I like it! 😀

Necrodude wrote on 2022-12-27, 09:37:

I like it! 😀

It would save me a lot of time if someone had a netted agp 1x/2x footprint laying around. Anyone? For kicad.

Can someone do some tests on some mobos that seem to need this? Put a overpowered card in it and hook it up to a scope and see if there are dropouts. I mean just as a confirmation that this will even help at all.

as for the pcb project files:

Might have to make the footprint myself unless someone has one.

found these:

I have an AGP footprint in Eagle, would that help?

Woohoo!

yes! Would you be willing to upload it in it's own project file with nets as public domain?

Sure thing 😀

Hope it helps!

Thanks SDZ you rock! as usual! 😃

Do I connect vio and vcc3.3 together? or make a option jumper? or a separate tap? or leave vio unconnected?

Any need to tap vcc5 or vcc12? only a few pins for those. doesn't seem like they are used heavily.

I need to check and correct orientation at this stage. any chance you could highlight 12v pin in 3d viewer on your project fr0m back side of board? or I can probe a mobo later. gotta be 100% sure. Silk next.

3.3V and VIO can be connected together for AGP 1.0 only, after that VIO is lower, 1.5V or 0.8V for AGP 2.0 and 3.0.
No clue how much the 5V and 12V rails are used on old AGP cards.

I can highlight, but a bit later today.

Okay, I just need to confirm orientation. 50/50 it is upside down right now 🤣. Figured easiest foolproof way is to bust out a ohm meter and find 12v on a mobo or highlight 12v in 3d on your device.

Not in front of a agp mobo atm. But maybe later.

Sorry if this is a dumb question, but is this intended to help with mobos that don't play nice with certain video cards?

I've got a GA-5AX rev 4.1 and, as I understand it, using certain Nvidia cards with it might be a bad idea.

If this fixes the problem, sign me up. 😀

Just a suggestion.

Since the pins for AGP are so much closer together than the pins for a PCI port, it might be easier for installation to use castellations on the PCB and just use the power connections which are on the outside of the pins.
Seems like half the pins for power are on the outside anyway

But regardless, whats the point of this? would it actually fix any issues similar to the PCI version of this board?

CalamityLime wrote on 2023-01-06, 09:52:

Just a suggestion. […]
Show full quote

Just a suggestion.

Since the pins for AGP are so much closer together than the pins for a PCI port, it might be easier for installation to use castellations on the PCB and just use the power connections which are on the outside of the pins.
Seems like half the pins for power are on the outside anyway

But regardless, whats the point of this? would it actually fix any issues similar to the PCI version of this board?

If traces are too thin, I can probably add a ring or go to a ring. Not a bad idea. Let me see what happens.

It has come up many times that people claim motherboards are unable to handle the power requirement of specific cards.

To my knowledge, this has never actually been tested or confirmed, but I guess it makes sense. If the traces aren’t beefy enough voltage sag might kick out the card.🤷‍♂️

If there are registers we can trigger with this adapter to set specific settings we can also do that.

So basically, it’s as much of an investigation as anything.

Looking for someone who has a card that isn’t working and they suspect power may be the culprit, maybe it is, maybe it isn’t. But beefing up the power supply shouldn’t hurt.

Suggest such a person starts by hooking up a scope so we can see what’s going on. Then we can move on to the adapter. The adapter might be a cart before the horse kind of deal.

CalamityLime wrote on 2023-01-06, 09:52:

But regardless, whats the point of this? would it actually fix any issues similar to the PCI version of this board?

Some early AGP motherboards couldn't deliver all the juice power-hungry graphics cards wanted to the AGP slot, or they were marginal overstressing the related components/tracks and eventually failing.

I never heard of this being really an issue at the time, I guess motherboards could handle existing cards. But when crazy guys in retro forums started experimenting with the last AGP card on the first AGP motherboard it became an issue.

I don't know how common it is though or whether this proposed solution will address it.

I noticed that I hadn't uploaded my progress here.

I haven't done much work on this in a while and probably won't do anymore before someone proves that a device like this will actually help. probably not me as I currently don't have a build in need of one. Anyway, here is a Beta draft of what I was last working on. I'm not aware of any issues with it, though there may exist some. ordering or building it isn't advised without further topic development.

there are long adapters from PCI to AGP , I seen them , and there are specs for all agp (data links) and specs for PCI (data links) , any of this would help ? should I start searching?

Everything already Exists... - should I search more?

This is even cooler , PCI32 to PCIexpress (if chipset supports in theory).

you can search some more if you wish

how cool is this :

Oh wow that is pretty cool.

This device is mainly to address the idea that some motherboards didn’t have enough power for some gfx cards.

The devices you listed are adapters from other busses not agp power augmenters

Sphere478 wrote on 2023-06-04, 20:09:

Oh wow that is pretty cool.

This device is mainly to address the idea that some motherboards didn’t have enough power for some gfx cards.

The devices you listed are adapters from other busses not agp power augmenters

OK now I get it , tnx for clarifying . (that is soldered under the board under the AGP soldering points).

Correct