VOGONS


ASUS K7M troubles

Topic actions

First post, by HanJammer

User metadata
Rank Oldbie
Rank
Oldbie

I have K7M which I owned since it was new. It have all original accessories, box and so on for it.
It was bought in ~99 or 2000 and It was used in several machines well upto 2013 I think. And since then it was in storage. Problem is that it's no longer stable now.
Well it's stable until it gets hot (chokes on the VRC get very very hot, but it always was like this, mosfets don't get too hot, some caps surrounding the VRC get hot because of the chokes). When I let it cool down for a few minutes it again is stable and once gets hot - freezes. Both Athlons I tested it with are fine and do work fine on my other Slot A board. I havn't tested adding some temporary radiator/cooler to the chokes yet. Caps do look fine and are made by good brands (Rubycon among others).

Do any of you guys have some experience with more detailed diagnosis of such problem?

New items (October/November 2022) -> My Items for Sale
I8v8PGb.jpg

Reply 1 of 24, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

My bet is the caps have dried out and will need to be replaced.

What happens if you point a fan towards those components that are getting hot?

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 2 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member

Indeed, can be capacitors. I have two dead "waiting to be repaired" ASUS motherboard of the same era (2000-2003). A A7N8X Deluxe that have a track who literally melt because of a dead cap. And another CUW-AM who got abnormally hot on the CPU area for a few days before stopping to work. It has problems on half of its caps . From what I can tell, if a capacitor is dead or almost dead, no or very few current pass through, and as the CPU needs the same amount of current to work, it will pump it on other tracks leading components on those tracks to have higher current than they expect and have to dissipate a lot of heat. All of that is not good and that leads to immediate problems such as freeze, instability but also to long term problems like chip failure. Best to test every cap and replace the defective ones. I'm not for a full recap as some can be in very good condition despite all this years.
Maybe I am wrong, but if anyone have a better understanding or can clarify things, I'm interested to know.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 3 of 24, by HanJammer

User metadata
Rank Oldbie
Rank
Oldbie
cyclone3d wrote on 2021-09-13, 18:17:

My bet is the caps have dried out and will need to be replaced.

What happens if you point a fan towards those components that are getting hot?

Like I said I havn't tested that yet. My guess is it will help abit... My first guess was replacing all the caps although I would like the board to be as original as possible and I can't find matching Rubycon caps. 🙁
Also I would need to desolder all the caps to test them as they are on a common rail (acting just like one huge cap). Especially those 4 sandwitched between chokes (which get literally so hot I can't touch them) and CPU (Slot A Athlons are very hot too) can be the suspect...

Attachments

New items (October/November 2022) -> My Items for Sale
I8v8PGb.jpg

Reply 4 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member

To test a capacitor, it must be removed, yes. I remember Socket A Athlon and Athlon XP are getting hot, that's normal. But as they have no overheating protection they need a good cooling.
For Slot A, I don't remember them getting hot. I have an Slot A Athlon 700 MHz in a machine. It have a radiator with a little fan (CoolerMaster) and the radiator can be touched without problem.
Do you have a good cooling for the CPU?
As the CPU drains the power, maybe it is not related to the capacitors. But the problem can be with a voltage regulator and not the cap. That is why I recommend to test the capacitors. If they are good, it's before them.
Without any measurement only assumptions can be made.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 5 of 24, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++
HanJammer wrote on 2021-09-13, 19:32:
cyclone3d wrote on 2021-09-13, 18:17:

My bet is the caps have dried out and will need to be replaced.

What happens if you point a fan towards those components that are getting hot?

Like I said I havn't tested that yet. My guess is it will help abit... My first guess was replacing all the caps although I would like the board to be as original as possible and I can't find matching Rubycon caps. 🙁
Also I would need to desolder all the caps to test them as they are on a common rail (acting just like one huge cap). Especially those 4 sandwitched between chokes (which get literally so hot I can't touch them) and CPU (Slot A Athlons are very hot too) can be the suspect...

I remember Slot-A Athlons running super hot as well. I don't remember the caps or chokes getting hot though and I was running a K7M at one point back in the day. Pretty sure I have a K7M in my current collection as well.

However, I modified an OEM Socket 939 low profile heatpipe cooler and mounted it to the heat plate on one of my Slot-A CPUs and it doesn't get warm at all.

There is also a big difference in the heat output between the original Slot-A Athlon chips and the Slot-A Thunderbird chips.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 6 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member

To remove a capacitor as they pass through the motherboard, beware to not damage the gates (the little cylinder that makes the link between the different layers of the motherboard). I recommend using a desoldering gun. Easy and fast.
To remove SMD like the voltage regulators, it require more experience as they are often around small resistors and capacitors. A mistake is quickly done, and replacing them is not easy.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 7 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member
cyclone3d wrote on 2021-09-13, 20:25:

There is also a big difference in the heat output between the original Slot-A Athlon chips and the Slot-A Thunderbird chips.

I had Thunderbird Slot A. I had Athlon 550 and 700, and both of them never get hot.

Last edited by Datadrainer on 2021-09-13, 20:36. Edited 1 time in total.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 8 of 24, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++
Datadrainer wrote on 2021-09-13, 20:29:

To remove a capacitor as they pass through the motherboard, beware to not damage the gates (the little cylinder that makes the link between the different layers of the motherboard). I recommend using a desoldering gun. Easy and fast.
To remove SMD like the voltage regulators, it require more experience as they are often around small resistors and capacitors. A mistake is quickly done, and replacing them is not easy.

What desoldering gun do you use? I have a really old one and it really struggles to melt the solder on holes that are on a ground plane.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 9 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member
cyclone3d wrote on 2021-09-13, 20:33:
Datadrainer wrote on 2021-09-13, 20:29:

To remove a capacitor as they pass through the motherboard, beware to not damage the gates (the little cylinder that makes the link between the different layers of the motherboard). I recommend using a desoldering gun. Easy and fast.
To remove SMD like the voltage regulators, it require more experience as they are often around small resistors and capacitors. A mistake is quickly done, and replacing them is not easy.

What desoldering gun do you use? I have a really old one and it really struggles to melt the solder on holes that are on a ground plane.

I'm using a GJ S-998P from Banggood for 2 years now. Great deal. Robust and easy to clean. But they are 220 V only with a Chinese plug. As I live in Europe it is not a problem, I just changed the plug.
It have a adjustable thermostat up to 480° C and I never had problem to remove anything even on ground plane. Sometime, for huge solder, tin must added though to make things easier.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 10 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member

I just checked because I was not sure but the Athlon 550 MHz is a first gen, not a Thunderbird. My mistake. But it never got very hot. If @cyclone3d says Slot A Athlon are heating up a lot, maybe that is the ones with high frequency. I don't know, never had one.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 11 of 24, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++
Datadrainer wrote on 2021-09-13, 20:55:

I just checked because I was not sure but the Athlon 550 MHz is a first gen, not a Thunderbird. My mistake. But it never got very hot. If @cyclone3d says Slot A Athlon are heating up a lot, maybe that is the ones with high frequency. I don't know, never had one.

I think the slowest I had was sold as a 500Mhz but had a 650Mhz core. Back then, most of them had higher rates cores than they were sold as. The ones sold as 650Mhz were often 800 or 850Mhz rated.

Pop off the heatspreader plate, check the engravings on the CPU core and then move the 0-ohm resistors to match the rated speed. The only caveat was that sometimes you got slower L2 cache than what was normally on the CPUs that were sold as that speed. And if I remember correctly, some of the slower ones actually had faster rated L2 cache as the divider for the L2 was lower than it normally was on the higher rated CPUs.

So much fun messing with those CPUs back in the day.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 12 of 24, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

I had/have a K7 500 (early one, so the chip is actually marked 500MHz, think it will go to 550 if I push the voltage up a bit) on an Abit KA7 motherboard (originally on a MSI 6167 board that I don't remember at all). That had most of the big caps (>1000uF) go bad around 2003ish, with the rubber seals blowing out. I'm not sure how to diagnose the problem properly if there's no obvious visible faults, probably get an oscilloscope and measure the ripple voltage, and remove a few of the capacitors to check their capacitance and ESR with a component tester.

Symptoms at the time were gradually increasing instability, until eventually it wouldn't boot. When I finally had a look I also found that the PCB under on of the power FETs had darkened from getting too hot. At the time I changed out the all the capacitors that look obviously broken, and any others of the same value, and it's been fine since. The only major snags I had were unsoldering from the ground plane with the gas iron I had then, and trying to find capacitors with the same spec that would fit in the space. There weren't any, so the diameter of the replacements was a bit too big, so the placement got a bit, er, creative.

I don't remember the heatsink on the CPU getting all that hot, but that would have been helped by the heat transfer goop they used on the cache chips not actually connecting the chips to the heatsink (discovered when I finally cracked open the CPU case).

Datadrainer, AIUI, the caps will have a rated voltage, temperature and lifespan. Often the lifespan at their limits isn't very long (couple of thousand hours). The caps are meant to reduce ripple in the voltage by shunting the ripple current to/from ground, and ideally there is no resistance to that ripple current, so there's no actual power burnt in the cap. But there is always some resistance. As the capacitor starts to fail then the resistance will increase, which means there will be more power burnt in the cap (so it gets hotter) and higher ripple in the voltage (so it'll run closer to it's voltage limit). Both of those things cause more current to flow through the other capacitors, which pushes those to fail faster as well. Eventually the ripple will get big enough that it starts causing glitches in the CPU, and probably causes the power FETs to switch more often, so they get hotter as well. So a slight increase in the resistance in a few caps can cause a load of escalating problems.

What are the capacitance, voltage and diameters of the caps near the CPU slot on this board? There must be some pretty close equivalents.

Reply 13 of 24, by HanJammer

User metadata
Rank Oldbie
Rank
Oldbie
Datadrainer wrote on 2021-09-13, 20:29:

To remove a capacitor as they pass through the motherboard, beware to not damage the gates (the little cylinder that makes the link between the different layers of the motherboard). I recommend using a desoldering gun. Easy and fast.
To remove SMD like the voltage regulators, it require more experience as they are often around small resistors and capacitors. A mistake is quickly done, and replacing them is not easy.

Well I do have experience with that and proper equipment (soldering station, desoldering gun and so on) although I pretty much hate working with anything more modern than 486 boards with all those huge ground planes between layers which suck the heat and Pb-free solder... Part of the reason why recapping it is a last resort for me...

Datadrainer wrote on 2021-09-13, 20:32:

I had Thunderbird Slot A. I had Athlon 550 and 700, and both of them never get hot.

I have 600Mhz since it was new and it always got moderately hot (although I never found any decent cooler for it) and 1000Mhz Orion which gets crazy hot.

New items (October/November 2022) -> My Items for Sale
I8v8PGb.jpg

Reply 14 of 24, by HanJammer

User metadata
Rank Oldbie
Rank
Oldbie
snufkin wrote on 2021-09-13, 21:44:

What are the capacitance, voltage and diameters of the caps near the CPU slot on this board? There must be some pretty close equivalents.

Here: https://drive.google.com/drive/folders/1i2ULj … 2sJ?usp=sharing

I think I bought almost all of the low ESR caps other brands though (still quality like Panasonic. I would really prefer them to be OE though...

Also it's noteworthy that the chokes whine almost like my girlfriend when I bring more vintage PCs to our apartment 😉 and it was never and issue when I used it before...

New items (October/November 2022) -> My Items for Sale
I8v8PGb.jpg

Reply 15 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member
cyclone3d wrote on 2021-09-13, 21:06:

I think the slowest I had was sold as a 500Mhz but had a 650Mhz core. Back then, most of them had higher rates cores than they were sold as. The ones sold as 650Mhz were often 800 or 850Mhz rated.

Pop off the heatspreader plate, check the engravings on the CPU core and then move the 0-ohm resistors to match the rated speed. The only caveat was that sometimes you got slower L2 cache than what was normally on the CPUs that were sold as that speed. And if I remember correctly, some of the slower ones actually had faster rated L2 cache as the divider for the L2 was lower than it normally was on the higher rated CPUs.

So much fun messing with those CPUs back in the day.

I read that a long time ago! I think it was possible on Socket A too. Never practiced that myself though.

HanJammer wrote on 2021-09-13, 22:30:

Part of the reason why recapping it is a last resort for me...

I understand perfectly but I don't know how to diagnose bad components efficiently, especially bad caps put in a row like on your motherboard in another way.

snufkin wrote on 2021-09-13, 21:44:

Datadrainer, AIUI, the caps will have a rated voltage, temperature and lifespan. Often the lifespan at their limits isn't very long (couple of thousand hours). The caps are meant to reduce ripple in the voltage by shunting the ripple current to/from ground, and ideally there is no resistance to that ripple current, so there's no actual power burnt in the cap. But there is always some resistance. As the capacitor starts to fail then the resistance will increase, which means there will be more power burnt in the cap (so it gets hotter) and higher ripple in the voltage (so it'll run closer to it's voltage limit). Both of those things cause more current to flow through the other capacitors, which pushes those to fail faster as well. Eventually the ripple will get big enough that it starts causing glitches in the CPU, and probably causes the power FETs to switch more often, so they get hotter as well. So a slight increase in the resistance in a few caps can cause a load of escalating problems.

Thank you for the explanation @snufkin. MOSFET are part of the VRM. On HanJammer MB, we can see five of them if I'm right. That was what happened on my ASUS CUW-AM, all the area got hot.
I wonder if it is possible to detect at least a group of caps with one or more faulty just by identifying the traces and measuring the current that flows avoiding to desolder anything or making complex oscilloscope measurement.
That is what HanJammer wants. Having a way to at least find areas where things are bad without removing anything. And as every components are visually good, maybe that is a way to get clues.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 16 of 24, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

If you want to keep the original part numbers then I think Rubycon YXG 1500uF 6.3V caps are available: https://uk.farnell.com/rubycon/6-3yxg1500mefc … v-20/dp/2346170 although they might have changed the case size (those ones are 10mm diameter and 20mm high).
Rubycon ZL 4700uF 6.3V https://uk.farnell.com/rubycon/6-3zl4700mefc1 … v-20/dp/1144675, 12.5mm diameter, 30mm high.

Not sure what availability is like where you are, but it looks like they're still made.

I quite like the Panasonic FS series, slightly lower ESR, smaller case sizes (1500uF is 8mm x 15mm, 4700uF 12.5mm x 20mm) and decent lifespan.

On the whining, could be that higher voltage ripple is causing higher currents, causing larger magnetic fields in the chokes, causing more movement, which you can then hear. Maybe.

Problem with testing capacitors in circuit is that they're all connected together. So if one gets hot because more current is flowing through it, that could be because another one has gone high resistance and has less current flowing through it. So the one that gets hot might not actually be the problem. I think there are some devices that claim to measure individual capacitance with the component in circuit, but I've never used one of those. I'd probably just go with replacing all the big capacitors in one go, but others may have better ideas on how to test the existing ones.

Reply 17 of 24, by HanJammer

User metadata
Rank Oldbie
Rank
Oldbie
Datadrainer wrote on 2021-09-14, 08:12:

Thank you for the explanation @snufkin. MOSFET are part of the VRM. On HanJammer MB, we can see five of them if I'm right. That was what happened on my ASUS CUW-AM, all the area got hot.
I wonder if it is possible to detect at least a group of caps with one or more faulty just by identifying the traces and measuring the current that flows avoiding to desolder anything or making complex oscilloscope measurement.
That is what HanJammer wants. Having a way to at least find areas where things are bad without removing anything. And as every components are visually good, maybe that is a way to get clues.

Mosfets just get warm, but that's all. Nothing unusally hot. I'm afraid I can't do much without desoldering the caps as they are connected in parallel. I loved when the motherboards were equipped with actual VRM (Voltage Regulator Modules) instead of VRCs (Voltage Regulator Circuits) like this board or most modern motherboards - in case of failure you could just replace the module, or service it outside the motherboards... I doubt they will ever bring VRMs back to consumer-grade products (in high end servers they are usually still present in one form or another).

snufkin wrote on 2021-09-14, 09:23:

If you want to keep the original part numbers then I think Rubycon YXG 1500uF 6.3V caps are available: https://uk.farnell.com/rubycon/6-3yxg1500mefc … v-20/dp/2346170 although they might have changed the case size (those ones are 10mm diameter and 20mm high).
Rubycon ZL 4700uF 6.3V https://uk.farnell.com/rubycon/6-3zl4700mefc1 … v-20/dp/1144675, 12.5mm diameter, 30mm high.

Thanks. I usually don't order passives out of the country because the shipping cost gets higher than the cost of the parts, but in this case I might do an exception. I looked in the Mouser some time ago, and I found these caps but they were on backorder without any date when they become available.

snufkin wrote on 2021-09-14, 09:23:

On the whining, could be that higher voltage ripple is causing higher currents, causing larger magnetic fields in the chokes, causing more movement, which you can then hear. Maybe.

Yeah, I thought so as well that this may be the case. Something odd is going on with this motherboards.

snufkin wrote on 2021-09-14, 09:23:

Problem with testing capacitors in circuit is that they're all connected together. So if one gets hot because more current is flowing through it, that could be because another one has gone high resistance and has less current flowing through it. So the one that gets hot might not actually be the problem. I think there are some devices that claim to measure individual capacitance with the component in circuit, but I've never used one of those. I'd probably just go with replacing all the big capacitors in one go, but others may have better ideas on how to test the existing ones.

I remember I measured the total capacitance of the caps in parallel and it was more or less what it should be (although it was when the board was cold). Another test I can do is provide some cooling to the caps and chokes and see if it changes something (I guess it will - the 'cold' motherboard works fine upto 5-6 minutes, after that it hangs, when I turn on the power for 1 minute - I get another 1 minute of stability and so on so it's definitely related to the heat. I guess replacing the first four 1500uF caps which are sandwitched together between the chokes and the CPU would be a good starting point.

New items (October/November 2022) -> My Items for Sale
I8v8PGb.jpg

Reply 18 of 24, by Datadrainer

User metadata
Rank Member
Rank
Member

@HanJammer. Yep Would means VRC, but for me, except one is embedded, it's the same. But yes, for latter 486 till Pentium Pro, plugable voltage regulation modules was great.

I'm asking myself a question for a long time now. Could it be done to use tulip socket to insert caps? Or is something wrong with that. That question came following a thought about having a way to easily replace all components prone to fail on old hardware. I got capacitor who inflate to the point they broke the trace where they where soldered. With socket they can just pop out and they are easier to check.

Last edited by Datadrainer on 2021-09-14, 12:34. Edited 1 time in total.

Knowing things is great. Understanding things is better. Creating things is even better.

Reply 19 of 24, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie
HanJammer wrote on 2021-09-14, 11:47:
snufkin wrote on 2021-09-14, 09:23:

If you want to keep the original part numbers then I think Rubycon YXG 1500uF 6.3V caps are available: https://uk.farnell.com/rubycon/6-3yxg1500mefc … v-20/dp/2346170 although they might have changed the case size (those ones are 10mm diameter and 20mm high).
Rubycon ZL 4700uF 6.3V https://uk.farnell.com/rubycon/6-3zl4700mefc1 … v-20/dp/1144675, 12.5mm diameter, 30mm high.

Thanks. I usually don't order passives out of the country because the shipping cost gets higher than the cost of the parts, but in this case I might do an exception. I looked in the Mouser some time ago, and I found these caps but they were on backorder without any date when they become available.

Don't know if it's just a language change and doesn't change the costs, but Farnell have a Polish website as well:
https://pl.farnell.com/rubycon/6-3yxg1500mefc … v-20/dp/2346170
https://pl.farnell.com/rubycon/6-3zl4700mefc1 … v-20/dp/1144675