IBMFan wrote on 2021-05-31, 19:07:

Okay, I accept the points. Also, do repair technicians replace everything to fix a defective part or do they find the defective part and replace it?
I know it's a hobby so to each their own. I still believe in diagnostics first.

Diagnostics can help if there is a chance of parts not failing. In case of these capacitors that chance is 0%.

evasive wrote on 2021-06-02, 06:22:

Diagnostics can help if there is a chance of parts not failing. In case of these capacitors that chance is 0%.

It's true with a counterfeit product but not true with an entire brand.
The "diagnostics" was made by the seller (board not booting) so that could mean anything: bad HDD, bad OS, bad cable, bad board etc.
Teapo = bad, Nichicon = good is not how people perform diagnostics especially knowing that Nichicon made the infamous HM/HN/HZ/KZG series and Rubycon was responsible for the mass-failures of the MCZ series. Brand reputation is misleading and that's where you start looking for the real culprit.
It's like calling the TV guy to fix your TV and he charged you $100 extra for replacing a bunch of otherwise fine parts because they COULD go bad in the future. Most people would be livid.
Many of us sat behind the TV guy when he fixed the living room set and watched him probing and measuring the big panels to find the 1-2 parts that caused the malfunction.
We are talking about 15-25 years old boards so you might as well replace all caps on all boards regardless of brand and actual condition. Most of them are way beyond their life expectancy anyway. Panasonic, Rubycon, Teapo, Fuhjyyu, OST, doesn't matter. They are old and probably bad, replace them all. It's essentially the same thing.
I know that the 2 groups (replace all vs don't fix if it ain't broken) won't ever see eye-to-eye and they don't have to.
I doubt many people would start replacing 30+ caps anyway especially with basic tools.

IBMFan wrote on 2021-06-02, 07:17:

It's true with a counterfeit product but not true with an entire brand. The "diagnostics" was made by the seller (board not boot […]
Show full quote

evasive wrote on 2021-06-02, 06:22:

Diagnostics can help if there is a chance of parts not failing. In case of these capacitors that chance is 0%.

It's true with a counterfeit product but not true with an entire brand.
The "diagnostics" was made by the seller (board not booting) so that could mean anything: bad HDD, bad OS, bad cable, bad board etc.
Teapo = bad, Nichicon = good is not how people perform diagnostics especially knowing that Nichicon made the infamous HM/HN/HZ/KZG series and Rubycon was responsible for the mass-failures of the MCZ series. Brand reputation is misleading and that's where you start looking for the real culprit.
It's like calling the TV guy to fix your TV and he charged you $100 extra for replacing a bunch of otherwise fine parts because they COULD go bad in the future. Most people would be livid.
Many of us sat behind the TV guy when he fixed the living room set and watched him probing and measuring the big panels to find the 1-2 parts that caused the malfunction.
We are talking about 15-25 years old boards so you might as well replace all caps on all boards regardless of brand and actual condition. Most of them are way beyond their life expectancy anyway. Panasonic, Rubycon, Teapo, Fuhjyyu, OST, doesn't matter. They are old and probably bad, replace them all. It's essentially the same thing.
I know that the 2 groups (replace all vs don't fix if it ain't broken) won't ever see eye-to-eye and they don't have to.
I doubt many people would start replacing 30+ caps anyway especially with basic tools.

You make good points of course.
And I know my thread title makes it sound like Teapo's are the root of all bad things happening to old hardware =)
Tho the times I just replaced obviously bad caps, I usually have to go back and redo the rest later.
And this KR7A might not be an exception.
Because I'm having issues.
Thing is, its been absolutely rock solid in all my installing, gaming and bench marking.
But Memtest86 and Memtest86+ both throw tons of errors when doing the "Block Move" tests.
But ONLY when using a T-Bred CPU... with a Palomino its just fine...
Even tested a Thorton just for the lols, its not recognized ofc but ran just fine and also had the same mem errors.
Tried with the latest beta BIOS "EJ", and a few non-beta versions before that, no change.

Now I haven't had time to test all various mem settings yet, so there might be something it doesn't like.
I have tested many different RAM sticks tho, and they all give the same errors.

In my mind theres probably a few options here:
1. The KR7A might not be fully compatible with (faster versions?) T-Bred's.
2. The pretty Rubycons and Nichicons also have to go.

But the next step is to bring out my older board and do the same tests there just to make sure.

IBMFan wrote on 2021-05-31, 19:07:

Okay, I accept the points. Also, do repair technicians replace everything to fix a defective part or do they find the defective part and replace it?

Depends on technician, device and defect.
In many cases replacing a component is more economical than a repair. The time it takes to properly diagnose and repair a part is often just not worth it. Plus giving a warranty on a repair oof a previously failed part isn't without risk for a business.
If you have a broken ATX PSU in a somewhat recent computer I can hardly imagine anyone would repair the unit instead of replacing it as a whole. (Otoh. if it's just a noisy fan in an otherwise good 1000W PSU it would make sense to replace the fan.)
But if you have something highly integrated like a notebook or phone pcb where you can not just modularly change a mainboard or CPU with relative ease component level repairs can make a lot more sense. Especially if the device is still relatively new and valuable or if it's just hard to get replacements (i.e. CRT PCBs).
I would consider preemptively replacing parts that are known to degrade (like doing a full recap on a functional 20+y/o board) to be a remanufacture/refurbishment or overhaul rather than a repair though.

Btw. I really enjoy that we can have different view points and still have a civil discussion without having to completely agree with one another.👍 This is really not a given in most online communities!

bofh.fromhell wrote on 2021-06-02, 10:50:

And this KR7A might not be an exception. Because I'm having issues. Thing is, its been absolutely rock solid in all my installin […]
Show full quote

And this KR7A might not be an exception.
Because I'm having issues.
Thing is, its been absolutely rock solid in all my installing, gaming and bench marking.
But Memtest86 and Memtest86+ both throw tons of errors when doing the "Block Move" tests.
But ONLY when using a T-Bred CPU... with a Palomino its just fine...
Even tested a Thorton just for the lols, its not recognized ofc but ran just fine and also had the same mem errors.
Tried with the latest beta BIOS "EJ", and a few non-beta versions before that, no change.

Now I haven't had time to test all various mem settings yet, so there might be something it doesn't like.
I have tested many different RAM sticks tho, and they all give the same errors.

In my mind theres probably a few options here:
1. The KR7A might not be fully compatible with (faster versions?) T-Bred's.
2. The pretty Rubycons and Nichicons also have to go.

But the next step is to bring out my older board and do the same tests there just to make sure.

Back in the day the KR7A had some issues with multipliers higher than 12.5, seemed to leave the pin floating or something which led top random behaviour of any CPU that had a multi of 13 or higher. There were bridge and socket tricks to modify.

Further testing with my other board revealed that they both behave identically.
Dropped in a Barton M , which gets identified correctly which was neat.
Tried lots of different multipliers, no change.
Did a lot more testing with different mem settings, no changes in results.
Lowering mem frequency to 100 MHz made no difference.
Even dug up an older Memtest86+ (v1.7) and that also gave the same errors during the Block Move test.

Weird that the mem test errors doesn't manifest themself as instability in any way.

I had some errors with memtest86+ on a Abit KT7A that manifested at the 512 MB boundary. I'm not sure which test it was, but in normal usage I never had any issue. You could try to do a cpu/memory intensive task such as compiling a large piece of software like the Linux kernel; if it runs correctly, you are probably fine. In short, do not trust memtest86+ too much. Anyway afterwards I decided to only use 512 MB with one stick, and the memtest86 failure disappeared.

If memtest generates an error, you have a problem, even if you don't notice it in other use cases. It doesn't ever produce spurious errors as far as I know

Some sticks don't play nice together, I had to go through a literal shoe box full of RAM to get 4 sticks that passed memtest in a Neo2 a few years back.

I was playing with an nforce 2 board just recently. Epox 8DRA+ or something, having the same errors in memtest86+ test 5 - blockmove.

Definitely can confirm that some sticks just don't play nice together. You need to swap them out into you get something that works.

Onto my thoughts on recapping motherboards.

If you are going to start recapping a motherboard, you should almost definitely change all the caps on the same series, and preferably all capacitors 470uF and larger.

More than once, I've had to go back to a board multiple times to replace further capacitors months after the initial repair because other capacitors on the board have decided to spill their guts. If you want a recap job to actually stick, and not cause issues down the line, do the job in full at the beginning.

Onto how I do it:

First of all, you need a good soldering iron. Something like a KSGER T12 soldering iron or one of the other cheap soldering irons youtube people fawn over, they all seem to be inexpensive and do a good enough job. I use the D12 tip (and D16 too somethings). I also use solder braid (wick). I have a desoldering station, but don't use it for recapping because it can be rough on the boards, to easy to damage nearby traces compared with a soldering iron.

I normally work with the soldering iron set to 400C, but will boost it up to 450C when working on the well heatsinked pads around the CPU vrms. I add some fresh leaded solder to each leg first of all, this lowers the melting temperature, wets the old solder, and gives your soldering iron more direct contact with the solder joint you are trying to remove. I then heat each leg up and walk the old capacitor out of the board, by pulling on it from the other side. Then I will attempt to clear the old solder out using solder wick, attacking from both sides if I have to. I am usually able to clear at least one hole, but I don't waste too much time on this if I can't get the heavily heatsinked pads clear.

Puttng in the new cap, I have cleared both holes, I pretrim the legs down to almost the correct size, and put it in and then solder it down. I then trim the legs to the correct length, and reflow each joint. Cutting the legs will put stress on the solder joint, so reflowing after cutting is just making sure I didn't crack the joint or something while cutting the leg. If I didn't get both holes cleared, I trim down both legs, but make the leg that goes into the cleared hole a little longer to act as a guide. I then add some solder back to the hole I wasn't able to clear, and walk the new cap back into the board. The cut legs down and reflow both solder joints.

I only replace a few capacitors at a time, keeping an eye on the polarity of each capacitor and the marking on the board (be aware that there could be errors in the screen print of the board, check/document capacitor orientation before you replace any) Before powering up the board, I check I have resoldered all the capacitors properly, give it a bit of a clean with isopropyl alcohol, and double check the polarity of the capacitors.

canthearu wrote on 2021-06-03, 23:21:

I was playing with an nforce 2 board just recently. Epox 8DRA+ or something, having the same errors in memtest86+ test 5 - blockmove.

Definitely can confirm that some sticks just don't play nice together. You need to swap them out into you get something that works.

BitWrangler wrote on 2021-06-03, 22:06:

Some sticks don't play nice together, I had to go through a literal shoe box full of RAM to get 4 sticks that passed memtest in a Neo2 a few years back.

maxtherabbit wrote on 2021-06-03, 21:56:

If memtest generates an error, you have a problem, even if you don't notice it in other use cases. It doesn't ever produce spurious errors as far as I know

But why would it only throw errors with specific CPU's?
I mean this is before the memory controller moved into the CPU.
Palomino at any speed is fine, T-Bred B (Don't have any A versions =/ ) and Barton's give the exact same error in both my boards.
Down clocked to 1GHz ish at 100FSB and really slow memory timings, exactly the same errors.
Tried plenty of memsticks of various sizes and quality, same thing.

Unfortunately these are the only 2 KT266A motherboards I have.
Anyone else with a KR7A that can test this out perhaps? =)
Or maby some other KT266A MB

canthearu wrote on 2021-06-03, 23:39:

Onto my thoughts on recapping motherboards. […]
Show full quote

Onto my thoughts on recapping motherboards.

If you are going to start recapping a motherboard, you should almost definitely change all the caps on the same series, and preferably all capacitors 470uF and larger.

More than once, I've had to go back to a board multiple times to replace further capacitors months after the initial repair because other capacitors on the board have decided to spill their guts. If you want a recap job to actually stick, and not cause issues down the line, do the job in full at the beginning.

Onto how I do it:

First of all, you need a good soldering iron. Something like a KSGER T12 soldering iron or one of the other cheap soldering irons youtube people fawn over, they all seem to be inexpensive and do a good enough job. I use the D12 tip (and D16 too somethings). I also use solder braid (wick). I have a desoldering station, but don't use it for recapping because it can be rough on the boards, to easy to damage nearby traces compared with a soldering iron.

I normally work with the soldering iron set to 400C, but will boost it up to 450C when working on the well heatsinked pads around the CPU vrms. I add some fresh leaded solder to each leg first of all, this lowers the melting temperature, wets the old solder, and gives your soldering iron more direct contact with the solder joint you are trying to remove. I then heat each leg up and walk the old capacitor out of the board, by pulling on it from the other side. Then I will attempt to clear the old solder out using solder wick, attacking from both sides if I have to. I am usually able to clear at least one hole, but I don't waste too much time on this if I can't get the heavily heatsinked pads clear.

Puttng in the new cap, I have cleared both holes, I pretrim the legs down to almost the correct size, and put it in and then solder it down. I then trim the legs to the correct length, and reflow each joint. Cutting the legs will put stress on the solder joint, so reflowing after cutting is just making sure I didn't crack the joint or something while cutting the leg. If I didn't get both holes cleared, I trim down both legs, but make the leg that goes into the cleared hole a little longer to act as a guide. I then add some solder back to the hole I wasn't able to clear, and walk the new cap back into the board. The cut legs down and reflow both solder joints.

I only replace a few capacitors at a time, keeping an eye on the polarity of each capacitor and the marking on the board (be aware that there could be errors in the screen print of the board, check/document capacitor orientation before you replace any) Before powering up the board, I check I have resoldered all the capacitors properly, give it a bit of a clean with isopropyl alcohol, and double check the polarity of the capacitors.

there's the dental pick method for holes that you can't clear... a little crude but can do the job in a pinch (i.e. when lacking proper tools). IMO a lot cleaner than pushing cap legs through hot solder. but to be honest if you are suggesting that even with that "well-regarded" iron, braid and such high temps you have trouble clearing holes i would doubt that this iron is really fit for recapping motherboards. it can't be stressed enough how much nerves a desoldering gun can save when you don't have to pull 20+ hot capacitors on these boards, and for re-soldering with good quality joints, i think it really comes down to thermal transfer between the iron and the tip and that's where a lot of these cheap irons don't cut it. pros will even preheat these boards in some cases, which really shows that it's not really an easy job when you want to do it right.

i've never seen any benefit from using IPA to clean boards, all it seems to do is spread the flux residue around while leaving stains of its own. that's with the 99% stuff and 70% is even worse...