K7SEM capacitor identification

[Scenic]

Re: K7SEM capacitor identification

do you still have that 6800?
a shitload of these got thrown out because one (!) mosfet shorted out (so you'd end up with a full short on the 12V side of the additional molex connector.. urgs)

fixed quite a bunch of these..

oh and... that description of the additional LAN port of the K7SEM (it's called AD-LAN btw) sounded kinda familiar... so i went digging..

guess what?
http://bambooz.pytalhost.net/badcaps/K7SEM/DSC02519.jpg
http://bambooz.pytalhost.net/badcaps/K7SEM/DSC02520.jpg
http://bambooz.pytalhost.net/badcaps/K7SEM/DSC02521.jpg

board was thrown out because no matter what RAM you put in it, it always puked out errors in memtest86. got it for free with a 1.8GHz Duron Applebred (1.5V Vcore )

OSTs on VRM out were bulging, all others looked fine.
mostly recapped, and it's rock solid again. to really get this b*tch running stable, you have to replace ALL the tiny 16V 100uF craps. this thing is riddled with tiny dry-up-G-Luxon's
i used LTec TK because i had a whole bag (500pcs) of these that i bought for next to nothing. their datasheet is kinda ... interesting..

Designed for consumer products, which require to withstand high temperature and high reliability.

so... these things withstand high reliability..!? wut..

oh and removing the northbridge heatsink, getting the double sided tape crap off and putting it back with thermal glue also helps A LOT..
that tape stuff seems to be more of an insulator than anything else..

this is one of the rare occasions where i reused old good brand caps from dead (asus socket A) boards. didn't wanna waste money on this thing. turned out to work just fine..


edit: as for the GF6800.. the mosfet labeled Q509 failed shorted
http://bambooz.pytalhost.net/badcaps/6800/01.jpg
http://bambooz.pytalhost.net/badcaps/6800/02.jpg
http://bambooz.pytalhost.net/badcaps/6800/location.jpg

got a few of them (old pic)
http://bambooz.pytalhost.net/badcaps/GF6800_madness.jpg
(note that the 6800GT (the one with the blue cooler on it) has the exact same PCB layout as the 6800 Ultra just missing some parts)

and here's my thread about the fried mosfet (posted 2 years ago.. god how time flies)
https://www.badcaps.net/forum/showthread.php?p=60783

[xrror]

Re: K7SEM capacitor identification

Some "raw research" here. The clockgen is software programmable. Which shouldn't be new info since the HoneyX bios page "unibios" alludes to this. It's the normal "rediscovering what was already forgotten".

you probably will want download and run cpu-z so you know what's going on. And especially to see what is actually being set.

CPU-Z:
http://www.cpuid.com/softwares/cpu-z.html

You will need setFSB.

setFSB:
http://www13.plala.or.jp/setfsb/

Unfortunatly, the ICS 94208EF isn't directly supported in setFSB. But setting it to ICS94225AF you can mess around with the clockgen

the settings won't be right, but here's what i have so far. note I'm running the 133/33.3/133 setting which sets a 1/4 divider. This is all very preliminary, but I had to share.

In setFSB, select ICS94225AF. Click "Get FSB"
You will get what looks like correct settings, but they're not. However it does work.

So far, here's what I've got:

133=91fsb
150=100
166=111
190=126
200=133
204=136

210 reset, cause I still have the Palomino in here, and well pally's suck for overclocking

Don't you love cliffhangers? Pretty sure it's the processor, since this is a 2000+ with 12.5 mult. so at 133 it's at 1666. Weirdly all pally i've had tank over 1.7. So I'll be back with more results when I swap it for a Throughbred. B. heh.

In any case this is more a preview. Having it set in the BIOS is infinitly more useful than trying to software set it. Though I must admit so far, this board has been totally rock solid with me arbitrarily changing clocks, even stupid huge jumps (some boards hang if you make a very large change, like 20+ mhz bus change) so this might end up being be a valid option.

Still, it'd be nice for it to set in bios to make it a "set it and forget it" and OS-agnostic.

And sorry to spam this thread. Let me know if it annoys.

[xrror]

Re: K7SEM capacitor identification

Times like this remind me of just how awesome the T-Bred B was when it came out. Anyway, time to fill out the numbers.

210=140fsb
215=143
220=147
225=150
230=153

It then reset. note that i've been posting all this from the K7SEM. I'm probably running into the limits of the 512mb SDRAM stick I have. I seriously doubt it was ever intended for this. But at a CPU speed of 1875 as I type this, K7SEM rev1.0 isn't so bad =D No idea if this is actually load stable though.

So there's the preview. Gonna go see how bad it dies on 3dMark01

added: oh, yea the t-bredB i'm using just happens to be a 2000+ also. So it still has the 12.5 multi. As if anyone wondered =P

[Scenic]

Re: K7SEM capacitor identification

i don't have that board set up right now (6pm and i've got a really bad headache.. just after the weekend.. go figure )... but i might try it out tomorrow.
would be kinda interesting if the (used) Rubycons and other small caps make a difference in overclockability

[xrror]

Re: K7SEM capacitor identification

Just a FYI. 148mhz bus seems to be a sweet spot\OC wall for this board. Three different video cards (with known AGP overspec tolerance) all behave the same, regardless of sideband/fastwrites agp speed settings.

Taking a guess from the rebelshaven posts about strange PCI delay settings and other odd voodoo for getting 150mhz to work "when the phase of the moon is right" I think 150 is just outside the stable envelope for the chipset.

Considering i'm running 148 rock solid, with PCI speed (assumed) at 37mhz vs. 30 as what he might be running, I think it's just a chipset timing issue.

Normally in a better featured BIOS, you could just turn down the aggressiveness of the internal chipset timings, or even change FSB strap. But this is ECS so you get what you get.

So I think 145~148 is max at whatever internal timings are default. So if there is a modded BIOS, instead of a 150mhz setting 148/37/148 would be better.

Tomorrow I'll try some different combinations of vid cards/memory to make sure. I also want to see how badly I can roast the IDE controller if i use setFSB and START from the 100/33.3/100 BIOS setting lol. 1/3 divider here we come! It'll be like the old Intel BX days! 47mhz PCI is good for what ails ya! And everyone loves an AGP speed in the 90's right? errr....

Well time for me to call it a day. Good times this.

[xrror]

Re: K7SEM capacitor identification

Is there any further interest in this topic? I don't want to (further) spam it. I realize that badcaps.net is focused on the capacitors, so I suppose whatever result I get after the board functions isn't too relevant to this forum.

If I can actually FIND my Epox 8RDA (non-plus) and it actually still works after re-capping I could see if increasing capacitance on the board helps overclocking. But that's another topic and another day when I actually find that board. In general I'm thinking that it does.

I do know on the Geforce 6800 I linked to, the added capacitors - even though they were overkill and Teapo did prevent the dreaded "black screen" when 6800np AGP cards were overclocked "too much" back in the day. But I think that card also used (linear???) power regulation. I'm out of my league here, since I realize that sounds pretty dumb since that's something you normally only see these days with some amplifiers, where unused power is just "bled off" as heat.

The Geforce 6800GS I modded later was definitely a "switching/PWM" design, with less beefy parts. It didn't like having "stupid overkill" capacitance added to it. I'm guessing that design is more like what most modern motherboards would use, so just putting gigantic caps on there would be bad.

But once again I'm going off topic. I'm also guessing deconstructor isn't going to come back to this thread.

[PCBONEZ]

Re: K7SEM capacitor identification

We don't starch our collars.
Can always open a new thread in a general discussion area.
.

[Uranium-235]

Re: K7SEM capacitor identification

Thanks for offering to look at the BIOS. I think this is the "non-B" one. When I get a chance I'll see if I can confirm it. If all else fails I can dump the bios with awdflash or whatever this board uses.

I need to check to see if it has the latest BIOS anyway, so hopefully later today I'll get that info.

@Uranium-235
I wish our goodwill even had fun stuff like that. Then again you can't find any old video game stuff for anything in thrift stores around here which is really annoying. And the pawn shops around here are all "ebay highest price" = ripoff.

I worked at the computer works (meh, volunteered). this was where all the computer donations for the entire metroplex went

somewhere on this forum I have a thread started where I have tons of pics. here it is

https://www.badcaps.net/forum/showthread.php?t=6676

[Pyr0Beast]

Re: K7SEM capacitor identification

But I think that card also used (linear???) power regulation.

Afaik no. All 6800 used switching supply for ram and core

Black-screen was linked more to general instability or power consumption

[pfrcom]

Re: K7SEM capacitor identification

Is there any further interest in this topic? I don't want to (further) spam it.

Please do, makes a refreshing change from "My monitor is broken" threads

Badcaps was originally oriented towards motherboards

[PCBONEZ]

Re: K7SEM capacitor identification

Please do, makes a refreshing change from "My monitor is broken" threads

Badcaps was originally oriented towards motherboards

I'm with ya buddy!
GO MOBO!
GO MOBO!
.

[linuxguru]

Re: K7SEM capacitor identification

Keep on POSTing (silly pun)...

I'm currently running a T'Bred Athlon XP/1800+ underclocked to 11.5 x 100 MHz in mine, for use as a quiet/HTPC box. The mobo is a K7SEM branded as a Kobian 730s FSMX (basically a PCChips M810 V7.1A) and has had various cap upgrades over the years, and everything works - except that it no longer retains CMOS on power down, and changing the battery has no effect. There must be a leaky cap or diode on the backup battery rail somewhere, but I'm too lazy to open it up once again and track it down...

Edit: Just noticed, 1000th post of mine.

[Pyr0Beast]

Re: K7SEM capacitor identification

Check diode somewhere near the battery

[linuxguru]

Re: K7SEM capacitor identification

Check diode somewhere near the battery

Thanks, I'll check them next time I get it open for an upgrade. I remember seeing 4-5 1N4148 or similar diodes near the battery, but didn't have the patience to remove and check them last time.

[linuxguru]

Re: K7SEM capacitor identification

board was thrown out because no matter what RAM you put in it, it always puked out errors in memtest86. got it for free with a 1.8GHz Duron Applebred (1.5V Vcore )

OSTs on VRM out were bulging, all others looked fine.
mostly recapped, and it's rock solid again. to really get this b*tch running stable, you have to replace ALL the tiny 16V 100uF craps. this thing is riddled with tiny dry-up-G-Luxon's

Aha - I changed a few of the 100uF caps, but there are several more small caps near the Southbridge which I left alone.

I replaced some of the 1000/6.3 caps near the Northbridge and DIMM slots with 180/16V Chemicon PSA polymers, and it's rock solid in spite of the much smaller value cap replacements. Vcore OSTs were replaced with new 3300uF/6.3 Nichcion HM(M)s, 2005 date-code.

[xrror]

Re: K7SEM capacitor identification

Sorry I dissappeared for four months everyone. Trying not to be a "flash in the pan" Long story short - I got another job. Or as I like to say, I managed to get a real job again after I was laid-off two years ago. So hopefully that means I have more time/money to revisit these old hardware projects!

I sure hope so anyway

do you still have that 6800?
a shitload of these got thrown out because one (!) mosfet shorted out (so you'd end up with a full short on the 12V side of the additional molex connector.. urgs)

fixed quite a bunch of these..

I just now read this fully. NICE Scenic. Yea I'd be more that willing to revisit the 6800 assuming I can find it's remains. The 12v dead short would explain why the power supply went into "suicide protection" mode instantly. I'm pretty sure my 6800 was significantly over design power after mods.

I really want to get back to messing with the K7SEM. If I remember 140bus was working great with a T-Bred B installed. I said "rock solid"... well as long as you didn't restart. It would hang at POST @ 140 which really sucks since everything else worked awesome.

[linuxguru]

Re: K7SEM capacitor identification

I really want to get back to messing with the K7SEM. If I remember 140bus was working great with a T-Bred B installed. ..

Long bump up - I just pulled my K7SEM out of the system for an upgrade. It's not a huge upgrade actually, just used a leftover Gigabyte 7VM333M-RZ with a T'bred-B Sempron 2400+ and 2x DDR DIMM slots to avoid any big issues with the original SW install. The Gigabyte board had a lot of cap upgrades done to it over time - today, I replaced 10x 1000uF/6.3 OST RLS (scattered at various places around the Northbridge and DIMM slots) with 10x 1500uF/10/8mm Rubycon ZLH (yes, it actually is 8mm). Installed, booted up the custom Fedora 4 (yup, that was the old install) and did a minor tweak to xorg.conf to use the Via driver for the Unichrome graphics, and everything just works like a charm with a nice bump up in speed and 1 GB of DDR 333.

That left me with a working K7SEM with a T'bred-B AThlon XP1800+, so that's what I'm going to tweak - not sure what it's going to be, but a quiet fanless PC or router is within reach. It's rock stable at 133, but 100 FSB gets close to the realm of fanless with a suitably large heatsink. I do understand that this is mostly obsolete in the era of fanless dual-core Atoms and APU/GPU combos, but it's mainly the fun aspect of taking a nearly 15-year old board that was widely derided at that time for being cheap and unreliable, and keeping it going doing something useful at little or no cost.