Okay, with latest developments in the tillamook thread I am more fully understanding this,

Re: Tillamook 266MHz and working L2 cache?

It seems,

You want to be sure that BF0, BF1, BF2 all show about 330 ohms or so to VCC3 (3.3v)

I believe that will solve some of the issues that some mobos have with setting higher multipliers namely socket 5 boards (Can anyone confirm?)

Next, you want about 25 ohms on your three switches to ground (vss). And obviously you just add this switch to your BF2 pin on older mobos.

okay, I think I got it

how's it look? that look right?

of course any vss or vcc pin can be substituted for one another (respectively)

Edit: I’m a dumbass and connected them all together.. 🤣

I updated the diagram top show the WB/WT mod also. (from K6-2/3+ Success on Asus P5A 1.06 - simple mod)

This gets me to thinking, maybe we could make a small pcb that we can solder on the back of the board with little dip switches and smd resistors on it

Kinda like this Re: Adding 3.3v to pci

Btw, we may be able to change the resistor values on the bf mod to 1k and 10k or 100 and 1000 I suspect that the low values I saw on the kingston aren’t as high as will work. And higher is better, (safer) one might assume that values that work for wt/bw should also work for bf

Are there any other mods that could be added to this theoretical pcb?

https://www.pchardwarelinks.com/586pin.htm
Description of pins

Seems there is some image quality loss on the modified image, so here is a
Higher res pic:

KEN#, CACHE# Cache pins may be for enabling or disabling cache. It could be added.

SMI#, SMIACT# could be used to set SMI (No Idea of this would help with anything)

BHOLD, idk if this can be activated by a pin but some mobos have a special bus option for cyrix chips.

There are several things listed for dual processor functions on intel,

One in particular interest. Is how might we be able to set redundant processor mode. Perhaps by enabling of a signal on one of these pins?

Catsay shared this^

So pull down should be 500 or less
And pull up should be 2.2k or less

ilembitov on discord just used the guide to do this to a pentium mmx 😀 and says it’s working.

It appears they are pulling bf1 to ground which should be relatively safe, but a higher ohm resistor may have been another approach that may have safety benefits. (The kingston used 25-ohm, the intel doc recommended 500 ohm or less for pull down, where as the kingston pulled up with a 330 ohm and the intel doc said less than 2200ohm )

I also re did my tillamooks

The tillamooks put bf2 one pin in toward the core so a link needs to be run between it and the normal position.

Many motherboards, specifically socket 7 (non super) don’t pull the bf2 pin high. So the 0 ohm link isn’t enough to make it work. It also needs to be pulled high. But a 0 ohm link to bf2 would not be a good idea, if you were to pull bf2 low you would the. Have a short.

So you must use a resistor I used a 330 ohm like what I observed on the kingston interposer.

I tried it in my socket 7 mobo and 266 4x setting is working when the motherboard is set to 2x without any need to add a option to pull bf2 to ground.

Some tillamooks had a 4.5x multiplier, I am wondering if pulling it low will show that setting. (Something I will experiment on later)

I’d like to get some help in this thread with a pcb project that can simply be soldered to the back side of the mobo and would have smd resistors and dip switches. Or jumpers.

I’m curious, about something…

The 233mmx supported 2x-3.5x

But

Later on the tillamook intel thought to put the bf2 pin here: (shown is the 233mmx p55c not tillamook diagram)

The tillamook supports multipliers up to 4.5x
It’s possible that intel had planned on this prior to the tillamook, but never implimented it. Especially given that your typical pentium mmx can often handle 266 mhz some even 300mhz

….. any chance the 233 mmx p55c ended up with something there on that pin that we don’t know about? (Because ss7 mobos dont use that pin, and it’s marked as n/c, it may have gone un noticed) I know it says NC. But we could confirm..?

Tillamook SMP and multiplier unlocking investigation thread.

In this thread I discovered a few pins marked NC were actually connected to something.

It’s a long shot but…. Easy to try.

You'll need separate pull up and pull down resistors per pin. The diagram you show is effectively shorting all three BFx pins, making all have the same level

H3nrik V! wrote on 2022-03-20, 09:50:

You'll need separate pull up and pull down resistors per pin. The diagram you show is effectively shorting all three BFx pins, making all have the same level

Oh okay, so the pull up resistors should be on each line? Good catch. Thanks. I’m ashamed to say I missed that 🤣 it’s so obvious.

————————————————————————-

A special thanks to computerguy096 for testing the tillamook bf2 pin (but on a p55c) to see if there was a hidden bf2 pin on the p55c

He saw nothing but a open connection 🙁

Next let’s see if we can find the fabled 4.5x setting on the tillamook by giving bf2 a pull down

how's this?

ugh, I wish I could upload bitmaps

As for the pcb, this also could be implemented (tillimook mods)

I’m not sure if any of these links may interfere with different model processors though… obviously they don’t seem to be of any harm in different mobos when a tillamook is used.
There are some reports of people needing to disable some of these mods on tillamooks to make them work right from one mobo to the next. So making them switched may be a good idea.

I also got to thinking that castellated holes could be used to recess a dip switch block as low down as possible for more clearance so as to allow motherboard to be more easily mounted

another idea, remote switches so that they can be more easily changed in a case

Here's a KiCAD v5 PCB-only file you can mess with. I also generated the gerbers in there too. The board outline will probably have to be brought in closer to the pins to clear the ZIF lever.

If you were to build this, I'd recommend milli-max individual pins soldered through the board so it can go into the ZIF socket on the motherboard easily, then the CPU would be on the top side PGA style.

I just saw this thread, but it isn't clear to me as to what this mod actually does. Would someone please summarize it for me or link to another relevant thread?

wiretap wrote on 2022-03-20, 19:15:

Here's a KiCAD v5 PCB-only file you can mess with. I also generated the gerbers in there too. The board outline will probably have to be brought in closer to the pins to clear the ZIF lever.

If you were to build this, I'd recommend milli-max individual pins soldered through the board so it can go into the ZIF socket on the motherboard easily, then the CPU would be on the top side PGA style.

socket7-mod.zip

I was figuring this would be soldered to the BACK of the board 😀

Like the pci slot thing I linked

Which brings up another thing,

We should put a castellated edge with a 3.3v tap for those 3.3v slot things

But lower profile outline is a good idea.

Many, all of the holes will have to wider so that they can go around the solder domes and then solder can be poured into the ones we wish to connect to, many of the others can simply be left not plated as to make trace routing easier.

Kahenraz wrote on 2022-03-20, 19:21:

I just saw this thread, but it's it clear to me as to what this mod actually does. Would someone please summarize it for me or link to another relevant thread?

Did you mean that it “isn’t” clear?

Basically there are several mods you can do to socket 5/7/ss7

I’ve linked to several of these threads

One makes the p5a mobos work better by toggling wt/wb (which can be handy on other mobos as well.)

There are also mods that make tillamooks work on desktop boards, issues there are voltage detect, l2 cache, and multiplier pin issues

We can also tap the cpu for 3.3v for the pci slots, especially if the cpu is dual plane. The 3.3v is over powered for a dual plane cpu so there is amperage to spare and can be used for the pci slot

And finally there are the multipliers.
There are four pins that I am aware of related to this, (three normal ones, and one for tillamook)
On many motherboards eapecially socket 5 bf0/bf1 aren’t properly pulled high, it’s a easy fix, and on socket 7 bf2 isn’t pulled high. Again a easy fix.

My goal is here to combine all these into one easy to install device that simply solder a few points on the back of the motherboard and presto, you now have not only the mods but also ways to shut them off and change their settings via switches.

Some notes:

The voltage detect mod should probably have a resistor pad and a switch

Cache mod #1 should have a pad and a switch

Cache mod #2 should have a pad and a switch

I believe the tillamook bf mod might be able to just be a link, but to be safe we should give a pad and a switch.

The bf2 tillamook resistor mod will be addressed at the same time as the bf0,1,2 mods

The last schematic should be good for the bf mods

We will have to do some digging into how to implement the wt/wb mod in a way that will work with all mobos.

My consern is the mobo probably already has pull high and pull low resistors for it and we will have to over power them.. (while the system may be pulling the opposite way)so perhaps pads for both, and values can be determined later? It seems that a double throw switch is required here to over ride the existing pull up/pull down signals? Am I right?

Also another concern is the affect of tillamook mods on other processors

Sphere478 wrote on 2022-03-20, 10:41:

how's this?
803CC0C4-2F2A-42EF-9740-36A7AF19524C.gif

ugh, I wish I could upload bitmaps

You need 3 separate 25 Ohms as well 😀

[Edit; had posted my reply inside quotation]

[Edit 2 - Never mind, you don't need, I guess] 🤣

[Edit 3] You might not need 3 separate pull-downs per se, but you'll might experience unexpected behavior from having one shared, as current through it will change depending on how many switches are closed, i.e. better safe than sorry.

H3nrik V! wrote on 2022-03-21, 06:20:

You need 3 separate 25 Ohms as well :) […]
Show full quote

Sphere478 wrote on 2022-03-20, 10:41:

how's this?
803CC0C4-2F2A-42EF-9740-36A7AF19524C.gif

ugh, I wish I could upload bitmaps

You need 3 separate 25 Ohms as well 😀

[Edit; had posted my reply inside quotation]

[Edit 2 - Never mind, you don't need, I guess] 🤣

[Edit 3] You might not need 3 separate pull-downs per se, but you'll might experience unexpected behavior from having one shared, as current through it will change depending on how many switches are closed, i.e. better safe than sorry.

Yeah, I was figuring that one would do, I think the kingston is wired like that. But may as well add 2 more what can it hurt…

Can anyone think of any mods that we have missed?

Edit: an image of the prototype from wiretap

-

new schematic

. check me for errors. I couldn't decide if the wt/wb should be a tripple throw so it had a neutral position or if it should be two switches seems like two switches is best because it is easy. since it has resistors that should make it safe incase both are turned on. so I think this should work.

I haven't put in resistor values yet.

I made the bf more complicated but I like that this way everything on the board can be totally disabled just by flipping a switch and go totally passive
everything has resistor pads, whether they need it or not. I'd hate to find out later they were needed and not have them (if you don't feel like installing them bridge it with solder)

this all means 12 switches. gonna get a little crowded, but I think it's doable.

does anyone else have a suggestion for mods needed?

I assume there isn't a problem combining all the vcc3 into one rail and tapping it for pci 3.3v power (when using a dual plane cpu)

For use with these:

I did the resistor mod on my ASUS P5A recently, but a dip switch on the CPU would be a lot nicer. I would rather not have this bodge on my board if it can be moved to the CPU. That way it can be easily enabled or disabled for testing.

Kahenraz wrote on 2022-03-21, 13:46:

I did the resistor mod on my ASUS P5A recently, but a dip switch on the CPU would be a lot nicer. I would rather not have this bodge on my board if it can be moved to the CPU. That way it can be easily enabled or disabled for testing.

The idea is to have this on the back of the motherboard

Alternately, eventually, this may be able to be incorporated into catsay’s power leap interposer that adds cache.

It’s not impossible though that a version of this couldn’t be added to the cpu it’s self but the pcb would have to be very thin

There is just more room for switches on the back of the mobo, also breaking out the 3.3v for the pci is a secondary goal

“ Cat, what resistor values should be used on my schematic?”

Catsay wrote:

1 330Ohm and 1Kohm should be fine
2For other pullups/downs than BFx you can also use like 10K
3The important thing is though that fast pins that get latched after reset (used as strap bit value initially and then for fast data transfers) use 'fast' resistors, this meaning low value resistors