Unlocking 13+ Multipliers on a T-Bred-B

[SidVicious]

I am the proud owner of a 1800+ Thoroughbred-B CPU since last year, I've been running it stable @ 2GHz (160MHz x 12.5) on a 1.75 Vcore since day one.

Due to the limitations of the KT333 chipset on my A7V333-RAID rev 1.0, I've been unable to push the FSB beyond 160MHz, Prime95 Torture Test return errors after a few hours even when the CPU is clocked @ 833MHz (166MHz x 5). I know for sure that my memory modules are'nt causing those errors either, got two 512MB sticks of matched Samsung PC3200 TCC3 (2.5-3-3-8 by SPD @ 200MHz, they are running stable 2-2-2-5 @ 160MHz).

For financial reasons and because of the upcomming relase of Socket 939 and LGA755 CPUs & motherboard, I do not want to purchase any new core components but I feel the need to extract a few more MHz out of my CPU.

Problem is that the highest available multiplier is 12.5 and my chipset crap out whenever I exceed a 160MHz FSB. Do you guys know of any way to unlock a Thoroughbred-B so that the multipliers over 13 are available ?

My search for an answer left me with more questions, I read some references about re-wiring some pins or bridges but found no images to illustrate how such a thing can be done.


Fok Speling Misstake

 

[lwj81]

<A HREF="http://www.ocforums.com/showthread.php?s4b37ea7a0f29d8aeb8fac630c8473286&threadid=279392" target="_new">http://www.ocforums.com/showthread.php?s4b37ea7a0f29d8aeb8fac630c8473286&threadid=279392</A>

 

[phsstpok]

Click the following link. Look for the CPU and Socket diagram near the middle of the page. They illustrate two mods but neither of them is what you want. However, the diagram and legend will help you.

<A HREF="http://www.beachlink.com/candjac/Workarounds.htm" target="_new">http://www.beachlink.com/candjac/Workarounds.htm</A>

Basically, the mod you do want is to connect the 8X multiplier pin (AJ27) to VCore (nearby pins AK26 or AM26). This pulls the 8X bit HIGH, effectively enabling the high multipliers.

If you had an nForce2 board you mod just the one pin and you are in business. With other mobos there may be extra steps.

Here is the concern. By forcing the 8X bit HIGH you also change the default multiplier. If that's the only pin you change then your default 11.5X multiplier of your XP1800+ becomes undefined, a non-existent multiplier.

Some boards handle undefined multipliers by picking a default of their own. Other boards refuse to POST. If this is the case you have to mod one or more additional pins to produce a valid multiplier and still enable the 8X bit.

At this point you need a crash course in Athlon multipliers.

The link will get you started plus you can jump to <A HREF="http://www.beachlink.com/candjac/" target="_new">John Carcich's homepage</A> for detailed information.

If you have any questions I'll do my best to answer them. If I can't I'm sure someone else here can.

<b>56K, slow and steady does not win the race on internet!</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 04/06/04 11:46 PM.</EM></FONT></P>

 

[SidVicious]

Thanks a lot for the links, those feature more info than I tought I needed !

I read that I'll have to connect pin #AJ27 and #AH26 with fine wire u-loop in order to switch the 5th 8x bit from low to high.

This will turn

11.5X--O--C--C--C--C 1800+

into

03.5X--O--C--C--C--O Reserved (Remapped to 19X)

right ?

I have'nt found a single picture featuring a complete identification grid but figured out that the pins I need to connect with the wire loop are around the 4th empty "cross" when counting the pinless gaps from left to right. #AJ27 is "north" of the 4th "cross" and the closest Vcc/Vcore pin would be #AH26, the one just "southwest" of #AJ27, correct ?

I guess that I'll need to set a 100MHz FSB in the BIOS before attempting that trick, otherwise I'll end up with a 2533 or 3040MHz clockspeed if I set it at 133 or 160MHz.

BTW, any idea how my A7V333 will behave ? Will I be able to access all of the multipliers above 13 ?

Fok Speling Misstake

 

[TheRod]

<A HREF="http://www.ocinside.de/go_e.html?/html/workshop/pinmod/amd_pinmod.html" target="_new">Ta dam!</A>

--
Would you buy a potato powered chipset?

 

[SidVicious]

That sure is a great tool, I'm adding it to my Favourites on the spot !

OCinside forgot to include the #AJ27 8x bit to #AH26 Vcc upper multiplier unlock wire mod I was looking for.

The mod they show involve grounding #AJ27 with #AH28, that trick unlock the lower multipliers for CPUs that are limited to the upper multipliers.

I'll be doing the mod this evening, I'll keep you guys posted, thanks again !

Fok Speling Misstake

 

[SidVicious]

My experiment with the wire mod yield no results, I tied hair-thin copper wire from an old headset around #AJ27 and #AH26 and made sure it held into place before installing the CPU and HSF.

My A7V333 just booted as it usually do, nothing changed, I'm still stuck with a maximum multiplier of 12.5 in the BIOS.

I am too tired and frustrated to give it an other try.

I'll try connecting #AJ27 with another Vcc pin tomorrow evening.

Feel free to reply, your comments would be appreciated !

Fok Speling Misstake

 

[phsstpok]

Oops, I made a couple of mistakes! Glad you were paying attention!!!

- 19X. That's right. I forgot you weren't starting with an XP1700+ like I did. Mod'ing the 8x bit to HIGH leaves me with an undefined multiplier but leaves you with 19x.

- For convenient Vcore you can choose AK26 or AH26, northwest and southwest in the diagram. (I mistakenly typed AM26 and AK26. AM26 is Vcc but you don't want to cross over pins)

I'm not comfortable with using 19x. You could try presetting it before the mod but what happens if you have to clear CMOS? Will your board default to 100 Mhz or 133 operation? It's not likely you can POST at 19 x 133 on default voltage. On the other hand you never know what will happen. Read on to see what I mean.

You can access all high multipliers but VIA boards have issues.

Usually all multiplier 13x thru 24x work great at 100 Mhz.

Some legacy boards have problems with multipliers 13x thru 14x at bus speed of 133 Mhz or higher. Read about the issue right there at the link I provided.

15x and up are fine limited by overclocking and default voltage.

I can't predict what you will see. Each board has its own issues.

Abit VIA boards tend to default to 20 x 100. Even my old KT7 does this. Tbred B's were great for that old board, the 1.6 volt or higher CPUs at least. Some 1.5 volt chips don't want' to POST at 20 x 100 on stock voltage.

Tbred A's almost never worked because they rarely overclock to 2000 Mhz and probably never on stock voltage.

Other boards may have a different default.

My Epox 8KTA3PRO (KT133A) always defaults (upon cleared CMOS) to the multiplier I choose and at 100 Mhz. Makes playing with wiremods very easy. Even when I got into trouble (too much overclock) another feature of my board bailed me out. There are +0.1V, +0.2V, +0.3V, +0.4V modifier jumpers on the board, like a factory voltage mod. Any time I have POST problems related to too much overclock I just move the jumper and add 0.4 volt to the chips default voltage. Other than the earlier Epox 8KTA3, I've never seen any other board with this nice overclocking feature. Too bad my board is so outdated.

OCINSIDE.DE is a good resource. I just want to point out that doing a wiremod for all 5 multiplier bits locks you into a single multiplier. They override CPU default and BIOS settings multipliers.

Generally, you want to use as few wires as possible so as not to elminate too many BIOS multipliers. Avoid mod'ing the 0.5x, 1x, and 2X pins if you can. You want to be able to change these in BIOS. 0.5X and 1X you definitely need. For example if you mod the 0.5X bit then you can only alter by whole multipliers because the 0.5x bit is now locked (or rather forced) to what you just set.

<b>[Addition]</b>
Here are some of my tips.

Take careful note of your CPU socket. Make sure you are orientated relative to the two corners with the missing pins. You don't want to mod the wrong pins!

If you are using OCINSIDE pinmod guide don't forget to use the "socket view". I'm assuming you are doing the wire in socket mod.

Mark the multiplier pins right on the CPU socket. I used a Sharpie pen. Once marked correctly you'll never mis-count pin holes again. (Been there done that but I didn't fry anything. LOL) You might also want to mark VID pins (voltage). For that matter, it might help you to do a wire mod to force the voltage to about 1.8 volt. Might help with POSTing during the initial testing phase to have a guaranteed Vcore. You can undo it later.

When inserting the CPU sometimes the little wire loops get in the way. You may have to do a little fiddling but remember the CPU should fall into place. Don't try to force it. (A bent pin will ruin your day).

When removing your CPU watch for the wire being pulled out. A loose piece of bare wire could also ruin your day.

To get the wire out of the socket can be difficult. I find it difficult grabbing the wire loops even with tweezers or needle-nose pliers. I use scotch tape. Pressing the tape down over the socket cause the loops to stick. If the tape doesn't pick up the loops it usually lifts them enough to grab them with the tweezers. Again be carefull not to drop the wires. Very hard to find on your motherboard.

<b>56K, slow and steady does not win the race on internet!</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 04/08/04 03:21 AM.</EM></FONT></P>

 

[phsstpok]

Make sure you have the correct pins. Orientate relative to the two corners with the missing pins.

I find the wire in socket mod the easiest to do. Tricky getting the wires out though (see my tips in my other message).

You might check in at ASUS Motherboards forum at <A HREF="http://www.amdmb.com" target="_new">http://www.amdmb.com</A>. See what issues other A7V333 owners have had.

<b>56K, slow and steady does not win the race on internet!</b>

 

[SidVicious]

BTW, that's not a 1700+ but a 1800+ that I'm trying to mod, being a Thoroughbred-B, it is factory unlocked by default but the motherboard seems to completely ignore the wire mod.

I could already manage multipliers from 5 to 12.5 in the BIOS and FSB from 100MHz to 160MHz in 1MHz increments.

The chipset will return errors with anything over 160MHz, which make sense since that early revision of the A7V333 was'nt approved for 166MHz FSB, I know those are'nt caused by my memory.

I can also choose a Vcc/Vcore from 1.1 to 2.3 with a combination of jumpers (undocumented and official) and BIOS settings, Vddr can be set between 2.5 and 2.8V with undocumented jumpers.

Right now, I am limited by the CPU multipliers, 12.5 x 160MHz is the highest possible frequency I can reach right now. Obviously, I do not know the full OC potential of my CPU but I suspect that I may be able to get a couple hundread more MHz if I access multipliers over 13.

I will try a few more tricks when I get home this evening.

Fok Speling Misstake

 

[phsstpok]

We are playing a bit of message tag.

I eventually realized you were using an XP1800+ not an XP1700+. Again sorry for the confusion.

Try the wire in socket mod. It's pretty easy if you are careful and the connections are pretty realiable due to ZIF socket spring tension.

If you get a good connection the processor has to switch multipliers because you are pulling the bit to the HIGH. Remember you are connecting the 8X sense pin directly to Vcore. This has nothing to do with the motherboard, or which motherboard you choose. The pin on the chip, itself, becomes strapped to the HIGH state.

The only problem that arises is whether the motherboard POSTs with the new multiplier. It will or it won't but something has to change when the mod is in place.

In my opinion, the easiest mod was the bridge mod. Take an XP1700+, XP1800+, or an unlocked Barton XP2500+. Cut bridges L3-3 and L3-5 and now you are in business. Boots in almost every motherboard. The XP1700+ or XP2500+ would now be mod'd to a 13x multiplier. The XP1800+ to 13.5x.

What makes this mod the best is you still have all the multiplier options from BIOS. Changing the bridges just changes the default multiplier of the CPU. Unlike the wiremods, this does not force the multiplier bits HIGH or LOW so the BIOS multipliers are not restricted.

<b>56K, slow and steady does not win the race on internet!</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 04/08/04 01:46 PM.</EM></FONT></P>

 

[SidVicious]

Message tag indeed !

The "wire in the socket" method does sound a lot more reliable and easyer to do than tying a knot around 2 CPU pins with an hair-sized, uncooperative copper wire, I swear that wire was making fun of me !

So I just need to insert a small(ish) wire running from #AJ27 (8x sense) to #AH26 (the nearest Vcc/Vcore) to turn the 8x bit high ?

I'll try it when I get home !

BTW, you mention cutting bridges, do I absolutely need to do that to gain access to 13+ multipliers ?



Fok Speling Misstake

 

[phsstpok]

BTW, you mention cutting bridges, do I absolutely need to do that to gain access to 13+ multipliers ?

Well, that's the heart of that workaround page. It mainly for people who ran into two issues.

One, after the single (8X bit) wire/pin mod the multipliers don't work with FSB at 133 Mhz (and higher) while they do at 100 Mhz. Affects a lot of people.

Two, even when one or more of the multipliers work with pin/wire mods usually 13x-14x do not work, due to legacy Multiplier ID handling. (Don't ask me for details. I don't quite get that concept myself. It has something to do with the old Athon 1.3 (200) and 1.4 (200) and getting 13X and 14x multipliers before there was an 8X multiplier bit in the chips).

Sometimes the bridge mods avoid these problems by working at the new default. Not always true but sometimes. You can always tell what will work. If a true XP2100+ and XP2200+ works in you particular board then an XP1700+ or XP1800+ will work too. While sometimes, the pinmod will fail.

In each case of failure, the workaround procedure is simply adding a second wire mod on one of the low-order bits. This some how fools the multiplier circuit, allowing the high multipliers to work with 133Mhz FSB and also sometimes restoring 13x-14X multipliers.

<b>56K, slow and steady does not win the race on internet!</b>

 

[phsstpok]

I just thought of one thing. You said you used a strand of wire from an old headset. I hope it wasn't a winding from the speaker element. That copper wire would have an insulation coating.

Regular cheap speaker wire (not windings) is a good choice. The strands are usually just fine enough not to cause any chip insertion problems when using the wire in socket method.

<b>56K, slow and steady does not win the race on internet!</b>

 

[SidVicious]

It finally worked ! I went for the "wire in the socket" trick and tried the 13.5 mod which only require two wires, one running between #AN25 (2x) and #AM26 (Vcc), the other connecting the all important #AJ27 (8x) to #AH26 (Vcc).

While I was cross checking an L3 bridge diagram with the pin array, I figured out one thing that you guys pointed out in your posts but that I totally missed :

Some L3 bridges have to be cut in order for the wire mod to work. In my particular case, I had to cut L3 #3 and #5 for that mod to work. The straight edge razor blade cuts I made are'nt as clean as the laser etched L3 #1 but it seems like everything is fine.

I am now running my CPU @ 2080MHz (160MHz x 13), Prime95 has been running a torture test for a few minutes now, so far so good. I'll be playing around with the multipliers and voltage until I find the sweet spot.

Thanks again for your help guys ! I feel a little embarassed that I could'nt see the relation between the L3 bridges and the pins connected to each of them.






Fok Speling Misstake