No worries! Thanks for catching it! Look forward to new version!

Re: Socket 5/7/SS7 (Voltage Interposer) Tweaker. (Released)

Lime made a new version(s) of the power adapter

For the curious regarding how this comes about since PCI slots are keyed, as mentioned before the key is just for signalling, not power. Here's the 2.1 spec:

And here's the 2.2 spec:

So if you have a 2.1 or prior motherboard with 5V keyed slots, it's not required to provide 3.3V power and any 2.1 or prior cards that need 3.3V power and are keyed for 5V are required to produce their own. With 2.2, motherboards are required to provide 3.3V and 5V regardless of signalling and cards can assume it's present. So using a dual voltage 2.2 card on a 2.1 or prior motherboard can fail.

Fortunately for me, my motherboard (Micronics W6-LI) has all the 3.3V lines on the PCI slots tied together and has some decoupling on them. It doesn't seem to go anywhere else on the motherboard though. Just manually wiring up a single pin to the nearest source of 3.3V is enough to get things going with at least a somewhat limited amount of current.

Edit: out of date read farther
new version

-I removed the vsb disconnect pads all else is the same

made this change based on observation of the freeway bodge board prototype. board seems to be looking pretty good. but as of now, I have not ordered a prototype for this or the latest version in this post. so if someone is interested, order one and check it out, I'm kinda swamped with other stuff, lemme know if there are any issues. but I am kinda expecting that this is probably good to go. (be careful, test for shorts before powering on).

I will probably just make bodge boards based on this for my motherboards instead of these universal type adapters. so have at this one, lemme know how it goes.

Sphere478 wrote on 2022-07-10, 23:03:

Re: AGP to PCI adapters - any PCB designers here willing to make one?

Need to investigate pin for 66mhz

this adapter isn't doing anything to this pin therefore no need to look further into this.

I like it, but they tying of 3.3v aux to 3.3v power makes me a little nervous. Only a little though. For systems following PCI <= 2.1, 3.3v aux is marked as reserved, so it should not be present and using this shouldn't cause any problems. for PCI > 2.1, the 3.3V aux signal is present, but the 3.3v power rail is required so this card isn't needed. If there's any pathological designers out there that included 3.3V aux but not 3.3V power, shorting the two together would be bad and this would be difficult to test for as 3.3v aux does not come from the PSU, but from a regulator that steps down 5VSB from the PSU.

RussD wrote on 2022-08-22, 20:32:

I like it, but they tying of 3.3v aux to 3.3v power makes me a little nervous. Only a little though. For systems following PCI <= 2.1, 3.3v aux is marked as reserved, so it should not be present and using this shouldn't cause any problems. for PCI > 2.1, the 3.3V aux signal is present, but the 3.3v power rail is required so this card isn't needed. If there's any pathological designers out there that included 3.3V aux but not 3.3V power, shorting the two together would be bad and this would be difficult to test for as 3.3v aux does not come from the PSU, but from a regulator that steps down 5VSB from the PSU.

I removed it because the method I tried didn’t work in physical form.

I can add it back in another way.

We are talking about 3.3v vsb right?

Sphere478 wrote on 2022-08-23, 00:22:

I removed it because the method I tried didn’t work in physical form. […]
Show full quote

RussD wrote on 2022-08-22, 20:32:

I like it, but they tying of 3.3v aux to 3.3v power makes me a little nervous. Only a little though. For systems following PCI <= 2.1, 3.3v aux is marked as reserved, so it should not be present and using this shouldn't cause any problems. for PCI > 2.1, the 3.3V aux signal is present, but the 3.3v power rail is required so this card isn't needed. If there's any pathological designers out there that included 3.3V aux but not 3.3V power, shorting the two together would be bad and this would be difficult to test for as 3.3v aux does not come from the PSU, but from a regulator that steps down 5VSB from the PSU.

I removed it because the method I tried didn’t work in physical form.

I can add it back in another way.

We are talking about 3.3v vsb right?

Yes, this one here:

Just one of those many possible corner cases I suppose.

I believe the adapter posted earlier in the thread by scheiss_freak has the 3.3vsb hard wired to 3.3v wasn’t it?

In any case, I could try a larger gap or run a trace on a third layer. Running it on a 3.3v layer would cut too much copper.

It's probably fine to just leave it. I think the possibility of a motherboard with 3.3Vsb somehow connected but not 3.3v power is unlikely enough to not worry about.

What I would change though is the connection of IO power. If you are dealing with a system with 3.3V IO power, it already has 3.3V power connected. You don't need this board. If it has 5V IO power, you don't really need to bring it out. That would give you an extra layer to work with. Either to bring out 3.3vsb optionally, or add a ground layer in order to allow decoupling caps to be added at either end.

Could make a version of this with a 5v to 3.3v regulator on it. There was a regulator speced for powering 486 chips that would probably work. I think I may have posted it in This or the socket 1/2/3 tweaker thread. But running a line to the atx connector is probably a better idea for atx mobos.

There is also vcc3 on most mobos. The motherboard tweaker can tap that. But some have spoken against this idea, though I think it would probably work. Especially on a split rail cpu.

Check out the gigabyte iRam drive adapter project in my sig. it has the rest of the voltage layers. We could modify it or pull the ground portion from it.

The 3.3/5v layer is separated on the one in this thread to keep it separate if the mobo gives those pins 5v there are bridges at both ends to combine them if you wish. I haven’t prototyped this feature yet though.

Lemme know thoughts, we might make changes 😀

Given that it sits underneath the motherboard, I think the single purpose design works really well. I wouldn't add a regulator. As far as the 3.3V vs 5V, I'm just not understanding in what situations you'd want to tie IO power to anything.

I understand why some would caution against hooking into an existing 3.3V supply, as it may not be rated for what's needed on the expansion cards that might get used. I guess it depends on how willing someone is to risk it and potentially need to replace a regulator. Using an add in-regulator is always a safer bet. But often people will just need this for a single expansion card and things will probably "be fine" either way.

In the pci specs there are pins that are designated as 3.3v always (or open)

But there are also some that got designated as 3.3 OR 5v

In such a case it is left to the user wether or not they tie those to 3.3v which is why I left them separated. (Don’t tie if mobo has 5v there)

As for tying to 3.3vio (at the regulator or on motherboard tweaker.)

Your call, I haven’t tried it yet. But on a socket 7 dual plane cpu that rail is just sitting there not doing much.

I'm still a little confused, for the IO power pins, yes, the spec specifies they can be 3.3V or 5V, but you would never want to tie them to a 3.3V or 5V supply. They are already connected to one or the other. The possible cases for motherboards supporting PCI v2.1 and below are:

1a) IO power is 3.3V and both 3.3V and 5V rails are present (3.3V keyed slot)
1b) IO power is 5V and both 3.3V and 5V rails are present (5V keyed slot)
1c) IO power is 5V and *only* 5V rail is present (5V keyed slot)

For motherboards supporting PCI v2.2 and above, the possible cases are:

2a) IO power is 3.3V and both 3.3V and 5v rails are present (3.3v keyed slot)
2b) IO power is 5V and both 3.3V and 5V rails are present (5V keyed slot)

A failure occurs when you have a PCI v2.2 and above card that supports both 3.3V and 5V IO power, so it has both ends keyed and it requires 3.3V to be present. This fails in a PCI v2.1 and below motherboard in the (1c) case. Both the motherboard and card are following a version of the PCI spec, it's just that in this situation, they are incompatible. To fix this, 3.3V needs to be injected on the 3.3V power rail of the PCI slot and that isn't something available from the IO power rail, as it will be 5V.

I’m taking about vcc3 on the cpu. (V I/O)

Re: Socket 5/7/SS7 (Motherboard) Tweaker

I think we're talking past each other a bit here. I'm not referring to the sourcing of 3.3V. I'm talking about the pins brought out by the adapter board from the PCI slot. Specifically the IOPWR pins, A10, A16 and B19.

RussD wrote on 2022-08-29, 20:32:

I think we're talking past each other a bit here. I'm not referring to the sourcing of 3.3V. I'm talking about the pins brought out by the adapter board from the PCI slot. Specifically the IOPWR pins, A10, A16 and B19.

If a mistake has been made I’d like to correct it. I am listening.

http://www.interfacebus.com/Design_PCI_Pinout.html

B59 a59 these are also defined as such

Are you saying that they aren’t common. All the ones that say 3.3 or 5v

I have those separated from the ones that say 3.3v

Sphere478 wrote on 2022-08-29, 22:01:

If a mistake has been made I’d like to correct it. I am listening. […]
Show full quote

RussD wrote on 2022-08-29, 20:32:

I think we're talking past each other a bit here. I'm not referring to the sourcing of 3.3V. I'm talking about the pins brought out by the adapter board from the PCI slot. Specifically the IOPWR pins, A10, A16 and B19.

If a mistake has been made I’d like to correct it. I am listening.

http://www.interfacebus.com/Design_PCI_Pinout.html

B59 a59 these are also defined as such

Are you saying that they aren’t common. All the ones that say 3.3 or 5v

I have those separated from the ones that say 3.3v

Yes, sorry, A59/B59 as well. I mention IOPWR because I don't think there's a purpose to including them in the PCB at all. The PCB only needs to bring out the +3.3V lines and optionally 3.3V aux. Beyond bringing out 3.3V, ground might be useful because it would allow decoupling capacitors at either end of the PCB.

RussD wrote on 2022-08-30, 05:11:

Sphere478 wrote on 2022-08-29, 22:01:

If a mistake has been made I’d like to correct it. I am listening. […]
Show full quote

RussD wrote on 2022-08-29, 20:32:

I think we're talking past each other a bit here. I'm not referring to the sourcing of 3.3V. I'm talking about the pins brought out by the adapter board from the PCI slot. Specifically the IOPWR pins, A10, A16 and B19.

If a mistake has been made I’d like to correct it. I am listening.

http://www.interfacebus.com/Design_PCI_Pinout.html

B59 a59 these are also defined as such

Are you saying that they aren’t common. All the ones that say 3.3 or 5v

I have those separated from the ones that say 3.3v

Yes, sorry, A59/B59 as well. I mention IOPWR because I don't think there's a purpose to including them in the PCB at all. The PCB only needs to bring out the +3.3V lines and optionally 3.3V aux. Beyond bringing out 3.3V, ground might be useful because it would allow decoupling capacitors at either end of the PCB.

So it is already set up that those pins can be left divorced.

I guess maybe you are saying that there isn’t much point at all including it as it is one or the other and that is up to the mobo and no gain tying to it?

I was figuring if it read 3v that you could tie it. But such a mobo would already have 3.3v so maybe still no point? Am I following?

In that case all I have to do is program the flood for 5v/3.3v to ground instead and there are your decoupling pads. (And now gnd taps also)

So switch the 5v/3.3v layer to ground and re route to gnd pads. Also add a 3v vsb switch?

Right, if IOPWR measures 3.3V, the motherboard by all versions of the PCI spec must provide 3.3V power, so the card isn't needed.

I still go back and forth on making 3.3Vsb optional. It's not that is a particularly difficult decision, I just can't come up with any situations where it would actually matter. The only situation where it would matter to the end user is if they wanted to use a PCI card with wake support, such as wake on lan ethernet. In such a case you'd need it brought out to a separate regulator that is powered by 5VSB. Just connecting it to 3.3V wouldn't provide any extra functionality. It's an optional signal for all versions of the spec, so cards should function just fine without it (minus being able to wake the system from standby). And if there are cards out there that for some reason do break during normal operation is 3.3VSB is left open, tying it to 3.3V shouldn't break things unless there's some pathological motherboard designer that's wired up 3.3VSB but not 3.3V.

Okie dokie. I’ll play around with it, wanna prototype it when I’m done?