As a friend of mine from retroHW pointed out to me, in 478 uATX cards there are usually only two DDR RAM banks, this is quite common, maybe 🤔 there are also some with three or four RAM slots, however considering that the maximum installable, usually is 2 GB 🇬🇧, only two banks are enough to reach 2 GB 🇬🇧!
From what I remember, apart from the RIMMs, there are no other types of memory on the 478, except for the first versions that had SDRAM, I think there are also cards with SDRAM and DDR, they are probably PCCHIPS with non-Intel chipsets, on this I would like to briefly make a hint.
In my opinion, a card with an Intel chipset is preferable, but for non-specific uses, you can very well use a VIA or SIS chipset, I think there could be 478 cards with ATI or NVIDIA chipsets, but I'm not sure.
Instead, there are few motherboards, with PCI-E instead of AGP, I have an ASROCK with a VIA chipset, but I have to work on it to solve a problem, and make it work (let's hope so!).

Since today I came home almost immediately, I started work on an MB ABIT KG7, which involves the insertion of new electrolytic capacitors, since I got the big 4700uF 6.3V Nichicon ones, I gathered all the others, six 3300uF 6.3V and three 1000uF 10V, these capacitors are missing, because the previous owner had tried to replace them, but he couldn't clear the holes, with a little effort I did it!

As I often point out, I don't have particular or professional equipment, I cleared the holes with a tin-sucker pump, but a pair were not completely free, I had to tinker with them a little before getting a sufficient pass.

The three from 1000uF 10V, I don't know if I put them from 6.3V, since they are on a 5V line, I don't understand why they used the 10V(?), the 6.3V ones should be more than adequate, I don't think it's necessary to withstand overvoltages, indeed 6.3V should be beyond the tolerance of +5V, so right (!?!?)

I wanted to thank the precious help of Vogons, who provided me with an image with the values of the individual capacitors, and the relative voltages and Mosfet (or other) to which they are connected, even if I had already been able to understand which they were (almost all), the image can also be very useful on other motherboards, because it makes you understand which and how many capacitors are needed, for RAM CPU etc...

Wanting to make improvements to this ABIT KG7-Lite, it could be transformed into KG7, adding two more banks of DDR RAM, thus bringing from two to four RAM sockets, it could be very convenient in the absence of 1 GB RAM, you could use four 512 MB banks, to get 2 GB.

This could also be cheaper, because usually 512 MB modules, cost less than the 1 GB ones, in total four 512, they could cost almost half of two from 1, but it could also not be like that, it depends on the type of RAM, those of well-known brands, could cost a lot more, however in the end having four free banks, you can use the RAM that we keep aside, if for example we have two from 512, we can add two more from 256, that half gig more, could improve performance, maybe not much, but have 1.5 GB is better than just 1.

For now I really don't feel like making this addition, I should free more than 350 holes from the tin, later if the card works, but I'll think about it, a lot depends on what PC comes out, if for example it will serve as a Windows 98 PC, there is absolutely no need to have four banks of RAM, with two of 256 you get to the limit of 512 MB, which is the maximum memory for Windows 98SE.

Just when I thought I had finished work, surprise 😱!!!

In the back I saw some (a lot of) interrupted tracks, they are not easy to fix, but with the right patience it is a bit of luck 🍀, I could be able to fix all the interruptions, it is not a job that I do well, when the tracks are tiny, it becomes complicated to weld a wire in the middle, for this reason, and to study 📚 a solution, I take time to dedicate it to other motherboards, I want to look at this one first under the microscope 🔬, unfortunately 😣 if I had seen this problem before, I would not have soldered the new capacitors, which I could have used in other boards, however Now 👣 that there are, I'll try to go to the bottom, some of the broken tracks 😑 lead directly to the CPU, and so it's important to fix them.

Still in that damaged area, I noticed that the welds of some electrolytic capacitors, are in very bad condition, a group of four 470uF 16V and others nearby, needs a fix, but maybe 🤔 I would change those four, but only if I can fix the scratched tracks.

As I wrote before, if I had noticed the scratches on the KG7-Lite, I would have worked on another motherboard, but by now it went like this, I'm just sorry that I can't try it, unfortunately that type of fault is quite complicated to fix, because after a few minutes the eyes start to get tired, the use of the microscope is highly recommended, because the breakpoints are several, and some quite close, honestly I don't know if I'll be able to do something, with the tools I have, maybe I'll try it soon, but being able to fix it will be difficult

Another note not exactly pleasant, I only have 5 electrolytic capacitors of 3300uF 6.3V, beyond those 5 I have 2 more left, and I would say that I am forced to order more, because I have several boards to fix, and with those 5-7 I can only fix one card, I already have in mind which ones to take, I avoid stuff like TEAPO, which costs little but I don't have enough confidence, maybe they work but in the long run they could have problems? (!?!?)

The next card that needs a change of electrolytic capacitors (bulges), is a Gigabyte 478 GA-8I848PM, with an Intel Celeron CPU included, it happens to have five 3300uF electrolytic capacitors to replace, and having five available, I will use those, I hope this weekend to make the change, and start the benches and diagnostics, but everything depends on the success of the repair, if there is the BIOS screen on video, everything should be OK

I just saw on this site:

Https://www.retrohardware-reviews.de/re ... -kg7-raid/

An interesting review on an ABIT KG7-RAID motherboard, which would be the most equipped version of the KG7-LITE that I have in repair, in that article I learned very interesting things, and this convinced me to repair the tracks, or at least to try to fix the simplest ones, but to get it working again, you have to fix them all, and at work done there is always the unknown that there is some other problem, in short, no guarantee that it will work, and in any case it will take a lot of work and a lot of time to devote to it.

But I think it might be worth it, because from what I've read, it would be one of the first 462 motherboards with support for DDR RAM, and therefore with the first (or the first) new chipset, I honestly don't know how much it can improve performance, compared to the SDRAM chipsets, maybe there's not a big difference, but the DDR have the advantage of being more easily available, and the large cut ones (1 GB) don't cost stratospheric figures, but having to do the repairs in the back, I would also do the work of freeing the holes of the two Missing desks, to solder the two sockets, in total I would therefore have four desks available, and the maximum RAM would go up from 2 to 4 GB, which for a PC like this is double what you can normally put.

On the fact that there is no integrated audio, it is an all in all positive thing, because you can add a PCI card type SB, in the article they use a Live! But I think this also depends on what you have available, that is, if you have an Audigy or something equivalent, I think it would be fine.

As I mentioned in the other thread, I was confused by the use of lower capacitance caps used that was found on that page. I mean I get using poly caps for the lower esr but not knowing much about it thats quite the drop in capacitance.

Seems to work for him but for now I just need to test my KG7- raid and all 6 of the original Nichicon HM caps have popped their tops so I harvested some Rubycon mbz 3300uF from 3 old boards just to get the 6 I need. (All tested within spec).

I just need to test the board to make sure it works before I dump the extra $$ into it for caps.

Grem Five wrote on 2024-12-13, 17:45:

As I mentioned in the other thread, I was confused by the use of lower capacitance caps used that was found on that page. I mean I get using poly caps for the lower esr but not knowing much about it thats quite the drop in capacitance.

Seems to work for him but for now I just need to test my KG7- raid and all 6 of the original Nichicon HM caps have popped their tops so I harvested some Rubycon mbz 3300uF from 3 old boards just to get the 6 I need. (All tested within spec).

I just need to test the board to make sure it works before I dump the extra $$ into it for caps.

Thank you ☺️ very much for the link, which I found really interesting.
So in my opinion, reading what it says, he didn't replace them to reduce the capacity, but he changed them to reduce the height, and be able to mount a wider heatsink, exactly I don't remember which one, but I think the purpose was to overclock us, and in conclusion improve heat disposal 🥵
I saw some photos of these MB KG7-X, with some 2200uF capacitors, even if a couple were 3300, here maybe he thought that the 2200 ones were enough, honestly I don't know if they came out of the factory with all of 3300, or with two of 3300 and the others of 2200.
Well if you managed to recover some capacitors, and replace them (removing those ko), I think they could last quite a while, I've seen Nichicon swollen capacitors over time, but maybe no Rubycon (I find it's hard to find bulging), while with other brands, Panasonic I've seen very few swollen ones, I usually use either one or the other, to replace the swollen ones or out of tolerance.

Since you are going to work on this board I'm going to post any thoughts here.

Might not hurt to look into see if its worth adding the two other memory slots as the manual says this:

KG7/KG7-RAID Only: […]
Show full quote

KG7/KG7-RAID Only:

• Four 184-pin DIMM slots support PC1600/PC2100 DDR SDRAM modules
• Supports two DIMMs of unbuffered DDR SDRAM for a maximum of 2 GB (64, 128, 256,512 and 1024 MB DDR SDRAM modules)
• Supports four DIMMs of registered DDR SDRAM for a maximum of 3.5 GB (64, 128, 256,512 and 1024 MB DDR SDRAM modules)

I'm guessing thats why other boards like this one only has two memory slots is because of the memory gets unstable using more than two slots unless they use registered memory.

I have an Asus A7M266 which is the same north and south bridge as the Abit KG7 but Asus choose to go with only two memory slots and back in the day it retailed for more than the KG7. At that same site https://www.retrohardware-reviews.de/review-asus-a7m266/ he pointed out that when the Asus board was announced the previous year it had a pic of the board with four memory slots but between that announcement and when they released the board they dropped the four slots down to two slots. My thoughts is they might have done that for stability and thinking not every one is going to notice to run more slots would require registered memory over unregistered memory.

I might need to test that when I get the KG7-raid in my hands and recapped but that might not be a problem. https://www.anandtech.com/show/845/2

According to ABIT, one of the reasons for their late appearance in the 760 market was their decision to make a board capable of running stably with more than two DIMMs. In fact, ABIT went a step further with a total of four DDR DIMM slots on the KG7-RAID, supporting up to 4GB of memory. Most other AMD 760 boards only have 2 DIMM slots. The KG7-RAID comes with 4 DDR DIMM slots, ensuring that it stands out in this group. Now it's one thing to include 4 DIMM slots on a motherboard, but it's an entirely different thing to ensure that they all work properly, without sacrificing stability or performance. We had no problem populating all four slots at the same time on the KG7-RAID. This is quite remarkable, considering the tight timing required for DDR SDRAM.

From that same article it seems Gigabyte might not been as lucky.

The 7DXR features three DIMM slots, rather than the two found on most 760 boards. In fact, it's one of only two boards to offer more than two DIMM slots, the other being the ABIT KG7-RAID. Unfortunately, with all the three DIMM slots populated, we had to back off the CAS latency to 2.5 to maintain sufficient stability.

Grem Five wrote on 2024-12-13, 22:50:

Since you are going to work on this board I'm going to post any thoughts here. […]
Show full quote

Since you are going to work on this board I'm going to post any thoughts here.

Might not hurt to look into see if its worth adding the two other memory slots as the manual says this:

KG7/KG7-RAID Only: […]
Show full quote

KG7/KG7-RAID Only:

• Four 184-pin DIMM slots support PC1600/PC2100 DDR SDRAM modules
• Supports two DIMMs of unbuffered DDR SDRAM for a maximum of 2 GB (64, 128, 256,512 and 1024 MB DDR SDRAM modules)
• Supports four DIMMs of registered DDR SDRAM for a maximum of 3.5 GB (64, 128, 256,512 and 1024 MB DDR SDRAM modules)

I'm guessing thats why other boards like this one only has two memory slots is because of the memory gets unstable using more than two slots unless they use registered memory.

I have an Asus A7M266 which is the same north and south bridge as the Abit KG7 but Asus choose to go with only two memory slots and back in the day it retailed for more than the KG7. At that same site https://www.retrohardware-reviews.de/review-asus-a7m266/ he pointed out that when the Asus board was announced the previous year it had a pic of the board with four memory slots but between that announcement and when they released the board they dropped the four slots down to two slots. My thoughts is they might have done that for stability and thinking not every one is going to notice to run more slots would require registered memory over unregistered memory.

I might need to test that when I get the KG7-raid in my hands and recapped but that might not be a problem. https://www.anandtech.com/show/845/2

According to ABIT, one of the reasons for their late appearance in the 760 market was their decision to make a board capable of running stably with more than two DIMMs. In fact, ABIT went a step further with a total of four DDR DIMM slots on the KG7-RAID, supporting up to 4GB of memory. Most other AMD 760 boards only have 2 DIMM slots. The KG7-RAID comes with 4 DDR DIMM slots, ensuring that it stands out in this group. Now it's one thing to include 4 DIMM slots on a motherboard, but it's an entirely different thing to ensure that they all work properly, without sacrificing stability or performance. We had no problem populating all four slots at the same time on the KG7-RAID. This is quite remarkable, considering the tight timing required for DDR SDRAM.

From that same article it seems Gigabyte might not been as lucky.

The 7DXR features three DIMM slots, rather than the two found on most 760 boards. In fact, it's one of only two boards to offer more than two DIMM slots, the other being the ABIT KG7-RAID. Unfortunately, with all the three DIMM slots populated, we had to back off the CAS latency to 2.5 to maintain sufficient stability.

Thank you ☺️ very much friend.
Yes I knew this thing about the registered RAM, however I have to try to free the holes from the pond, if I can easily, that is, without the need for particular or professional equipment, well I continue with the marathon, I should clear more than 350 holes!
Regarding the other links, thank you in advance, I will need some time to read them and understand the interesting points, regarding the MB 462, I have several of both ABIT and ASUS, I also have a Gigabyte that I wrote about recently, also under repair, I will try to finalize this one as soon as possible.

Tomorrow I will try to fix the Gigabyte GA-8I848PM motherboard, with Celeron CPU, there is to replace all the swollen electrolytic capacitors, and having the remaining ones, I can use them all, except the two Rubycon, which I keep for any boards with one or two (max) electrolytic capacitors to change, finished the work, I have to figure out if I have a suitable heatsink, I don't have any memory, but I could adapt a different one, we'll see what I recover.

In general it is a card quite similar to others that I have working, even if this differs in the uATX format, in the smaller number of RAM slots, and PCI slots, the last difference is the CPU, I usually use the P4, here instead there is a Celeron, this type of CPU, over time I have always tried to replace it with faster P4, besides I made a comparison some time ago between my P4 2000/400/512 and a Celeron 2600/400/128, on the card the Celeron should have been a little better, in reality it was quite slower, at least visually, I did not do bench, but having remained Disappointed, I preferred to avoid reverse upgrades, that is, from P4 to Celeron, different matter if I initially had a Celeron (as in this case), but switching to P4 having one available, it would undoubtedly be better, especially if you use an FSB 533 CPU, which I think is the easiest to overclock, probably if that Celeron was unbearably slow, I would replace it with a P4.

I started work on the Gigabyte GA-8I848PM motherboard, with Celeron CPU, there is to replace all the swollen electrolytic capacitors, in total there are five from 3300uF, the others are visually fine, and I don't think we need to check them, even if as we will see one I had to remove it, but checking it with my ESR70, it says it's fine, so I'll just take a thorough look, in case you see something leaking from below.

First I had to remove the plastic part for the hooking of the heat sink, if you are familiar it is relatively simple and fast, in short you have to lift ℹ four plastic nails, and then from the back tighten and push the protruding parts.

I removed the five capacitors quite easily, in fact later I had to remove a sixth electrolytic capacitor, a Rubycon MBZ of 1500uF 6.3V, because it got in the way and I couldn't free the holes of the one in the middle, those removed are all Nichicon HM 3300uF 6.3V, they probably belong to series produced initially, which over time caused swelling problems, later someone said the problem was corrected, but I don't know if it was solved definitively, anyway I replace them with Panasonics, so no problem I hope.

Unfortunately I couldn't finish the work, because two holes remained occluded, it's something that strangely often happens in Gigabyte motherboards, the positive part must have a very large surface, and it makes the molten tin cool quickly, on the one hand it melts, on the other it doesn't!

For problem I will make other attempts, but not now, if I can today well, but I will look for some useful info, maybe there are some tricks that can help, maybe it would be to warm up the area, and in this regard, I noticed in the back under the Mosfets, burnished areas, a sign that the board was quite warm, I hope it is not the sign of extreme overclocking, and therefore there are no other problems besides the swollen capacitors.

A little while ago I finished the work on the Gigabyte GA-8I848PM motherboard, with Celeron D CPU, there were all the swollen electrolytic capacitors to be replaced, five 3300uF Nichicon, which I replaced with 3300uF 6.3V Panasonics, practically the same but of a different brand.

First, however, I freed the two occluded holes from the tin, at first I couldn't do anything, then I used a fairly simple system, a nice hot pin, which I insert into the molten tin, until it arrives on the other side, basically it doesn't eliminate the tin, but creates a sufficient hole to insert the condenser pin, in short, that's all.

I soldered the six capacitors, obviously being careful to put them in the right direction, some welds are not the best, but it doesn't matter, they all seem quite acceptable (even visually), after I did a cleaning with alcohol, finally I reassembled the tile to hook the CPU heatsink, just insert it into the holes, and lower the four plastic studs, the last operation, a superficial cleaning to the CPU, which is not at all as bad as I thought, it is a Celeron, but it is a Celeron D 2.66GHz/256/533, it should not be much slower than the P4 of equal frequency, I will see if the board works, to try Overclocking at 3.33GHz, raising the FSB to 667, on an ASUS P4B533 with a P4 worked great, you just have to use 333 or 400 DDR RAM, the 266 ones can create problems with PC locking.

To finish and to see what the final result was, I mounted on the MB the indispensable to try the boot, a 478 CPU heat sink, a DDR RAM (or rather some), the 8MB PCI ATI Rage XL video card, and the usual ISA PCI Post Card, but first I checked the power lines, and they were fine, no short or strange values, connected the power supply, I pressed PWR-ON and the fans went on, the PC unfortunately NOT!!!

I made some attempts but I didn't see any change, the Post Card always marks — —, no code displayed, and some of the LEDs on, but the others off, while among those off, there should be at least flashes, which at the moment I haven't seen.

Unfortunately this MB has the BIOS chip soldered, I will still try to reset the CMOS, I don't think anything will change, but otherwise there could be some wake-up signal???

I conclude by saying that, after a few other attempts, if I don't get results, the card would go among those that don't work, I forgot that I did a quick voltage check on the PIN 2 of the Mosfet, everything seems ok maybe the VCORE is lower, I think around 1.35V, I don't know I should do an in-depth check, and measure all the PINs, to understand if there is an anomaly, it seems not to me, however I will try one of my CPUs, in case it is that Celeron D that does not start the PC.

Since with the Gigabyte I had no luck 🍀, despite some attempts I didn't get anything, I decide to suspend any other attempt (at least for the moment) to start, I dedicated enough time to it, the only strange thing I found is the VCore at almost 1.3V, the value seems too low to me, even if it should allow startup, usually it should be at about 1.5V, however I will see if changing CPU, there is some sign of awakening, latest news, the board should come from a Fujitsu-Siemens PC, strangely they use a motherboard not of their manufacture, but overall it is a card that I like 😻 and I would like to make it work.
But since I had time, I decide to change the motherboard, always 478 but ATX, I put the DFI NB78-BL motherboard, it already has a CPU installed and a RAM, I add the 8MB VGA PCI ATI Rage XL, I connect everything (or rather I thought I had connected everything), but I check the manual, and the Celeron D 2.66GHz/256/533 CPU, it is not supported, to avoid possible problems, I put a P4 1700/256/400 (Willy), and I try to start: No sign of life, ditto like the other MB.
I check the Mosfets of the VRM, and I don't have any voltage 😬, very strange, but after a careful look, I understand the problem, I forgot to connect the 4 PIN of the ATX power supply, so I try again:
IT WORKS!!!
I see the codes scroll and the Logo screen appears on the video, followed immediately after by the main BIOS screen, just 😡 press F1 or Delete, to continue or enter.

PC@LIVE wrote on 2024-12-15, 20:11:

...IT WORKS!!!...

congrats 😀

Ujeen wrote on 2024-12-15, 23:30:

PC@LIVE wrote on 2024-12-15, 20:11:

...IT WORKS!!!...

congrats 😀

Thank you very much ☺️ friend
Today I will do some upgrades if I can, and some tests with bench and diagnostic programs.
Gigabyte I have to do other checks, it is possible that there are some bad capacitors 🦹‍♂️ with ESR a little high, and this gives as a result, VCORE values below 👇 than normal, maybe that's why it doesn't start, but here too I should try another CPU, at least to understand if the Celeron D that is working or not.

Today I started doing some bench on the DFI with P4 1700, mostly I used Phil's benches, but not all of them, the lighter ones, I didn't use them, even if maybe 🤔 the scores were the same, however I collected some info, to then be able to compare them with the CPU that I will put shortly, looking among those in the collection, I realized that I don't have as many P4 533 as I thought, and that I have quite a bit of Celeron, about half are Celeron D, I think I bought them years ago at a symbolic price, today I would pay them for More, but it doesn't matter, I just have to reorder those CPUs, to see which ones I have double, or even triples!

I chose a CPU I would say definitive, a P4 NW 2.66GHz/512/533, of these CPUs I have one that I should repair, a PIN is missing, to fix it, you need to solder a very thin PIN taken from a destroyed CPU 😭, currently I don't want to do it at all, maybe in the future with the right tools 🛠.

On a similar MB of ASUS, a 2.66 I brought it to 3.33, and it works very well, in practice I raised the FSB from 533 to 667, here in the DFI I don't know if it's possible to do the same, but I think you can easily get around 3.00GHz, which for me is enough.

For today I finished with the P4 1700 + DFI 478, I took note of all the scores of Phil's bench, some seemed modest to me, not at the level of a P3 1000, it's strange but being under DOS, it is possible that some functions or instructions are not used, because there was nothing like that, before the P.MMX there was only Processor and Coprocessor, so no SSE or similar stuff, so everything was quite simple (I guess).

I tried at the end of the benches, the ASTRA program, to see the info it gives, I didn't expect anything special, the images speak for themselves

Since I have the possibility to work on other motherboards, today I started the repair on a QDI P6V694T/A10T, a socket 370 with a CPU already installed, I don't know what it is, but we will find out at the moment of startup, even if the premises are quite interesting 🧐, it should be a Tualatin, if I wanted I could remove the heatsink to find out what it is, but I would ruin the surprise 😱!!!

The visible problem of this motherboard, are three swollen electrolytic capacitors, 1000uF 6.3V from Fujicon, aesthetically similar to Panasonic, but apart from this similarity, I think there is a nice difference, after removing the three, I removed a fourth capacitor, to understand in what conditions the others may be, after the checks, from the displayed values, I think they are all to be changed, even the others still ⚓️ present on the motherboard, the ESR is high and the capacity is a little low, I think that if the board was still working, it had some big instability problem.

How many are in total? Of the same capacity voltage and size are nine in total, four already removed, plus five to be removed.

For the moment I'm just changing only these 1000uF, but there are other bigger ones with 1500uF 6.3V, which at the moment I haven't decided whether to leave them or not, they are signed FJ (Panasonic), but I'll make the decision later.