• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Cutting bridges (Epox 8KTA3 won't go to 133Mhz)?

A

Arbiter

Member
I'm building a system for the first time, and yesterday the last component (FOP-32) arrived and I was all set to build. Got my 1.2G T-bird up and running at 12x100 on my Epox 8KTA3. Was real nervous about the HSF install but was psyched to see the CPU temp was only reported at highest 34C in the BIOS once up. Eventually tried to go to 133Mhz by first lowering the multiplier to 9, power down, set jumper from 100 to 133, reboot and... NOTHING. Major downer. Tried two vid cards, tried other multipliers, no go. The POST LEDs stayed at FF, which tells me POST never got running anyway. Strangely, the power toggle button and reset button stop working as well - have to manually toggle the PS at the back.So I drop into forums, do some rooting around, and find out the 1.2 is trying to boot at 12x133 temporarily and shutting down, despite BIOS settings. Yes, I've tried bumping up the voltage - no go. Memory not the problem (Crucial CL2 PC133). It appears I can wait for a new BIOS, or cut the bridges myself. I've always wanted an excuse to buy a dremel... 🙂 - Can somebody explain to me clearly what bridges need to be cut? - This just changes the -default- multiplier the CPU will come up at, it doesn't permanently set it to that multiplier does it? - Has anyone heard from Epox regarding a new BIOS to address this?Thanks a bunch!
 
It appears I can wait for a new BIOS, or cut the bridges myself. I've always wanted an excuse to buy a dremel...

Oooo.... I've read that that is hard to do! Some have managed to do it with the smallest engraving tool (preferably diamond) and then had to modify even that... 🙁

You may want to check out this article. Not sure if how to do it was here but I did see somewhere some photos of someone who had done it...
 
Arbiter, you may not have to cut the bridges. Several BIOS options are not compatible between 100 and 133. Make sure that you set the BIOS options to Default before you try switching. Connecting your L7s to get the extra voltage kick on power up will also help.

Try setting your mobo at 100fsb and use HSTCLCK+PCICLK to test your memory at 133. If this works, make sure you reset back to default before setting the FSB to 133.

BTW, the Dremel is NOT used to cut the bridge. The Dremel is used to modify a diamond tipped 'burr' into a chisel type point. The modified 'burr' is then mouted in a handle and is used to manually cut the bridge. John Carr has posted links to his site which details the methods. The 4X multiplier on the L6 bridges would be the one to cut.
 
DaddyG, have also tried using the default BIOS setting for multiplier, as well as explicitly set to 9x - no go. I seem to be having the exact same problems that 'Droo' saw on the amdzone forums.

> Try setting your mobo at 100fsb and use HSTCLCK+PCICLK to test your memory at 133.

Can you explain that one? Not familiar with 'HST' clock, and what the memory has to do with the PCI clock? The memory is 384Mb Crucial CL2 PC133. I noticed when I boot at 100Mhz that the memory is reported in BIOS as CAS3 and 100Mhz; didn't seem right. However, I am fairly certain memory is not the problem - the POST LEDs indicate that the DRAM is first touched at 'C1', and I can see it go through several codes before reaching C1 when I boot at 100Mhz. When I boot at 133, the POST LEDs start at and do not budge from FF - indicates to me the POST is stopped dead at the gate before even going out to size and init memory.

> BTW, the Dremel is NOT used to cut the bridge. The Dremel is used...

Yep, understood, have seen John's website.

> The 4X multiplier on the L6 bridges would be the one to cut.

Thanks for the help DaddyG, appreciate it.
 
Arbiter, HSTCLK+PCICLK is a term used to set the memory at 133, with the FSB at 100. Some mobos use different terms but they mean the same. Maybe yours has FSB+33, its the same.

I have another idea. Instead of trying to cut the L6, try messing with the L1 brides. Can't remember the combos but, certain multipliers are not available if all the L1s are not connected properly. You only need to take off like 2X and your cpu will probably post OK at 8X default. Give it a try.
 
DaddyG,
The L1's are already unlocked (all connected), so I'd have to cut one of them right? I'm also a little confused, from looking at diagrams of the L6 and such, what you mean by '4x' on the L6 and '2x' on the L1.These are what they look like:L1: . | | | |L6: . | : | |

[edit] I read and reread the Droo thread, and finally was able to follow what you and John were talking about (his website certainly requires a little concentration!), and I understand now that she cut the last bridge and if I do so, it will put mine at 800. So now I have to decide if I want to do the cut, cross fingers for BIOS update, or RMA the board and roll the dice again with a new one. I'm afraid I don't have the patience for the last two and too much of an attraction towards power tools...
 
Thanks Arbiter, I'll check the thread. The reason I ask is because I'm thinking of exchanging my KT7-Raid (w/Tbird800@1GHz) for a KT7A-Raid. It shouldn't be a problem in my case though, since this CPU should do 1064MHz for a few millisecconds...
 
Arbiter, you wrote..."...finally was able to follow what you and John were talking about (his website certainly requires a little concentration!), and I understand now that she cut the last bridge and if I do so,it will put mine at 800"

Referencing...."and I understand now"....that's the whole point of how the site is structured....we've done our job...you don't need the site anymore 'cause you understand the 4 bit binary control logic and circuits. (Now you know why we hate incomprehensible settings tables).
Thanks for concentrating and confirming,;-)
John C.
 
Arbiter, sorry, from some reason I thought you had a 1 gig BIRD not an unlocked 1.2. Makes little difference since a bridge is a bridge. I'd suggest you have patience, yes the same thing is happening to ABIT as well, but they seem to be a little more tolerant. You might want to RMA your mobo and see if you can get a mobo with dips. ASUS and Gigabyte have dips and possibly Soyo. So far, I only have a theory that dips will solve the problem, but if you followed Johns site, you'll see that on the ASUS, opening two of the switches effectively disconnects the L2,L3 and L6 and replaces the Binary value with the dip setting. I don't see any way that the cpu could initialize at the default multiplier under this arrangement.

Another option is to wait for a response from EPOX, hacking up bridges on a Duron is OK but 1.2gig BIRDS are kinda expensive.

The IWILL KK266 also uses BIOS overclocking but has conservative memory settings and seems to do OK with 1.2 BIRDS. IWILL also uses version A1 of the Northbridge, all version A0's were bought up by the big guys. Version A1 seems to have a slight edge over A0.
 
Arbiter, I think I need to clarify the L6 bridge settings.

L6 bridge settings of '0' (all L6 connected) thu '1.5' get a 11X adder. Looking at your L6s we see that you have 1X, add the 11X and you get 12X or 1.2 gig.

L6 Bridge settings of '2' thru '7.5' (all L6 open) get a 3X adder.

Now, 'DROO' has a 1.1 BIRD, multiplier '0' (all L6s closed). By opening the 4X multiplier, the chip became 4X plus the 3X adder = 7X.

So, opening the 4X L6 bridge (FID 3) on your cpu will result in the following : 4X plus 1X plus the 3X adder = 8X.

One alternative to cutting the FID 3 bridge would be to insulate the corresponding pin on your cpu. You can do this with clear nail polish. Make sure it dries well, before inserting your cpu. Your system should then boot up at 800 in 100FSB mode. Change the FSB to 133 and you should be OK. A Digital Multimeter will help confirm the correct pin on the cpu.
 
Didn't understand the L6 bridges until I saw the Athlon spec on the FID, I'm very familiar with binary tables but the table finally clued me in that the scale doesn't just simply start at 0 and go up!

I might just try that nail polish trick, only question is, at the temps we're talking, could the polish possibly soften and permanently ruin that pin connection? Just want to be aware of what I'm getting into.

 
Arbiter,
2 other options...the diamond edged tool to cut ceramic trench, see p2 at http://members.nbci.com/candjac/index.htm Duron OC article

Or carefully open a trace from the appropriate socket pin, mobo backside. Site above has cpu pinout pic which can ID FID3 socket. As you know, don't cut...to avoid runners underneath surface...scrape carefully...opening traces has been done, see link to Cyrano's site,
also on page 2.
Le Vieux
 
Arbiter,

Don't think the temps will be a problem with the nail polish. Another option could be Shellac, I'm pretty sure that Shellac was used to insulate transformer wires in the good old days. I'd only cut the mobo trace as a last resort.
 
Thanks to both of you guys for the suggestions. I think I'll give the nail polish a try, it seems the least risky. At least if I can convince myself with a hair dryer and a test run that it won't melt or anything! 🙂 (When you say shellac, are you talking plain shellac or is there some specialized product called 'Shellac'?)

Instead of cutting a trace on the mobo, couldn't I follow it to the multiplier control circuit and do a pin lift? Seems easier to repair. Anyway, now I have an excuse to go out and buy a digital multimeter...

John, question: when you talk about 'ringing' out those boards on your web page, is that essentially probing around looking for 0 ohm connections and mapping out the CPU pins to where they end up on the board and bridges?
 
Arbiter, sorry I should have just said shellac. BTW, 'ringing' is an old term, original meters had an audible alarm for zero ohms.
 
Arbiter,
Yes, as DaddyG said...ringing = 0 ohms, (didn't realize we were so old);-) And yes that's how we "verified" all the cpu control circuits, L1, L3, L4, L6, L7, bridges. We old-timers try never to take anything for granted. In Hindi there's a saying..."Satan' nahi' sota'"..."The devil never sleeps".

"Pin lift" does same as trace break...but have you considered the easy way to cut bridges with the hand tool mod to a diamond tipped burr, described and linked at our site...page 2?? Cuts as though the ceramic was butter. Kit has 20 pcs, write off the cost for Dremel type work on other projects.

Final tip...don't place probes on any bridging material such as the defogger repair kit or silver conductive inks, as that doesn't check the connection below to open bridge ends...and probing material will destroy the fragile dried material. Always probe for good connection across painted bridges from respective pins to L3/L4 resistors or to ground pins, all identified with pics/diagrams/etc at site.
John C.

 
I have considered the cutting of the bridge, in fact I have the Dremel (already had fun cutting extra air holes in the front of the case). I also bought a small set of two diamond tips from Sears, they were only a few bucks and I figgered worth a shot. Haven't seriously used them yet, only enough to find that one cutting the other isn't a quick job. One has a round bulbous head, the other is pencil-tip shaped. Not sure that one can file the other down though, but will give it a shot and maybe try the kit otherwise.

My point with lifting the pin is that I can always lower the pin and RMA if need be; cutting a trace or a bridge is a little more permanent. Thanks for the ringing advice.

I'll let you guys know how it all came out in the next few days...
 
Someone please clue me ...as I guess I am clueless.
I thought that the KT133 chip was not designed to go to 133. That is what I have on my MSI Pro 2, and I now have it at 105x8 (650 Duron).
Seems like I read somewhere that the 1/3 the FSB rule does not apply with this board. I don't wnat to go too high as to damage my periherials (Maxtor 8.3 DMA33, ATI Rage Fury 16 Mg.)
Neos
 
You are correct Neos; the KT133 only supports 100Mhz on the FSB. The chipset we're talking about though is the new KT133A, which does support a 133Mhz FSB. I believe when set to 133Mhz it automatically divides the clock a bit more (to 1/4) to keep the peripherals happy.
 
Where did you read that when the 1.2 T-bird set at 9x133 boots as 12x133 ?? This was believed to occur with all T-birds, however if you go at www.overclockers.com you'll see that it sais that this is true only for locked cpu's i.e cpu's with no L1 bridges. Then the default multiplier will be used at boot no matter what the bios seting is. However the 1.2 T-birds are unlocked, so that should not be the case, at least not according to that article.
Have you tried 8.5x133 ? Does this work ?
Moreover, there are also rumours that VIA changed something in the manufacturing process of the KT133A chipset at about week 52 if i remember correct. What is the week of your motherboards chipset.
Finally www.overclockers.com sais that in many instances cpu's that worked at fsb 133 on one motherboard didn't work on another and recommends a motherboard change in situations where one cpu can work at a given MHz level at 100 fsb but not at 133 fsb (with different multipliers usually).
From what i have read in this thread i don't see why the problem would not lie on the motherboard or the memory rather than the cpu.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top