MSI 6340 recap

[trodas]

Some five years ago my friend Peter give my this MSI 6340 micro ATX mobo with the Duron 750 CPU. It allegedly worked, but very, very unstable. However I got intersed, because I see mainboard back from 2001 for Socket 462 processors - Duron, Athlon - that got interestingly designed Vcore regulator and mainly a polymer Fujitsu caps. The yellow ones are polymers - regardless that they have the top perforations like elektrolyte caps:



Also as they say - everything small is nice. And this board is really small, almost like just fit tomy hand:



The fact that the mainboard is unstable is understandable. It is only enough to look, what caps are on it - many bad caps like these Chhsi ones:



And these terrible caps are combined with polymers, witch is trully interesting combination, witch in the end probably allow the board to survive all the time working. At least sort of...



Near Vcore output coil, witch get pretty how when the caps in Vcore are bad, the Chhsi cap is leaking now:



...but it looks like these two good polymers hold him pretty well, so the computer somehow worked.



Except quality polymers there are on the mainboard also good caps - Chemicon KZE - as input filter caps, witch sure worked well. Even I did not trust Chemicons much, the bad batches of them are only the KZG, KZJ, TMV and TMZ series, not the KZE. And on top of that, they are nicely green:



However all that is not going to stop the instability of chipset, witch power voltage is "filtered" Chhsi terrible cap...



But all it all this looks like a decent Vcore design (for 2001):



560uF 4V Fujitsu polymers and 2700uF 6.3V Chhsi terrible caps is almone relatively qualite Vcore filtering, unless they start to break down, of course:



For rams and USB ports voltage filtering are used these bad caps Tayeh:



Of course the big problem is, when you push to big and heated coil a capacitor. This cap is really having a troubles, when the whole cascade is start to overhat:



However as you can see, bad caps are bulging even when they are long away from all typically overheating componets, like coils and mosfets:



MSI 6340 v1
--------------
2x 4700uF 6.3V d12.5 (Chemi-con KZE) -> 2x 4700uF 6.3V Panny FM d12.5 - P12347-ND
4x 2700uF 6.3V d10 (Chhsi) -> 4x 3300uF 6.3V Samxon GC d10
2x 820uF 4V d10 (Fujitsu) -> 2x 2700uF 2.5V Samxon ULR d10
2x 560uF 4V d8 (Fujitsu) -> 3x 1000uF 4V Samxon ULR d8
6x 1000uF 6.3V d8 (Chhsi) -> 6x 1000uF 6.3V Samxon GC d8
3x 330uF 6.3V d6.3 (Tayeh) -> 3x 470uF 6.3V Samxon GD d6.3
2x 47uF 16V d5 SMD -> 2x 47uF 16V Panny FK SMD d5 (16V) - PCE3397CT-ND
4x 10uF 16V d4 SMD -> 4x 10uF 16V Panny X5R SMD ceramic (4V) - PCC216CT-ND

(one polymer I added near the CPU, because it was removed and these 47 and 10uF SMD little caps I did not yet replaced, as I did not have anything to replace them with ATM)

[trodas]

So years ago I already gather for this recap the caps, witch give me Big Pope - at lest these Samxon caps. Adding the 1000uF 4V cap increased the total Vcore output capacity to 18 300mF in 8 caps! (for example the famous Socket 462 mobo DFI Lanparty B have Vcore output wth only 4 caps and 13 200mF total capacity - and no polymers!)



Whole look at the Vcore part of the mainboard after recap:



Look at the bottom caps from Vcore and for the AGP powering:



Much smaller todays polymers (560uF was - 1000uF is) for the same voltage is really just small "bits" compared to these 3300uF Samxon GC caps near them:



On the other hand, a 2700uF Samxon URL polymers are quite big pieces of caps:



And at the end, whole look at the MSI 6340 mainboard - little mobo:




Fun fact - after powering the mobo for the first time it show up, that for whole five years the real time clock is running. And on top of that, it even show good time - only +30min, witch is for 5 years w/o usage and with desoldered caps something amazing, I did not expect that And mobo is working quite well after the recap, even that the new CPU-Z version does not detect FSB, witch is weird. But the CPU-Z autor is already asked for some info to fix this, so there is a hope that this get fixed:
http://valid.canardpc.com/vblm4v
FSB 0 is not looking really truthfully
Also is worth noting, that the mobo has a pretty detailed setup in the bios (especially considering that this is VIA KT133 chipset and SDRAMs) and even overclock possibilities, where one can choose from Default setting (100MHz) to 117MHz FSB (37MHz for PCI). However the result from this last settings (with so quality caps I did not expected and problem so I tried this right away) is, that CPU is working at 256MHz when using this setting (30x7.5) Well, there is a good deal of fun with this little MSI mobo...

[bluetooth]

Too much baed aps

[Tweakurr]

If the time was almost correct after several years of inactivity, you had a really good battery on that board.

[trodas]

bluetooth - Indeed, there is pretty much all caps that are bad... except the Fujitsu polymers :) But that is why I recapped it. However I also hoped for some overclocking, when I unlocked it in BIOS! And that is where it fail...


Tweakurr - indeed I was very surprised about it. But that is what happend. Now I seeking to get a better bios for is, to support AXP CPU's, because it looks like that when the AXP is "unknown CPU", then the performance suffer bellow Athlon speeds...!

wPrime 32M test on the very same hardware - a ancient MSI 6340 mobo, same ram settings, same FSB, same everything: except for a CPUs!

Athlon 900MHz: 182sec

Athlon XP 1100MHz: 228 sec


...now could anyone explain me, how 200MHz slower CPU with half the cache could be 46sec faster?! WTF!


http://hwbot.org/submission/2436983_...9_182sec_972ms

Barton core suxx so much, compared to Thunderbird A4 core or wPrime have probem with Barton CPU's, killing their score somehow?!



...

Similary weird is the MaxxMem score, even there is the difference minimal:

Athlon 900MHz: 17.1 Marks

Athlon XP 1100MHz: 16.4 Marks




Lack of detection from the mainboard cannot (IMHO) answer this, not to mention the slow-down is minimal, not 46sec more on 182sec score...!

[Tweakurr]

I had not used that version of prime test before, so I downloaded the latest version for a comparison. The 32m test completed in 6.869 seconds, so I'm not sure we are using a comparable version unless your board is really that old. I recall my Asrock 939 would take 30 mins. or more to run the original 32M Prime test designed back in 90's. Now my system takes less than 8 minutes to run all 24 iterations.


I suspect the bios of your MB can not take advantage of the extra bandwidth of the higher cpu as it may not be able to address all lanes available, or it is designed for a board that has more lanes. The OS driver could also be at fault. Maybe run some DOS based benchmark and see if the results are the same to eliminate driver problem.