So after welding the new electrolytic capacitors, I tried to see if one of the motherboards works.

As we will see, the replacement work was successful, for the moment I have not completed the work on the ASUS P4SDX, while the BIOSTAR is waiting to be verified and subsequently tested.

But what motherboard is on the bench?

An Intel D845GEBV2, with CPU P4 2400/533, in this board I replaced three swollen electrolytic capacitors, with three others the same but from Nichicon, definitely better than the ones that were installed.

As soon as I started the power supply, the motherboard started without pressing the power button, soon enough the sequence of post codes stopped, I think the RAM was not compatible, however without RAM the PC beeps, a sign that both the BIOS and the CPU are working.

I summarize briefly, avoiding commenting on the various steps and startup attempts, after many attempts, I found a perfect combination that made the PC work, but unfortunately there remained a tendency of the PC to crash quite soon, so I decide to connect the usual IDE CF adapter with DOS, to run the bench, after some startup attempts, a fatal error (DMA) appeared, after that screen, nothing appeared on video anymore, and the post codes stop quite soon, in short, try and try again, but nothing no progress, and since before it was practically unusable though Starting out, I decided not to waste any more time, and I brought it back among the motherboards to be repaired.

Continuing with the 478 motherboards to try, because for the moment I will not do other repairs, I started from the BIOSTAR that I fixed, the board is P4M80-M4, for the test I use a P4 2.80GHz/512/800, and I start adding what is needed, to do a boot test.

As soon as I have everything ready, I turn on the video and the power supply, close the contacts for the PWR/ON, and nothing happens!

I check in the PINs and in one I find 3.25V, this is fine, unfortunately when it does not start, it is not at all easy to understand the reason, but it is not said that the motherboard cannot be fixed, after a quick check, I find on the back of the SB, a missing SMD, from the initials I understand that it is an SMD capacitor, I think that adding one, it will hardly work again, if instead it had been an inductor, that would have been the problem (missing).

No luck then with this motherboard, which was already in difficult condition, and that I had fixed hoping to be able to make it work (?!).

But I have another motherboard of the BIOSTAR S.478, it's an ATX P4TSP-D2 VER:1.3, with Intel 848 chipset, I don't really think I've done any work on this motherboard, in short, it looks like it's in excellent condition, and just try it to see if it works?!

I reuse the CPU and everything else used in the other motherboard, and I try to boot, the card starts, but in reality the card does not work, because there is no post code visible, everything normal on the LEDs except the codes that are not there!

Later I performed some tests, but nothing has changed, of course I did not check the tensions, so there may still be something to be done, we will see to do further investigations in the future, but for now it has not been possible to do anything else, and unfortunately neither of the two has shown any encouraging signs, to be able to make some progress.

I resume a job started some time ago, the motherboard was ready to weld the new electrolytic capacitors, but also the new ATX 20 PIN socket, before welding everything, I noticed that one or maybe two SMD capacitors had jumped off, right next to the ATX socket, I imagine it was the clumsy result of replacing the socket, because it had blackened pins, anyway in short, now I can't weld the connector if I don't fix those two missing SMDs first.

For the replacement, I don't know the original values, so I just remove two from a scrap card, visually they seem quite similar to me, this obviously does not guarantee that they are right as substitutes, but they could be installed and work, from what I see they are connected on two PINs, one on PIN 18 and the other on PIN 9.

I haven't checked which power supply cable matches, but the track is pretty wide on the 9, while on the 18 it's thin, this may be unimportant, but it could help when we re-send, we'll have to be careful not to damage the thinner track.

As soon as I finished re-soldering the SMD, it will be the ATX 20 PIN connector, I disassembled one from another motherboard, then it will be the turn of the electrolytic capacitors, I removed six in total, two of 1200 uF 6.3V and four of 3300 uF 6.3V, the original ones were from GSC and CHOYO, no need to comment on these two brands.

The motherboard, however, is a Gigabyte SKT 462/A, GA-7VAXP REV:1.1, with integrated VIA chipset, audio and LAN, three DDR RAM slots, one AGP and five PCI, apart from this damage to the SMD, the rest seems to be in good condition, and this gives hope that after welding everything, it should work again, at least we hope so.

I made an attempt, trying to re-weld the two SMD capacitors, first I cleaned the tracks, removing the pieces of those that jumped off, then I welded on the seats of the cool tin, so it remains to attach the two SMDs, unfortunately I couldn't, for a couple of reasons, one for the presence of other nearby SMDs, and for the size of the welder tip (too big, even if the tip is thin), there is also a capacitor that prevents a little, in short it was almost done, unfortunately my equipment limits what I can do.

For the welding of SMDs, especially the small ones, some use hot air, or a tweezer welder with two tips, maybe it could be done with a very thin welder tip, surely the use of an enlarger (or microscope) would avoid involving nearby SMDs, then anyway watching YouTube videos, everyone uses a preferred technique, not necessarily the same for everyone, in short personally I can only watch, but I can't imitate.

For now I want to rearrange the ideas, because even if the result was aesthetically horrible, I care relatively, in the past I have re-welded some SMD capacitors in various MB, here I have found some unexpected difficulties, I hope not insurmountable.

However, I would avoid trying again, without having done any tests, on how to position the tip of the welder, and I think it might be better to add some tin to the SMD, so that it can be held still with the tweezers, and with the other hand the welder.

I hope to succeed by this week, it would be nice if this Gigabyte SKT 462/A works again, even if I have other 462 motherboards, I find this one aesthetically beautiful, and I wouldn't mind having it among the working PCs in my collection.

I just finished welding the two SMDs, the final result is that they are present, aesthetically I am the first to say that they are sorry, but if after a check they are not short, I would say that apart from the painful aesthetics, they should do their job, I do not know if good or bad, this depends on the values of the original ones, which I do not know what they are (?), and not having tools 🛠 that allow me to do it better than this, I am forced to continue welding everything else, that is ATX connector and electrolytic capacitors, finally do the final test, does it work? Or the substitutions had no effect!?

Normally I would say that the chances that it will work, and often perfectly, are quite high, but there is always an unknown to keep in mind, that is that while doing a great job (and here I would say that it is enough), it may happen that there is some other non-visible problem, that is, solving the visible problems (SMD capacitors etc...) it is not that the others are automatically solved (if any), but certainly the use of a post card, can help to understand if there is any problem, or possibly if there is something that prevents normal operation, but perhaps there are more professional post cards, which can identify with More precision where the problem may be, in the one I usually use, instead there is a paper manual whose reliability is somewhat uncertain, but anyway at least for me, it is useful to understand what the PC does during startup, or if it does something!

We'll see if by the evening I can finish with the other welds, I leave the ATX connector for last, I wouldn't want to have to disassemble it after welding it, as soon as I finish with the welds, I would have to look for a CPU to put it, usually for testing I use a Duron, but then as a definitive CPU, I prefer a Sempron or an Athlon XP, for the RAMs I could start directly with DDR 400, but on the final quantity, a lot will depend on what use it will have.

I've almost finished the work on the Gigabyte GA-7VAXP motherboard, I've soldered all the capacitors (electrolytic and SMD), only the ATX 20 PIN connector is missing, for that I'll need another tin, I've almost exhausted the spool that I always use, and I wouldn't want to be left without the work started, moreover the wire is quite thin (0.5 mm), and this is useful to me when there is something tiny to weld.

Lately I've made a nice selection by trying various motherboards, some could be interesting, unfortunately many didn't work or don't work at all, for some I think attempts could be made in the future, even if it's quite difficult to understand what the problem is.

An interesting thing about this Gigabyte, is the dual BIOS, basically it has two welded PLCC32 chips, here I hope that the previous owner, didn't mess with the BIOS, but if by chance it did something wrong, the second BIOS chip, it could be the best solution for the restoration, except that obviously it didn't mess that up too, I really hope not!

Another interesting thing, there are 4 IDE 40 PIN ports, theoretically up to eight installable drives, plus two other floppy disks, more realistically they could be used, four drives all like Master, maybe two HD and two CD or DVD players/writers (or both), or a ZIP100 or LS120 Floppy player if available.

But it doesn't end here, in addition to the two USB sockets in the shield, we can connect other USB ports, through two USB sockets on the MB (yellow), while the three in gray are FireWire sockets, probably brackets were provided together with the motherboard, but maybe they were reused on other motherboards.

Last consideration, the AGP is an 8X, you can read it on the chipset sink, I don't think it makes much difference between a 4X or an 8X, but in any case I will try to use a VGA AGP 8X, I can't say which one now, first we'll see how it goes (I hope well), and then I'll decide the final configuration for this PC.

After replacing the capacitors (electrolytic and SMD), on the Gigabyte GA-7VAXP REV: 1.1, today I had to complete the welding work, of the ATX 20 PIN connector, I had to buy in a hobby store, a little tin to fill the big holes on the motherboard, I did not find what I was looking for, but I bought something quite similar (?), normally I use a 0.5 mm 60/40 Sn/Pb wire with rosin, but for this work (and other similar ones in the future), I took 20 g in diameter 1.5 mm 50% Sn, the other 50% Sn it is not written if it is Pb or other.

This allowed me to complete the work, because I almost finished the 0.5 mm wire, and I prefer to use it for tiny welds, however using the 1.5 mm one, I saw that to fill those big holes, it was both fast and simple, which instead would have taken a lot of time, using the 0.5 mm one.

Now I start to prepare the motherboard for a first boot, as I said before, I like this board and I put it (a long time ago) as an image in my group "PC@live Whatsapp", in particular I really like the colors used, blue for PCB and green for SKT AGP and IDE, too bad there is no ISA slot, but I think they were abandoned when you switched from SDRAM to DDR, strange also that there is no AMR or CNR, useful if you have a modem of this type.

Choosing the CPU, it is initially very simple, I will use an AMD Duron with FSB 100, or alternatively an Athlon FSB 100, in reality that 100 would be 200, just to clarify, this motherboard has jumpers for the selection of 5X and 5.5X, as far as I know there are no 500 or 550 MHz CPUs (SKT 462/A), but maybe if we use one with unlocked multi, it could be possible to select those multi, to experiment or make reduced frequency benches.

Currently, the CK RATIO (1,2,3,4,5) is in Auto, so it would be able to automatically read the multi, while the CPU Clock (SW1) is 166 MHz, so I will have to reverse the switches to lower it to 100 MHz, very strange instead the Dual BIOS chip, it has very few connections both at the top and on the back of the PCB, I suspect that the connections are internal, obviously I would have preferred that one of the two chips was removable, but here they are both welded!

Before starting, I'll check the lines, just to see if there's anything strange, and I hope everything is ok.

In summary, having little time available, I focused the work on a motherboard, it is not an ASUS, but the Gigabyte GA-7VAXP REV: 1.1.

The work I started some time ago, was completed today, with the welding of the ATX connector to 20 PIN (recovered from another board), while yesterday I had welded the two missing SMDs, with others of recovery, I put these blindly, not knowing the original values, and not even those of the substitutes.

The swollen electrolytes have been replaced with the same ones, but of better quality, or so it seems, because apparently, there are copies around, which resemble the original ones.

After gathering everything I need for a first boot, I install an AMD Duron 850 MHz CPU, a temporary CPU that I used in the past on other motherboards.

I try to start, and it starts but there is no code on the PCI post card, you can see the usual — —, I quickly check the tensions on the Mosfet, and everything seems right, except for the VRM area that in a short time heats up a lot! Shouldn't it be so hot?!

At this point, now disappointed by the result, I make one last attempt, I change the Duron with a Morgan 1000, I try the start again, I have the codes, it starts and it works!!! Mosfet temperatures are normal, now it's fine.

I continue, with the bench DOS, first at the frequency of 500 MHz, which is the minimum for this card, then at 1000 MHz which is the CPU frequency, finally I change the VGA PCI with an ATi Radeon 9200 AGP.

Surely after the initial attempt went wrong, having obtained the full functionality, is a good satisfaction.

Now for the moment I run a series of benches, which I had to redo, because I changed the configuration three times, the first with 500 MHz CPU and VGA PCI (8MB Rage XL), the second with 1000 MHz CPU, the third with 1000 MHz CPU and VGA AGP (Radeon 9200).

Some don't show much difference, after the change of VGA, in fact the PCI seems to make excellent scores, even better than the AGP, however there could be a certain difference using a different VGA AGP (?).

I haven't tried if selecting other multi, it accepts them and it works (I think so), if I wanted I could try a little overclock, for example with multi 11X, if it's okay it means I can manually select what I want.

I will start to evaluate which CPU to put definitively, the RAM instead could have increased a lot, and then also the VGA could be replaced, with a better one, for the CPU I have some Barton XP FSB 333, maybe it would be interesting to try a 2500+ and then raise the FSB to 400, it should reach more or less the same level as a 3200+!

Going back to the Bench, here are the results of Phil's, in the three configurations (see above

2) 333.7 - 405.4 - 255.7
3) 211.4 - 257.0 - 288.6
4) 68.0 - 82.2 - 50.4
5) 132.9 - 186.6 - 199.9
6) 35.2 - 38.4 - 25.0

A) 377.22 - 426.80 - 367.93
B) 82.35 - 90.10 - 77.88
C) 104.4 - 146.6 - 156.7
D) 46.4 - 63.6 - 61.9
E) 19.0 - 19.8 - 47.2
M) 331 - 359 - 558
N) 598.98 - 1197.67 - 1197.71

As I anticipated, I wanted to try other multi on the Gigabyte GA-7VAXP, I did some tests, and I found most of the selected multi working, this indicates that the CPU is unlocked, and consequently you can select a lower multi with higher FSB, so instead of 10X and 100 (200), you can use 7.5X and 133 (266), which always corresponds to 1000 MHz!

But I did other tests, using multi over 10X, and all of these worked: 10.5X 11.0X 11.5X 12.0X.

For now I have reached 1200 MHz, without any overvolt, all at standard voltage, and after doing the usual bench, I wrote down all the results, some are almost the same, but I don't know if the limit is the VGA or there is something else that influences the results, on the 486 PCI I found big differences between the RAM settings, however in short here I didn't change anything, only the multi, so I don't know if it can be done better, with faster settings.

Since it is possible to go to higher frequencies, I would be curious to try, the multi 5X with the FSB 200 (400), the result would always be 1000, but the greater speed (memories) would be a nice jump, even if I don't know how much the scores could improve, the result could be really interesting, but I don't think it's twice as much, more realistic I think it can be 10-15% more, at most 25% but I don't think beyond that.

I will do progressive tests, going up one step at a time, so first 133 then 166 and finally 200, which correspond to 266 333 and 400, perhaps for some benches it will not be possible to get improvements, because many benches can be mostly based on the speed of the CPU or VGA, in short I will be able to see differences only in the benches that make a lot of use of RAM.

Not happy, in fact I would say the 1200 MHz, I tried the last possible multi from the jumpers, that is the 12.5X which corresponds to 1250 MHz, an overclock of 25%, at standard voltages, that of the CPU is about 1.6V, but so far all the bench performed have worked, and the temperatures seem to be normal, even if perhaps a heat sink with the copper contact part comes to the help (to see the color), in short it could be a little more efficient, but the maximum would be a heat pipe, in any case after starting up, I see that even at 1250 MHz the PC It works very well, and the usual benches, show a certain difference, even if some seem limited by VGA, which honestly I didn't understand how fast the AGP goes, maybe in the absence of special drivers it works at only 2X?.

As far as I understand, multi from 13X up are possible, but only after running a mod., or if the CPU has left the factory with higher multi, it is not necessary to make any modification, but you just have to find out which selection corresponds to the multi of our CPU, if we have a Duron 1300, and we want to get to 1500, we will have to look for the 15X to which it corresponds, maybe 7X or 8X or something like that.

Some time ago on an ABIT, I had forced a 2400+ to over 2000 MHz and FSB 266 with this system, if I wanted here too I could do the same, but certainly not at more than 2000 MHz, because the Duron CPU I use goes to only 1000 MHz, of course it could be possible with particular cooling systems and overvolts, but I prefer CPUs that do not need any overvolt, and at most I can put a more generous heatsink (larer dimensions), or more efficient, so as to keep the temperatures lower.

Soon, I will group the results of the bench, at the various frequencies, then I should start doing those with the highest FSB frequencies, always starting from 1000 MHz, and for comparison I will get to 1250 MHz (or a little more), so even if I could go much higher, for the moment this is not necessary.

But I already plan to work on an ABIT SKT 462/A, so if I can finish fixing it, I will temporarily pass this Duron on to the ABIT, and here I would put a Barton 2500+?

For the record, I continued with other settings, and for the first time I lowered the multi and raised the FSB, well yes I must admit that this Duron Morgan 1000 CPU, has always responded present and has never disappointed, of course to put it in difficulty I think I should go up to 1500 MHz, but for the moment up to 1250 MHz, I got there without any problem, it remains to be seen if there may be any problem during the installation of Windows, after all bench aside, no I did nothing special in DOS.

So far I'm collecting information, doing bench at various frequencies, first I tried the 750 MHz (7.5X 100), nothing in particular to report, later I'll see to make a picture 🖼 with all the frequencies and scores, but why try the 750 MHz?

Because the multi 7.5X allows with the FSB 133 (266), to obtain the original frequency of 1000 MHz, I believe that even the RAMs can work faster, but I'm not sure about this, I think so if the RAMs work in sync, maybe the doubt can be dispelled by looking in the BIOS, or just read what one of the diagnostic programs reports.

Looking at the results at (7.5X 133) = 1000 MHz, I notice that many scores are similar to those obtained at 1250 MHz, or even better, this I think depends on how much the memory weighs in those benches!

I will not change FSB (266), and I will try to raise the multi to 8X or 8.5X, and then also to 9X and 9.5X, in some cases I would get slightly different frequencies from those tried so far, but all in all they should guarantee much better results, but be careful no doubling of the scores, then of course you have to consider that any result at a certain frequency, could be improved by changing the times of the RAM or other in the BIOS.

I continue to experiment with some overclocking, with the Duron 1000 Morgan and the Gigabyte GA-7VAXP, but for the moment I do not report the values here, because I want to group everything, and so we will have to wait for these experiments to finish.

An interesting thing is that if the values do not change much (at the same frequency), it means that the benches do not make much reference to RAM, so in practice, using faster or slower settings, it would not change the final results.

So now, let's see the frequencies I've already experienced, and those to try, so let's start from the FSB 100 (200), I went from 10.5X up to 12.5X (maximum selectable here), and I tried some lower multi, but for completeness I'll make additional bench with multi lower than 10X, so as to cover all the frequencies obtainable!

With the FSB 133, instead I started from 7.5X, and I went up so far to 9X, I have to try the 9.5X, the 10X maybe I can try it, it depends if it goes to that frequency without any problem, I don't want to go up anymore, even if it were possible, just because I have no idea if the limit is close, that is, the higher you go, and at what frequency you need an overvolt, which for me it would also be possible to do, but I want to see what is the maximum frequency that can be obtained (without overvolt)!

The next step is to go down with the multi to 6X, and go up with the FSB to 166 (333), in this way we always start from 1000 MHz, and then we try to go up a little, with 6.5X 7.0X and later maybe 7.5X or more.

If those at 166 MHz are fine, I would have one last frequency to try absolutely, it would be the one at 200 MHz (400), here of course only multi 5X and little more, already with the 6X we are at 1200 MHz, with the 6.5X even at 1300!, In short, to go further, I don't think it's possible, or rather if it is, you can do it from BIOS and not from jumpers, but this should raise beyond the FSB to also the PCI and AGP RAM frequencies, so you could reach the limits quite easily, but in the past I did something like that, like passing From 166 MHz (333) to 175 MHz (350), or maybe even higher, like 180 (360) but I don't remember exactly.

I just finished with the benches at various frequencies, or rather I did several at low frequency, plus others in high frequency, summarizing I tried all the frequencies with FSB 100 (200), and they range from 500 MHz to 1250 MHz, in case we find a Duron 1300 between our paws, we will have to select the 5X and theoretically, the rest he does by adding 8X to the multi 5X, if instead we wanted to use CPUs with higher multi, in this motherboard we can get to the maximum of 20.5X (12.5+8)!

It seems very interesting but I don't think you need multi so high, because from what I see, this motherboard has various selections for the FSB, from 100 to 200 MHz, which correspond to 200 and 400 MHz, plus the intermediate ones of 133 and 166 MHz (266 and 333), as we will see I made the 1000 work with standard FSB (100-200), and also at 133 (266), finally even at 166 (333), while the 200 (400) has not yet been ️ tried.

Both at 133-266 and 166-333, I tried slightly higher multi, without going too much beyond 1250 MHz, first I want to complete the lower multi benches, that is, with frequencies below 1000 MHz, then we will see to be around 1300 MHz, so far this CPU and this motherboard, they have always responded very well, without any error in the benches and without missing starts or resets.

And this possibility of easily changing the frequency, I find it quite useful, because it could make a program run well, in the sense that if it is too fast, you can slow down, maybe even from BIOS you could lower the FSB (below 100?), in short, at this moment I can go up from 500 to 1250 MHz, or on the contrary I can go down from 1250 to 500 MHz!

I publish the image with Phil's benches, plus HWINFO 3.05, at various frequencies with FSB 100, I included the values obtained with the 8MB ATi Rage XL PCI, while the rest is obtained with an ATi Radeon 9200 AGP.