than watching and LCD monitor power ON/OFF.

My little nephew (4 yr old) is always a bit terrified and amazed at the same time when one of my CRT monitors turns ON, especially when I do it in the dark and you can see all of the arcs on the tube as it charges up. He also finds the static noise heard after the CRT turns off very interesting as well.

When he grows up a little, hopefully I can get him interested in building his own computer... and by that I mean a proper desktop. Maybe with a CRT too . I feel sorry for him right now for only having a tablet to play with at home. These things are absolutely terrible and so limited in anything you can do. Not to mention that the kids constantly get bombarded with advertisements, whether it's watching YouTube or playing a "free" game, or whatever else they do.

YT went down the "tubes", so to speak, once Google acquired them some ten years ago or so.

And yes, LCDs are boooooooooooring in comparison to CRTs.

[momaka]

Re: is nichicon hm compatible with ucc kzg?

You can use special plugins on Firefox to bypass the HTML5 on YouTube and force the use of Flash.

No need.
I'm still using Opera 10/12 and FireFox 24 for the most part, so the default on those browsers is still Flash Player if the system has it (which I do, typically 10.3 or 11.4). That said, even YouTube with HTML isn't that bad. It's other websites (line news and shopping websites) that are the worst offenders. Was looking for an RC car for my nephew yesterday and visited several websites, including ToysRus, Target, and Walmart. Needless to say, most were slower than watching dough rise on my Pentium 4 PCs. Target even refused to work with anything but the newest browser.

Anyways, I'm getting way off-topic here. The point is, you DO need at least a dual-core CPU nowadays to browse comfortably anywhere.

This is why they added VRM heatsinks for the P5x series VRM FETs.

Or good copper layout on the board.
The P5GC-MX I have uses only 2 FETs per phase (yes, just two), 3-phases total. No heatsinks. Yet, that motherboard is actually designed to handle every Netburts, including the hottest of the hottest Pentium D and even Pentium Extreme Edition chips, IIRC. The VRM does run warm, but nothing is scorching hot. Rather just the entire board gets warm. ASUS actually designed the VRM pretty well there.

Well, I could see the "Press-hots" dying if they run around 80˚C (if you use a junk thermal compound!).

But I DO use "junk" thermal compound ... or at least some no-name eBay stuff. It's very runny, but has worked great on everything I've applied it on so far. Doesn't dry out either.
That's not to say that every cheap no-name compound is good. Early in the days of eBay, I tried this brand called "Spectra Cool" - now that was a JUNK thermal compound. It felt like applying chalk and water, and behaved much the same way after a few days of use. Even the unused compound in the tiny syringe got dried up after about a year. Whereas the compound I am using now is maybe 3-4 years old and still fine.

That's not bad. But I recall some Socket 462 CPUs running quite a bit hotter than that (over 50˚C idle and over 60˚C load).

Probably people using Pentium 3 class heatsinks or something. I have a standard all-aluminum heatsink about the size of a standard socket 754/939/AM2 heatsink.

I guess I have an unrealistic phobia when it comes to BGA - but don't we all?

Not really. BGA is actually not that hard to work/rework. In fact, you don't need one of those rework machines at all. Just a temperature probe or two, and something powerful enough to heat the bottom of the board. Gas stove/burner is perfect. If you got hot air for the top, you are all set.

A thru-hole CPU socket would be much harder to rework, IMO, because after you remove the socket, then you will need to clean all of the holes for the new one. And I think that's where it can get tricky.

I recall that your junk bin is dominated by Teapo el cheapo Teapo.

Actually mostly CapXon and Sacon FZ. Teapo comes in a distant third and probably fighting for that spot with Fuhjyyu and Chemicon KZG.
Then again, everyone's junk bin will be dominated depending on what equipment they get to recap. If you do a lot of power supplies, you're bound to be full of Teapo and CapXon. With motherboards, it would probably be Chemicon KZG.

That board powering its CPU from +5V might be why it uses a through-hole socket instead of a BGA socket. I say that because Intel specified the BGA socket after the advent of +12V powered CPUs.

Well, I've also seen thru-hole socket 775 boards, but they are rare. I think the Jetway 775GT1-LOGE was one of them. Got outbid on the same board on eBay.

Ah, so they used diodes for the low side MOSFETs instead of FETs... can anyone say no synchronous buck converter on that board? I'd say this is how they get away with using freestanding FETs in that particular position as the high side FETs are ran quite a bit harder.

No free-standing FETs/diodes on the P4VXASD2+. All TO-263 parts are soldered to the board properly.

Another question: I know the low side "freewheeling" FETs and output capacitors partially exist to provide a place to dump the "energy" from the inductors. Is this true in a SMPS that uses the magamp circuit on the output as well?

I don't know about the mag-amp typically used on the 3.3V rail, but I do know that in forward converter topologies, one half of each output rectifier is used as a free-wheeling diode for the output inductor.

And yes, the bzzzt-hum-click noise CRTs make as they degauss and first start is awesome! So much more thrilling and empowering

You're right, it is much more eventful than watching and LCD monitor power ON/OFF.

My little nephew (4 yr old) is always a bit terrified and amazed at the same time when one of my CRT monitors turns ON, especially when I do it in the dark and you can see all of the arcs on the tube as it charges up. He also finds the static noise heard after the CRT turns off very interesting as well.

When he grows up a little, hopefully I can get him interested in building his own computer... and by that I mean a proper desktop. Maybe with a CRT too . I feel sorry for him right now for only having a tablet to play with at home. These things are absolutely terrible and so limited in anything you can do. Not to mention that the kids constantly get bombarded with advertisements, whether it's watching YouTube or playing a "free" game, or whatever else they do.

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

And yes, the bzzzt-hum-click noise CRTs make as they degauss and first start is awesome! So much more thrilling and empowering than the haunted-house-totally-not-there-weak-lame-ass "noise" LCDs make when they "turn on".

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

It's one of the many reasons I still keep (and use) my CRT monitors - I just can't turn ON a computer and not hear the static electricity build-up of a CRT. The two are like peanut butter and jelly.

Barring possibly dried up capacitors, a well designed CRT will likely outlast any LED monitor or TV. Probably will outlast any CCFL LCD monitor too.

Multi-core CPUs aren't just for people who multi-task. Rather, it's almost a necessity nowadays when it comes to online surfing, especially websites with embedded video. Neither Flash, nor HTML5, nor the various scripts that are often (over)used by websites are kind to single-core CPUs. As an avid Pentium 3/4 and Athlon 64 (single-core) CPU user, I can tell you how much I hate "modern" websites with heavy scripts and HTML5.

You can use special plugins on Firefox to bypass the HTML5 on YouTube and force the use of Flash. But even that is inefficient by comparison to just downloading the MP4 or FLV directly from YT and playing those in VLC media player.

I suspect this causes way more damage to the BGA than heat from the CPU.

Yes, it seems the BGA package suffers just as much from shoddy mounting as it does thermal cycles (or rather, whatever "warps" the board first - it might as well be the same failure mode).

Yeah, I think either my meter wasn't reading the caps right or that KZG and the several Sacon FZ I tested were all were open-circuit, and perhaps that's what the meter shows for O/C.

Well, even KZGs are actually better than Sacon FZ. That's not saying too much, though.

Not that hot, *I think*.
Will know more when I assemble a PC with the motherboard that they were in. The motherboard wasn't discolored or anything, though, so again, I think they didn't run that hot. Not like an HP DC7700 motherboard I got that had quite a bit of discoloration under the CPU, NB, SB, CPU VRM area, and a few other "hot spots". The board was NOT at fault, though - most likely HP did a terrible job with the case cooling. Many of the Rubycon MCZ were bulged and leaking (i.e. fully cooked). The MBZ, no... but I didn't pull them to check their ESR.

Even the Panasonic FLs that "survive" in the position that MCZs didn't were probably dried up. Same goes for the MBZ further down the board. What I noticed about the P4x series Asus boards is that they run VERY hot in the VRM area because of undersized copper pads for the VRM MOSFETs. So if MCZs are used in the VRM output or KZGs - they are bound to die. Even if the VRM high capacitors are KZEs or ZLs, they are bound to dry up from all that heat (same goes for FL in the VRM output as stated afore). This is why they added VRM heatsinks for the P5x series VRM FETs.

Yes. That said, don't wait on those "Press-hots" to die anytime soon - they are some tough fellas. Plus, even though they ran "hot" most of the time, I haven't seen many go past low/mid-60s °C under max load (assuming the right heatsink was used). As it is, they are 10 years old already and still going. Not to say that there aren't any Pentium 3s running either, but many are retired now. Whereas with the Pentium 4's, I still see the odd business using an Opti-Plex GX270 and GX280 for front desk use.

Well, I could see the "Press-hots" dying if they run around 80˚C (if you use a junk thermal compound!).

Also, in my case, the Athlon XP 2500+ I have wasn't running anything excessive either: 47-49°C idle and 53°C under load max. That's in the summer now with around 28-29°C room temperature.

That's not bad. But I recall some Socket 462 CPUs running quite a bit hotter than that (over 50˚C idle and over 60˚C load).

I still doubt it. If they were soldered directly to the board using BGA - then yes I can see that happening (like it does on many modern high-power GPUs). But the contact area between the CPU pads and socket pins is way too tiny to conduct heat quickly. So even if the CPU heats up like crazy very quickly, it will take a loooong time for the CPU socket to heat up to the same temperature, simply because of poor thermal transfer between CPU<->CPU socket.

Well, if it ran 24/7 for many years folding with a very hot Prescott or Pentium D, then after a long time cooling down.... but it takes more than one thermal cyce to break BGA. Maybe if such a task was performed over the course of a century. The reason that I thought it to be possible was that BGA seems to be very fragile, leaded or unleaded, and I've seen too many chipsets die because of it. I guess I have an unrealistic phobia when it comes to BGA - but don't we all?

What, the P4VXASD2+?
Believe it or not, ever single one of the OST and Luxon caps I pulled from it tested good. And that motherboard was powered by a cheap-o PSU (the CyberLink CWT-320ATX I posted on the forums here) most of its life that had only 2 (two) 470 uF caps filtering the 5V rail and no PI coil. The shocker is that the 5V rail is used by that motherboard for powering the CPU .

Heh, I bet if they used Teapo all over the board, not ONE would test good. I recall that your junk bin is dominated by Teapo el cheapo Teapo. At least those 6.3mm-8mm... but I don't think the 10mm-12.5mm+ Teapos are THAT much better, to be honest. I class Teapo right down with the likes of Hermei, Fuhjyyu, and CapXon, honestly. That board powering its CPU from +5V might be why it uses a through-hole socket instead of a BGA socket. I say that because Intel specified the BGA socket after the advent of +12V powered CPUs.

But the icing of the cake is the CPU VRM: oldschool 3-phase single-ended buck design with a KA7500 PWM controller and Schottky diodes instead of lower MOSFETs. Did I hear anyone say inefficient?
Oh yes, it IS. And those caps on the high-side of the VRM were stuck right between them Schottky diodes, typically running around 50-55°C in the winter.

Ah, so they used diodes for the low side MOSFETs instead of FETs... can anyone say no synchronous buck converter on that board? I'd say this is how they get away with using freestanding FETs in that particular position as the high side FETs are ran quite a bit harder.

Another question: I know the low side "freewheeling" FETs and output capacitors partially exist to provide a place to dump the "energy" from the inductors. Is this true in a SMPS that uses the magamp circuit on the output as well? Does the toroid need to "dump" energy into the output capacitors that would otherwise qualify as voltage spikes (thus specifying a minimum capacitance)?

[momaka]

Re: is nichicon hm compatible with ucc kzg?

Also is is worth upgrading to pentium d 945? They are dirt cheap these days (like 2 bucks).

Yes.
The Pentium D 8xx and 9xx series are actually truly dual-core CPUs, whereas the Pentium 4 670 is a single-core with HT. Don't expect miracles, though - the Pentium D 8xx and 9xx CPUs are about as good as a low-end Core 2 Duo, and that is only in some cases. In many cases, the 8xx and 9xx PDs will perform about the same as AMD Athlon 64 X2 3600+ to X2 4000+.

That said, that particular CPU - the Pentium D 945 - is rated for 95 Watts TDP, hitting 115 Watts max under load. That's about 20 Watts lower (both TDP-wise and max.) than your current P4 670, so you should get lower temperatures too.

So overall, I still do think it is worth getting if you really can get it cheap and your motherboard does not support Core 2 Duo CPUs.

Also I love these old school rigs they bring me back to childhood.

Yeah, me too.

It's one of the many reasons I still keep (and use) my CRT monitors - I just can't turn ON a computer and not hear the static electricity build-up of a CRT. The two are like peanut butter and jelly.

The stock heatsink (pic 1).

Ah yes, looks like stock Intel heatsink. Does it have a copper-insert for the contact area over the CPU? I have one like that that does, and my CPU (a Pentium D 830, a 130 Watt TDP CPU) idles in the high-40's and low-50's °C in idle mode. I reckon I'd probably get similar temperatures to yours then under max load.

Also, I feel stupid now asking you for a picture of the heatsink, because I see that there were pictures of it on page 1 of this thread.

depends on what u use it for. if u multi task on it heavily it might be worth it...

Multi-core CPUs aren't just for people who multi-task. Rather, it's almost a necessity nowadays when it comes to online surfing, especially websites with embedded video. Neither Flash, nor HTML5, nor the various scripts that are often (over)used by websites are kind to single-core CPUs. As an avid Pentium 3/4 and Athlon 64 (single-core) CPU user, I can tell you how much I hate "modern" websites with heavy scripts and HTML5.

but i recommend getting a better cpu cooler (third party). the stock heatsink just mangles the board...

I agree 100%.
Either get a better CPU heatsink with a proper CPU backplane bracket or at least make a packplane bracket for the current heatsink you are using. Those stock Intel push-pin heatsinks really do warp the board too much. I suspect this causes way more damage to the BGA than heat from the CPU.

Not to mention how fragile those push-pins get with time and temperature. Bump the computer a little more by accident and they will come undone and leave your CPU to overheat.

Wow, 150Ω.

The electrolyte must be all but gone and/or dried up as it turned from liquid to gas. I would guess that they are completely open (infinite resistance).

Yeah, I think either my meter wasn't reading the caps right or that KZG and the several Sacon FZ I tested were all were open-circuit, and perhaps that's what the meter shows for O/C.

I wonder how hot those KZGs ran.

Not that hot, *I think*.
Will know more when I assemble a PC with the motherboard that they were in. The motherboard wasn't discolored or anything, though, so again, I think they didn't run that hot. Not like an HP DC7700 motherboard I got that had quite a bit of discoloration under the CPU, NB, SB, CPU VRM area, and a few other "hot spots". The board was NOT at fault, though - most likely HP did a terrible job with the case cooling. Many of the Rubycon MCZ were bulged and leaking (i.e. fully cooked). The MBZ, no... but I didn't pull them to check their ESR.

True. But any CPU could eventually die from excessive heat.

Yes. That said, don't wait on those "Press-hots" to die anytime soon - they are some tough fellas. Plus, even though they ran "hot" most of the time, I haven't seen many go past low/mid-60s °C under max load (assuming the right heatsink was used). As it is, they are 10 years old already and still going. Not to say that there aren't any Pentium 3s running either, but many are retired now. Whereas with the Pentium 4's, I still see the odd business using an Opti-Plex GX270 and GX280 for front desk use.

Also, in my case, the Athlon XP 2500+ I have wasn't running anything excessive either: 47-49°C idle and 53°C under load max. That's in the summer now with around 28-29°C room temperature.

I was more referencing the hottest of the Prescotts and Pentium Ds - those could eventually run hot enough to break the BGA socket.

I still doubt it. If they were soldered directly to the board using BGA - then yes I can see that happening (like it does on many modern high-power GPUs). But the contact area between the CPU pads and socket pins is way too tiny to conduct heat quickly. So even if the CPU heats up like crazy very quickly, it will take a loooong time for the CPU socket to heat up to the same temperature, simply because of poor thermal transfer between CPU<->CPU socket.

I recall that particular board was fitted with OST RLX and G-Luxon LZ (and SM) capacitors from the factory - a surefire recipe for premature failure with those stock parts.

What, the P4VXASD2+?
Believe it or not, ever single one of the OST and Luxon caps I pulled from it tested good. And that motherboard was powered by a cheap-o PSU (the CyberLink CWT-320ATX I posted on the forums here) most of its life that had only 2 (two) 470 uF caps filtering the 5V rail and no PI coil. The shocker is that the 5V rail is used by that motherboard for powering the CPU .

But the icing of the cake is the CPU VRM: oldschool 3-phase single-ended buck design with a KA7500 PWM controller and Schottky diodes instead of lower MOSFETs. Did I hear anyone say inefficient?
Oh yes, it IS. And those caps on the high-side of the VRM were stuck right between them Schottky diodes, typically running around 50-55°C in the winter.

[ChaosLegionnaire]

Re: is nichicon hm compatible with ucc kzg?

depends on what u use it for. if u multi task on it heavily it might be worth it... but i recommend getting a better cpu cooler (third party). the stock heatsink just mangles the board...

[nick122]

Re: is nichicon hm compatible with ucc kzg?

The stock heatsink (pic 1). Also is is worth upgrading to pentium d 945? They are dirt cheap these days (like 2 bucks). Also I love these old school rigs they bring me back to childhood.

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

Well, I only had some bulged 6.3V 3300 uF KZG caps to test, and from those, one of them read 150 Ohms on my transistor tester's ESR function. But I think 150 Ohms is the max ESR that the transistor tester can measure anyways, because I tried a bunch of failed Sacon FZ caps, and those also appeared with the same exact ESR of 150 Ohms. When I tested these caps with the auto-test/detect feature on the transistor tester, the transistor tester couldn't even decide what kind of parts they were, often showing "No, unknown, or damaged part". So there is actually a possibility that these caps were open-circuit. But from the 10 or so KZGs I tested, only one showed that 150-Ohm ESR on the trans. tester. The other ones had anywhere between 1 and 4 Ohms ESR and 1/5 to 1/2 of the printed capacitance.

Wow, 150Ω.

The electrolyte must be all but gone and/or dried up as it turned from liquid to gas. I would guess that they are completely open (infinite resistance). I wonder how hot those KZGs ran. Then again, the 3300uF 6.3V and 11.5mm high 820uF 6.3V KZGs seem to be the worst of them.

It would be interesting to see how high ESR KZGs could become without going completely open.

Yeah, Pentium 3 CPUs are wonderful.
AMD Athlon XP CPUs also have very little difference between idle and load. But they don't run that cool, though, due to higher TDP.

True. But any CPU could eventually die from excessive heat.

Well, not really, because even if the CPU runs hot, the thermal contact between the CPU pins and the CPU socket isn't that great.
Also, there are a few motherboards out there with thru-hole CPUs sockets (like my ECS P4VXA-SD2+).

I was more referencing the hottest of the Prescotts and Pentium Ds - those could eventually run hot enough to break the BGA socket. I suppose a BGA socket could survive with a Northwood or even a Willamette, though. I recall that particular board was fitted with OST RLX and G-Luxon LZ (and SM) capacitors from the factory - a surefire recipe for premature failure with those stock parts.

[momaka]

Re: is nichicon hm compatible with ucc kzg?

Probably not that high. I think you would have to really cook a KZG to reach those levels of out-of-spec failure, and it doesn't take much heat (or time on the shelf) to kill KZGs.

Well, I only had some bulged 6.3V 3300 uF KZG caps to test, and from those, one of them read 150 Ohms on my transistor tester's ESR function. But I think 150 Ohms is the max ESR that the transistor tester can measure anyways, because I tried a bunch of failed Sacon FZ caps, and those also appeared with the same exact ESR of 150 Ohms. When I tested these caps with the auto-test/detect feature on the transistor tester, the transistor tester couldn't even decide what kind of parts they were, often showing "No, unknown, or damaged part". So there is actually a possibility that these caps were open-circuit. But from the 10 or so KZGs I tested, only one showed that 150-Ohm ESR on the trans. tester. The other ones had anywhere between 1 and 4 Ohms ESR and 1/5 to 1/2 of the printed capacitance.

And the difference between idle and load temperatures in Pentium III and II CPUs is even less. And they run cool as a cucumber with smaller heatsinks. No wonder they last forever.

Yeah, Pentium 3 CPUs are wonderful.
AMD Athlon XP CPUs also have very little difference between idle and load. But they don't run that cool, though, due to higher TDP.

What worries me a bit about hot-running Pentium 4 CPUs is the fact that they (Intel) specified the usage of BGA solder balls for the actual socket-to-motherboard connection. So even though the CPU itself was PGA, the connection to the PCB itself was BGA. With such hot running Pentium 4 CPUs swinging between room temperature and high temperatures (thermal cycles between the stabilized operating temperature and storage time), it was bound to cause a BGA-related failure over time.

Well, not really, because even if the CPU runs hot, the thermal contact between the CPU pins and the CPU socket isn't that great.
Also, there are a few motherboards out there with thru-hole CPUs sockets (like my ECS P4VXA-SD2+).

Now what is worrisome about those BGA sockets is that they get quite abused by the crappy design of Intel's heatsink retention mechanism - it often put so much pressure on the motherboards that they warp quite badly under the CPU socket. What I do with those is typically just sand/grind down the lobes on the plastic tension handles. That way the heatsink isn't so tight anymore.

but I think something else will probably give out before that does... like a hot running BGA RoHS chipset.

Exactly.
You're much more likely to see the chipset BGA fail than the CPU socket BGA.

Absolutely. That mobo ^ is currently running with p4 670 3.6 ghz prescott cpu. I cant get it under 46c even in bios
. Under load its usually between 65 and 70c and thats pretty hot for a cpu. Preshot

Well, that's a 115 Watt TDP CPU, so those temperatures are about right with a small stock heatsink. Speaking of which, what kind of heatsink are you using and what is your room's ambient temperature? A CPU like that needs a serious heatsink to stay cool.

[nick122]

Re: is nichicon hm compatible with ucc kzg?

For me, the best of the Pentium 4s was the Northwood. They didn't overclock well ("Northwood Sudden Death Syndrome") but they ran very cool and for the time were good performers.

Absolutely. That mobo ^ is currently running with p4 670 3.6 ghz prescott cpu. I cant get it under 46c even in bios
. Under load its usually between 65 and 70c and thats pretty hot for a cpu. Preshot

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

Alright, challenge accepted.
I'll try to measure mine tomorrow and see what I get.

Probably not that high. I think you would have to really cook a KZG to reach those levels of out-of-spec failure, and it doesn't take much heat (or time on the shelf) to kill KZGs.

Whaaaaat!!! Might as well substitute a resistor then.

I can understand using electrolytic caps for 1 uF and above. But below that, I'd just go with polypropylene, polyester, or some other poly- cap and be done with it.

Some of the values specified for the larger can sizes in the datasheets are more reasonable. Like 0.232Ω for a 12.5x20 1000uF 25V KME, or 0.265Ω for a 10x20 1000uF 25V KMG. Although some of the smaller values (100uF 25V and 470uF 10V) are surprisingly high (2.652Ω for a 6.3x11 100uF 25V SMG capacitor). Guess they really want to exaggerate the difference between a GP and low ESR capacitor.

For CPUs, I think up to 100-120 Watts is okay and not that hard to cool. Anything above that, though, will require a big heatsink with several heatpipes.

Maybe even water cooling. But that only works to keep the CPU cool and not the capacitors.

What worries me more is the difference in power consumption we see nowadays on modern CPUs: they can go from just a few Watts of power in idle mode to well over a 100 Watts during full load. That IMO is bound to create a lot of thermal stress on the die, no matter what heatsink you are using. In comparison, a good ol' Prescott doesn't have that great of a difference between idle and full load power consumption. As such, the CPU die isn't that stressed (sure it's running hot all of the time, but that's better than constant thermal cycling). So don't be surprised if you see a Pentium 4 CPU still chugging along 10-20 years from now.

And the difference between idle and load temperatures in Pentium III and II CPUs is even less. And they run cool as a cucumber with smaller heatsinks. No wonder they last forever.

What worries me a bit about hot-running Pentium 4 CPUs is the fact that they (Intel) specified the usage of BGA solder balls for the actual socket-to-motherboard connection. So even though the CPU itself was PGA, the connection to the PCB itself was BGA. With such hot running Pentium 4 CPUs swinging between room temperature and high temperatures (thermal cycles between the stabilized operating temperature and storage time), it was bound to cause a BGA-related failure over time. I think Kaboom had a Pentium 4 Prescott with this issue, with the crappy retention brackets on the Socket 478 heatsink exacerbating the BGA failure. I understand that all modern motherboards attach the socket to the motherboard by way of the BGA package, but I think something else will probably give out before that does... like a hot running BGA RoHS chipset.

For me, the best of the Pentium 4s was the Northwood. They didn't overclock well ("Northwood Sudden Death Syndrome") but they ran very cool and for the time were good performers.

[momaka]

Re: is nichicon hm compatible with ucc kzg?

Although I think the highest recorded ESR reading of failed KZGs (bulging) was around 100Ω by PlainBill.

Alright, challenge accepted.
I'll try to measure mine tomorrow and see what I get.

According to UCC, the highest acceptable ESR value for a 5x11 0.1uF 50V SMG is a whooping 1,989.0Ω ESR.

Whaaaaat!!! Might as well substitute a resistor then.

I can understand using electrolytic caps for 1 uF and above. But below that, I'd just go with polypropylene, polyester, or some other poly- cap and be done with it.

It's more the heat output of high TDP CPUs (and GPUs) that irks me. Although I suppose that's nothing a good ghetto cooling solution can't fix around.

For CPUs, I think up to 100-120 Watts is okay and not that hard to cool. Anything above that, though, will require a big heatsink with several heatpipes.

What worries me more is the difference in power consumption we see nowadays on modern CPUs: they can go from just a few Watts of power in idle mode to well over a 100 Watts during full load. That IMO is bound to create a lot of thermal stress on the die, no matter what heatsink you are using. In comparison, a good ol' Prescott doesn't have that great of a difference between idle and full load power consumption. As such, the CPU die isn't that stressed (sure it's running hot all of the time, but that's better than constant thermal cycling). So don't be surprised if you see a Pentium 4 CPU still chugging along 10-20 years from now.

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

That said, I am curious to see if these failed KZGs caps will pass the spark test - that is, charge each of the caps with a 12-15V source / power supply. Then remove the caps from the source and press both of the cap's terminals on a metal surface. If you see/hear a spark, they are either still working on not completely bad yet. But if you don't see or hear a spark, they are definitely gone. I can tell you from personal experiments that a 16V 1500 uF low ESR cap will definitely produce visible and audible sparks (but nothing loud, though).

If they read half the capacitance and 0.5Ω ESR, that tells me they're definitely goners. Although I think the highest recorded ESR reading of failed KZGs (bulging) was around 100Ω by PlainBill.

That being said, you have a point. If you go look at the UCC (not NCC) datasheets for their general purpose capacitors (85˚C or 105˚C general purpose), they actually list the highest acceptable ESR values at 120Hz and 20˚C. According to UCC, the highest acceptable ESR value for a 5x11 0.1uF 50V SMG is a whooping 1,989.0Ω ESR.

Hey, don't be dissin' the Prescotts! FYI, this post was brought to you by a 2.8 GHz P4 Prescott HT.

Although the single-core Prescott is rated for 89 Watts TDP, its maximum power consumption is only 100 Watts, which isn't that terrible. There are modern CPUs out there nowadays that are much more power-hungry than that - even more than the Pentium D dual-core CPUs (i.e. 2x Prescott dies). And modern GPUs? Don't even get me started on that!

It's more the heat output of high TDP CPUs (and GPUs) that irks me. Although I suppose that's nothing a good ghetto cooling solution can't fix around.

[momaka]

Re: is nichicon hm compatible with ucc kzg?

After recap it runs like charm.

Nice! Looks like another motherboard saved here .
Keep us posted if that really was the solution to your problem (longterm)... although with those KZGs reading nearly half their capacitance and high on ESR, I wouldn't be surprised if they have failed and caused the problems you were having.

That said, I am curious to see if these failed KZGs caps will pass the spark test - that is, charge each of the caps with a 12-15V source / power supply. Then remove the caps from the source and press both of the cap's terminals on a metal surface. If you see/hear a spark, they are either still working on not completely bad yet. But if you don't see or hear a spark, they are definitely gone. I can tell you from personal experiments that a 16V 1500 uF low ESR cap will definitely produce visible and audible sparks (but nothing loud, though).

I suppose this is the one good thing about high TDP CPUs... assuming of course those high TDP CPUs are worth anything. *cough*prescott*cough*

Hey, don't be dissin' the Prescotts! FYI, this post was brought to you by a 2.8 GHz P4 Prescott HT.

Although the single-core Prescott is rated for 89 Watts TDP, its maximum power consumption is only 100 Watts, which isn't that terrible. There are modern CPUs out there nowadays that are much more power-hungry than that - even more than the Pentium D dual-core CPUs (i.e. 2x Prescott dies). And modern GPUs? Don't even get me started on that!

[nick122]

Re: is nichicon hm compatible with ucc kzg?

Ive actually deleted that pic (idk why) so i took a pic of panny fj because its bung looks the same as nichicon hn's and those little caps are pretty hard to unsolder.

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

That last picture looks like a Panasonic capacitor from the bottom, not a Nichicon. A Panasonic FJ, to be meticulous. Three of the four KZGs look like they have slightly bulging bungs (picture 4). Glad to hear you fixed it.

[nick122]

Re: is nichicon hm compatible with ucc kzg?

It started freezing today so i decided to finally recap it. Pulled those li'l suckers out and ONLY 1 was good on esr meter. Other 3 showed 800uf instead of 1500uf and 0.5 ohm resistance. After recap it runs like charm. Can u tell which one is good and which one is bad? (pics 2 and 3 also sorry for bad focus on pic 3). Pic 5 hn from the bottom.

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

I guess perhaps this is because CPUs have a low static resistance, thus keeping the power load on the CPU VRM sufficiently high regardless of CPU load. So there's always somewhere for the energy in the inductors to go, even if the caps went bad.

Yup, that's Ohm's Law for you. I suppose this is the one good thing about high TDP CPUs... assuming of course those high TDP CPUs are worth anything. *cough*prescott*cough*

RAM, on the other hand, ofter doesn't use much power and doesn't support much of a voltage variation. So that could well be the reason why RAM is more likely to die from a bad FET on the buck regulator circuit.

Or from bad caps. I guess it wouldn't be such a big deal if it was Qimonda/Infineon or Buffalo RAM, or Elixir or Corsair RAM. Best to put crap brands out of their misery I suppose.

[momaka]

Re: is nichicon hm compatible with ucc kzg?

But CPUs are tough bastards anyway.

Yes, that's a good way to put it.

Had a Core 2 Duo withstand MOSFETs shorting on a Intel DG33TL-equivalent board multiple times. One of the phase's MOSFET drivers must have been bad (I say *must*, because they all appeared okay on my multimeter). But unfortunately, it cost me a lot of MOSFETs to realize this. Each time a FET shorted, the PSU would enter into short-circuit protection. Surprisingly, the CPU is still alive and working. I guess perhaps this is because CPUs have a low static resistance, thus keeping the power load on the CPU VRM sufficiently high regardless of CPU load. So there's always somewhere for the energy in the inductors to go, even if the caps went bad. RAM, on the other hand, ofter doesn't use much power and doesn't support much of a voltage variation. So that could well be the reason why RAM is more likely to die from a bad FET on the buck regulator circuit.

[Wester547]

Re: is nichicon hm compatible with ucc kzg?

I agree 100%.


Well, this is less likely to happen to the CPU VRM, since it usually has a more advanced chip than your typical PWM controller. But I agree that capacitors on the input of a buck regulator indeed should not be deemed less important than the output.
As you know (I'm sure I mentioned it more than once ), I've had a router die from bad caps on its buck regulator input. And that's just once instance off the top of my head. Had a few other similar cases in other devices.

True. The CPU VRM should shut down in the case of a short circuit. Also, PSUs with good short circuit protection should catch that relatively quickly as well and shut down in an instant (having good overcurrent protection would help as well). It's true that CPU PWM ICs are more advanced than PWM ICs for simpler chips. I suppose it's the same for GPUs - not as likely to die as quickly as less advanced chips from bad caps on the input or output of a buck regulator. Of course that's not talking about the RAM... the RAM is much more prone to failure when buck regulator caps go bad. But CPUs are tough bastards anyway. I don't think I've ever seen one die from anything other than broken pins, bad BGA contacts, severe overvoltage, failed overclocking attempts, or from overheating for too long (rather, that's referencing very early Athlons which ran very hot). You are more likely to blow a MOSFET than a CPU when the caps go bad (since when the ESR skyrockets, the duty cycle of the MOSFETs increases until they burn out, and same goes for the coils).

The risk with capacitors on a buck regulator going bad is plummeted capacitance which ultimately results in fried chips (by way of voltage spikes) since the energy released from the inductors has nowhere to go (or rather, there's nothing standing between that energy and the ICs).