- Joined
- Jan 3, 2007
- Messages
- 454 (0.07/day)
| System Name | celer |
|---|---|
| Processor | Pentium 4 650 3.4GHz 2MB L2 |
| Motherboard | MSI PM8M3-V |
| Cooling | Thermalright SI-128 SE |
| Memory | 2048MB OCZ 2-3-2-5 2T at 200Mhz |
| Video Card(s) | Sapphire Radeon R9100 |
| Storage | 250G Samsung 850 PRO (MZ-7KE256BW) - 1024G WD Black (WD1003FZEX) |
| Display(s) | 19' iiyama ProLite E1980SD 1280x1024 75Hz DVI |
| Case | Eurocase moded |
| Audio Device(s) | onboard |
| Power Supply | Enermax 620W Liberty |
| Mouse | Logitech MX510 red |
| Keyboard | eTech PS/2 keyboard |
| Software | Win XP SP3 |
| Benchmark Scores | http://hwbot.org/submission/2455634_ |
MSI 6163 is for a MSI board pretty unusual thing. It is still a mere old Slot 1 board, but in it's time it has a very modern design and with surprising overclocking features. It does support Pentium II and Pentium III (!) CPU's up to 700Mhz. ( http://global.msi.com.tw/index.php?func=proddesc&prod_no=340&maincat_no=1 ) I got a revision 2.0 of the board, so it should finally work well 
What unfortunately stays like classic MSI is the usage of bad caps. JPcons in pictured mainboard and in the exact same mobo (also revision 2.0) are Chhsi caps. Chhsi caps are also known bad caps and the mobo ended up like that - bad. ( http://i4.photobucket.com/albums/y134/trodas/MSI6163withChhsicaps.jpg -=- http://i4.photobucket.com/albums/y134/trodas/MSI6163withChhsicaps2.jpg ) Upper mosfet desoldered from the board and fried in the process. It took away even the blocking diode that is parallel to the main Vcore coil... And mech13 is definitively not alone, who is having problems with this mobo, thanks to the bad caps, just check there (czech): http://www.svethardware.cz/disc_doc-005111219273077183262DBC23098A5.html
(english): http://www.badcaps.net/forum/showthread.php?t=1222
On the another side, Vcore regulator is very well designed and if you did not torture him by Pentium III CPU's and you get along with not overclocked Celerons - well, then it might run for you for some time. Of course this does not fix the bad caps and the board is designed for overclocking. And this is very tempting. In bios you can setup FSB from 66Mhz to 155Mhz!
This is exceptional for a Slot 1 mainboard. And it is probably the reason (the exceptional overclockability of the Intel BX440 chipset as well) why MSI build such nice mobo with so well designed - sadly still only one phase - Vcore regulator and did not skip any caps
This should be reason to thank MSI for, as that is a very rare sight indeed.
On the other hand, what is very very common is bad caps. Whole mobo is full on bad caps - JPcon. (another one is full of Chhsi caps anyway)
And to make matters worse, then input caps is only bigger size, not capacity! All are 1000uF. Only the input ones are 10V, witch "really helps" when the CPU Vcore is powered from 5V rail. Too few uF on the input, right? And the tiny wiring on the input coil, well, this is definitively not designed for Pentium III overclocking! Witch is probably why on the mobo with Chhsi caps it almost burn to black even with the core of the coil... (Pentium III was overclocked in the board by mech13)
On South Bridge you can find a beautiful sticker - probably instead of heatsink On the other hand, all swiches and connectors are properly labeled and described. At least something.
Bios is tested AWARD bios (this does no mean that MSI did not add some kinks to it, one being default enabled bios flash protection - if you change it to disabled, be ready, all your bios settings are lost and never ever remembered... including this particular setting) and there slots for SDRAMs accept up to 384MB of ram (3x 128MB) - or up to 768MB if you use 256MB SDRAMs with ECC. Where to get such ones, hell, I don't know From my testing 512MB one it took only the 128MB. And this laughable small NB heatsink will be necessary to change, if you plan run on FSB 155Mhz for sure
So, there we have an overclocking mainboard with bad caps. What to do? Of course, we replace the caps for a good Samxons! Let's start with Vcore regulator, where it is most need. Cap right in the very front should be d10. Next time
Around rams and chipset is is also very important - at least if we want higher FSB, then it is not going to be any other way.
Input caps should be much bigger. 2x 1000uF?! You kidding me? Let's use 2x 3300uF!
Good capacity of Vcore input caps play major role in the Vcore stabilization, so, we better overshoot this a little more, nothing bad could happen here. The more the better.
That of course does not mean that we neglect other caps. AGP cards like clean, no rippled voltage too.
Neighbourhood of South Bridge is looking very nicely color full, after recap. I don't think that for running at FSB 155Mhz will be necessary to add a heatsink, but I could be wrong. Maybe when the load is higher, maybe then...
Small caps are not good to omit. There are only annoying things happening because of that. Some kink there, some elsewhere. All because of not recapped small caps, so, I say - replace them all. Samxon GK 22uF 25V is a good choice.
Look at the recapped Vcore regulator - there is a slight problem with the much bigger Vcore input caps - they are getting too close to the Slot 1 space and with the d10 caps in place the Pentium II Klamath barely make it.
But it made good contact and run well as well, as other CPU's I tried in the board. Mendocinos are lazy a$$ses, however even in the reduction they can be clocked pretty well in this board. From 466Mhz to 583Mhz:
(notice that the Vcore is increased to 2.40V and that SpeedFan reporting is exactly like that - that means the sensors are good/well calibrated and the regulation is exceptionally good - besides being still only one phase - very likely thanks to Samxon GC caps)
So that is what I begin overclocking with the board. The little Mendocino ended up at 90 x 7 - 635Mhz: http://valid.x86-secret.com/show_oc.php?id=351368
Then I took a Celeron Covington without L2 cache (!) and he made it from 266 to 460Mhz: http://valid.x86-secret.com/show_oc.php?id=351231
And for the 3rd time - Pentium II Klamath 233Mhz got raised FSB from 66 to 124Mhz (!) and ended up at 371Mhz: http://valid.x86-secret.com/show_oc.php?id=350966
(CPU-Z wrongly identify it as 200Mhz one, probably because the multiplier is set to the lowest 3x and not to 3.5x as it should be)
Old good Pentium II Klamath has a serious overclocking problem with L2 cache. There are used 7nS chips witch top at somewhere near 200Mhz. Frequency divides for 233Mhz ones is 1,75x and for 266Mhz ones is 2x. It is pretty obvious witch one will get clocked much better because of that. Unfortunately I had the 233Mhz version and then even the brutal voltage increase from 2.80V to 3.40V (!!!) does not help single bit. 371 / 1.75 is 212 and 212Mhz is absolute top for 7nS chips.
Instead, limit of Celeron Covington is purely in the bios itself. Like Mendocino, from default 2.00V it can be increased only to 2.40V...! Now that is very serious disadvantage of this great (for Slot 1) board.
If someone can remove these limits from the bios, then there could be some records broken with this mobo To make my point, the Celeron Covington w/o L2 cache made for me w/o modifications 448Mhz ( http://valid.x86-secret.com/show_oc.php?id=350945 ) and with modifications like lapping the core to the cooper, using Arctic Silver 3 TIM instead of the rubber crap and replacing 18pcs of the ceramic caps on it's Slot 1 board that stabilize the Vcore with new, better and bigger ceramic caps + adding the two ones missing I made the CPU work at 460Mhz.
This might sound as great improve, but it is not. It is just one step up in the FSB. Either 112 or 115Mhz. On 117 it always reset when Windows booting...
This Celeron used 6pcs of 2.56uF ceramic caps + 12pcs of 1uF ceramic caps for add voltage filtering.
All that was replaced with Panasonic X5R 5.6uF 6.3V (PCC1937CT-ND) ceramics caps + two missing ones are added. And because the overclocking result nearly not increased at all, it means that the Vcore regulation is excellent and stable and make it even when overclocking hardly.
Let's hope it made it even when his owner mech13 start torturing the board with Pentium III.
What unfortunately stays like classic MSI is the usage of bad caps. JPcons in pictured mainboard and in the exact same mobo (also revision 2.0) are Chhsi caps. Chhsi caps are also known bad caps and the mobo ended up like that - bad. ( http://i4.photobucket.com/albums/y134/trodas/MSI6163withChhsicaps.jpg -=- http://i4.photobucket.com/albums/y134/trodas/MSI6163withChhsicaps2.jpg ) Upper mosfet desoldered from the board and fried in the process. It took away even the blocking diode that is parallel to the main Vcore coil... And mech13 is definitively not alone, who is having problems with this mobo, thanks to the bad caps, just check there (czech): http://www.svethardware.cz/disc_doc-005111219273077183262DBC23098A5.html
(english): http://www.badcaps.net/forum/showthread.php?t=1222
On the another side, Vcore regulator is very well designed and if you did not torture him by Pentium III CPU's and you get along with not overclocked Celerons - well, then it might run for you for some time. Of course this does not fix the bad caps and the board is designed for overclocking. And this is very tempting. In bios you can setup FSB from 66Mhz to 155Mhz!
This is exceptional for a Slot 1 mainboard. And it is probably the reason (the exceptional overclockability of the Intel BX440 chipset as well) why MSI build such nice mobo with so well designed - sadly still only one phase - Vcore regulator and did not skip any caps
This should be reason to thank MSI for, as that is a very rare sight indeed.
On the other hand, what is very very common is bad caps. Whole mobo is full on bad caps - JPcon. (another one is full of Chhsi caps anyway)
And to make matters worse, then input caps is only bigger size, not capacity! All are 1000uF. Only the input ones are 10V, witch "really helps" when the CPU Vcore is powered from 5V rail. Too few uF on the input, right? And the tiny wiring on the input coil, well, this is definitively not designed for Pentium III overclocking! Witch is probably why on the mobo with Chhsi caps it almost burn to black even with the core of the coil... (Pentium III was overclocked in the board by mech13)
On South Bridge you can find a beautiful sticker - probably instead of heatsink
On the other hand, all swiches and connectors are properly labeled and described. At least something.
Bios is tested AWARD bios (this does no mean that MSI did not add some kinks to it, one being default enabled bios flash protection - if you change it to disabled, be ready, all your bios settings are lost and never ever remembered... including this particular setting) and there slots for SDRAMs accept up to 384MB of ram (3x 128MB) - or up to 768MB if you use 256MB SDRAMs with ECC. Where to get such ones, hell, I don't know
From my testing 512MB one it took only the 128MB. And this laughable small NB heatsink will be necessary to change, if you plan run on FSB 155Mhz for sure
So, there we have an overclocking mainboard with bad caps. What to do? Of course, we replace the caps for a good Samxons! Let's start with Vcore regulator, where it is most need. Cap right in the very front should be d10. Next time
Around rams and chipset is is also very important - at least if we want higher FSB, then it is not going to be any other way.
Input caps should be much bigger. 2x 1000uF?! You kidding me? Let's use 2x 3300uF!
Good capacity of Vcore input caps play major role in the Vcore stabilization, so, we better overshoot this a little more, nothing bad could happen here. The more the better.
That of course does not mean that we neglect other caps. AGP cards like clean, no rippled voltage too.
Neighbourhood of South Bridge is looking very nicely color full, after recap. I don't think that for running at FSB 155Mhz will be necessary to add a heatsink, but I could be wrong. Maybe when the load is higher, maybe then...
Small caps are not good to omit. There are only annoying things happening because of that. Some kink there, some elsewhere. All because of not recapped small caps, so, I say - replace them all. Samxon GK 22uF 25V is a good choice.
Look at the recapped Vcore regulator - there is a slight problem with the much bigger Vcore input caps - they are getting too close to the Slot 1 space and with the d10 caps in place the Pentium II Klamath barely make it.
But it made good contact and run well as well, as other CPU's I tried in the board. Mendocinos are lazy a$$ses, however even in the reduction they can be clocked pretty well in this board. From 466Mhz to 583Mhz:
(notice that the Vcore is increased to 2.40V and that SpeedFan reporting is exactly like that - that means the sensors are good/well calibrated and the regulation is exceptionally good - besides being still only one phase - very likely thanks to Samxon GC caps)
So that is what I begin overclocking with the board. The little Mendocino ended up at 90 x 7 - 635Mhz: http://valid.x86-secret.com/show_oc.php?id=351368
Then I took a Celeron Covington without L2 cache (!) and he made it from 266 to 460Mhz: http://valid.x86-secret.com/show_oc.php?id=351231
And for the 3rd time - Pentium II Klamath 233Mhz got raised FSB from 66 to 124Mhz (!) and ended up at 371Mhz: http://valid.x86-secret.com/show_oc.php?id=350966
(CPU-Z wrongly identify it as 200Mhz one, probably because the multiplier is set to the lowest 3x and not to 3.5x as it should be)
Old good Pentium II Klamath has a serious overclocking problem with L2 cache. There are used 7nS chips witch top at somewhere near 200Mhz. Frequency divides for 233Mhz ones is 1,75x and for 266Mhz ones is 2x. It is pretty obvious witch one will get clocked much better because of that. Unfortunately I had the 233Mhz version and then even the brutal voltage increase from 2.80V to 3.40V (!!!) does not help single bit. 371 / 1.75 is 212 and 212Mhz is absolute top for 7nS chips.
Instead, limit of Celeron Covington is purely in the bios itself. Like Mendocino, from default 2.00V it can be increased only to 2.40V...! Now that is very serious disadvantage of this great (for Slot 1) board.
If someone can remove these limits from the bios, then there could be some records broken with this mobo
To make my point, the Celeron Covington w/o L2 cache made for me w/o modifications 448Mhz ( http://valid.x86-secret.com/show_oc.php?id=350945 ) and with modifications like lapping the core to the cooper, using Arctic Silver 3 TIM instead of the rubber crap and replacing 18pcs of the ceramic caps on it's Slot 1 board that stabilize the Vcore with new, better and bigger ceramic caps + adding the two ones missing I made the CPU work at 460Mhz.This might sound as great improve, but it is not. It is just one step up in the FSB. Either 112 or 115Mhz. On 117 it always reset when Windows booting...
This Celeron used 6pcs of 2.56uF ceramic caps + 12pcs of 1uF ceramic caps for add voltage filtering.
All that was replaced with Panasonic X5R 5.6uF 6.3V (PCC1937CT-ND) ceramics caps + two missing ones are added. And because the overclocking result nearly not increased at all, it means that the Vcore regulation is excellent and stable and make it even when overclocking hardly.
Let's hope it made it even when his owner mech13 start torturing the board with Pentium III.