3dfx Archive - AGP.5 Release.

osckhar God Member Offline www.3Dfx.es Posts: 812 SPAIN Gender:	Re: AGP.5 Release. Reply #60 - 04.12.13 at 14:50:40 goriath wrote on 04.12.13 at 14:37:48: Does this imply that the power supply circuitry on the AGP.5 adapter is redundant? No, since without it the VGA will not boot up. Internal power supply send 3.3V signal to 10 agp pins A/B needed. - Oscar.
Back to top	3Dfx Site - Museo 3Dfx - 3Dfx Collection.&& IP Logged

goriath

God Member

Offline

Ruggedized Retro Gamer

Posts: 599
Italy/Slovenija
Gender: male

Re: AGP.5 Release.
Reply #61 - 04.12.13 at 15:35:05

What I meant to say was:

As far as I can understand, the power supply circuitry is there to provide the needed 3.3V from a source of 5V (PCI slot and/or molex), right? So, if this circuitry is involved when feeds the card plugged into a 5V PCI slot, if you don't implement level shifters because you are planning to use it into 3.3V PCI (because of 66MHz), why should you keep the power supply circuitry since the card could be fed through the 3.3V PCI slot itself?

Since plugging the adapter (with 3.3V card) into a PCI 5V ONLY slot doesn't seem safe without implementing level shifters, then you can remove the power circuitry from the adapter and use it only into 3.3V PCI slots. In this way you feed the card through the 3.3V PCI slot itself. In the other thread Komponent built a very simple adapter that shows what I'm referring to.

Back to top

« Last Edit: 04.12.13 at 15:36:51 by goriath »

OLD SCHOOL: Because it was done right the first time

IP Logged

osckhar

God Member

Offline

www.3Dfx.es

Posts: 812
SPAIN
Gender: male

Re: AGP.5 Release.
Reply #62 - 04.12.13 at 16:09:33

I hope I can explain me since technician words on english are not my hard.

Here how I understand AGP.5 works:

Internal Power Supply is all the time working giving 3.3V VGA device. Does not matter if you are using a PCI_5V or PCIx_3.3V.
AGP.5 adapter does not have traces from PCI_PIN to AGP_PINS in power rail. Thanks internal Power Supply adapter can be used in both PCI/PCIx slot with right voltage. Yes, now we must to know if really the SHIFTERS are needed when adapter is pluged in PCI 5V.

Quote:

why should you keep the power supply circuitry since the card could be fed through the 3.3V PCI slot itself?

If for example you only want to run the AGP.5 adapter on PCIx_3.3V. Yes, you can omit all the internal Power Supply but traces must be created from PCI_PIN to AGP_PIN adapter. So VGA will take the voltage directly from the PCI slot and you can cost saving in material

For example the Adapter version from Dominik. He had a great idea since his adapter does not need internal power supply for 3.3V Rail since a sata connection will do the work.

Quote:

BUT! Most (if not all) more modern mainboards, that are PCI2.1/2.2 compliant will work with the selfmade adapter and don't harm the card. Why? Simply because the southbridges are designed to work with 3.3v cards even when they are keyed for 5V.

If Dominik is right and thanks southbrige adapter does not need the SHIFTERS for transfering data signals in correct voltage. My question is how shouthbridge detects if it needs to send signal data in 3.3V when adapter is pluged into 5V PCI slot?

Thanks,
Oscar.

Back to top

« Last Edit: 04.12.13 at 16:16:55 by osckhar »

3Dfx Site - Museo 3Dfx - 3Dfx Collection.&&

IP Logged

Komponent

Senior Member

Offline

3dfx modder

Posts: 282

Re: AGP.5 Release.
Reply #63 - 04.12.13 at 16:25:14

Yes, you can do without dedicated power supply for something like a Voodoo5 5500, if run inside a server/workstation designed motherboard, into a PCI-X slot, but I can't know if it will be enough to run a Voodoo5 6000, or with a low quality motherboard into a "regular" PCI slot.

If someone knows for granted that the motherboard he intends to use with AGP5 does 3.3V level logic, sure it is safe, no matter how the slot is keyed. That was never the point. The case is about 5V PCI. If you're not sure (or don't know) what you have or you know to have a 5V part, it's not safe to test your luck with AGP5 or bad things could happen Shocked

With level shifting you don't care about those issues Wink

Back to top

IP Logged

goriath

God Member

Offline

Ruggedized Retro Gamer

Posts: 599
Italy/Slovenija
Gender: male

Re: AGP.5 Release.
Reply #64 - 04.12.13 at 16:38:47

OK, I understand now.
I was thinking that adapter was designed to change 5V to 3.3V only when 5V source was used, while keep feeding the card through the PCI slot if 3.3V source.

In the picture I saw the adapter working into (I guess...) 3.3V PCI slot and with the molex unplugged. When the molex comes to?

osckhar wrote on 04.12.13 at 16:09:33:

If Dominik is right and thanks southbrige adapter does not need the SHIFTERS for transfering data signals in correct voltage. My question is how shouthbridge detects if it needs to send signal data in 3.3V when adapter is pluged into 5V PCI slot?

That's what I was wondering too.

Back to top

« Last Edit: 04.12.13 at 16:41:47 by goriath »

OLD SCHOOL: Because it was done right the first time

IP Logged

Komponent

Senior Member

Offline

3dfx modder

Posts: 282

Re: AGP.5 Release.
Reply #65 - 04.12.13 at 18:38:37

It doesn't switch. If it signals at 3.3v level that is compatible with 5v logic (the thresholds for logic high and low are met). From 3.3v initiator to 5v target it's compatible. The problem arises in reverse, from 5v to 3.3v, if the 3.3v target is not 5v tolerant. Then level shifters are required.
In simple words, if the motherboard signals at 3.3v all the time AND its inputs are 5v tolerant, then it works with both 5v and 3.3v PCI cards, no matter the slot keying. A number of newer motherboards (BUT NOT ALL) do work that way. Those are safe to use with AGP5 and equally with my passive AGP2PCIX adapter.

Back to top

IP Logged

Loeschzwerg

Senior Member

Offline

AAlchemy PSU maker :-)

Posts: 396
Germany -> Bavaria -> Augsburg
Gender: male

Re: AGP.5 Release.
Reply #66 - 04.12.13 at 18:47:27

The PCI 2.1 specification recommends using dual voltage buffers that can work with 5v/3.3v, that's probably why 5v keyed slots can also work with 3.3v signaling.

Edit: From Spec 2.1

Edit2: KT133, KT266 are compatible. 440BX is also compatible. All famous plattforms for 3dfx gamers will do fine. Every 2.2/2.3 compliant board will anyway. My advice: Always check the data sheets of the board.

We could do a compatibility list.

Back to top

« Last Edit: 04.12.13 at 19:15:49 by Loeschzwerg »

IP Logged

goriath

God Member

Offline

Ruggedized Retro Gamer

Posts: 599
Italy/Slovenija
Gender: male

Re: AGP.5 Release.
Reply #67 - 04.12.13 at 19:54:18

Komponent wrote on 04.12.13 at 18:38:37:

It doesn't switch. If it signals at 3.3v level that is compatible with 5v logic (the thresholds for logic high and low are met). From 3.3v initiator to 5v target it's compatible. The problem arises in reverse, from 5v to 3.3v, if the 3.3v target is not 5v tolerant. Then level shifters are required.
In simple words, if the motherboard signals at 3.3v all the time AND its inputs are 5v tolerant, then it works with both 5v and 3.3v PCI cards, no matter the slot keying. A number of newer motherboards (BUT NOT ALL) do work that way. Those are safe to use with AGP5 and equally with my passive AGP2PCIX adapter.

So, let's see if I got it:

We have basically an (could I call?) *interruptor* that does work this way:

0V - 0 binary - LOW - so it's OFF
3.3V/5V - 1 binary - HIGH - so it's ON

Now, let's say we have 3.3V source and we need to feed a 5V target - in this case there aren't problems since the 3.3V (MAX) that the source can provide are actually within the range of tolerances (0V<3.3V<5V); so, since it's non-zero signal it acts as HIGH anyway.

On the other hand, we could have a 5V source to feed a 3.3V target - in this case singal is still a non-zero so it acts as HIGH BUT we could run into troubles since the 5V that the source provides is actually out of the spec's range and could not be tolerated by the target. Here is where the level shifters come to.

Is it correct?

Back to top

OLD SCHOOL: Because it was done right the first time

IP Logged

osckhar

God Member

Offline

www.3Dfx.es

Posts: 812
SPAIN
Gender: male

Re: AGP.5 Release.
Reply #68 - 05.12.13 at 09:33:24

If I understand it right. Since PCI 2.1 compliance the southbridge will send signal data in 3.3V even for V5 PCI slot. Am I correct? Or this feautre is added in PCI2.2 compliance?

Any motherboard manufactured from 1999 year must be compatible since PCI2.2 was introduced in 1998 year. Am I correct?

Yes, make a compatible motherboard list is a good idea.

Thanks,
Oscar.

Back to top

« Last Edit: 05.12.13 at 13:42:14 by osckhar »

3Dfx Site - Museo 3Dfx - 3Dfx Collection.&&

IP Logged

osckhar

God Member

Offline

www.3Dfx.es

Posts: 812
SPAIN
Gender: male

Re: AGP.5 Release.
Reply #69 - 27.01.14 at 13:09:54

AGP3.3 DUAL VERSION

- Limited Edition black PCB / Flash Golden letters.
- Fixed some minor bugs.
- Modified for using the adapter without internal Power Supply (only PCIx 3.3v)
- Added an extra IC regulator on back of the adapter to help IC1 to controll better the heat.
- Added new holes for improvement the heatsink than vers. AGP3.0
- Via jumpers (Open / Short) you can set the adapter according as needed for each user.
- Different range price.

I modified the PCB for different settings. What it means different range price.

LOW COST ADAPTER:

Not internal POWER SUPPLY. It only compatible in PCIx slots. It could be the best solution for all V5-6000 users since If PCIx can handle the power rail needed when 6k is runing 8xFSAA. It means, cheap adatper and not heat problem. Only a motherboard with PCIx slot is needed.

MEDIUM COST ADAPTER (users dont own V5-6000 card):

Internal Power Supply. It means it can be used in standar PCI slot with 5V. Be ware here. Make sure your PCI slot is at least 2.1 compliance. I still did not find a motherboard is not compatible with adapter.

HIGH END DEVICE DUAL SETTINGS VIA JUMPERS + FRONT / BACK HEATSINK ON IC + EXTRA IC REGULATOR (Only recommendable for V5-6000 users):

It would be a deluxe version dual. Settings via jumpers where you can choose what manner you want to use it. Problems? Yes, if you not use correctly the jumpers VGA can be killed.

Best regards,
Oscar.

Back to top

3Dfx Site - Museo 3Dfx - 3Dfx Collection.&&

IP Logged

osckhar

God Member

Offline

www.3Dfx.es

Posts: 812
SPAIN
Gender: male

Re: AGP.5 Release.
Reply #70 - 29.01.14 at 11:59:03

Hi,

I built a fast sample only for testing if AGP3.3 DUAL ADAPTER with not internal power supply could be handled for PCIx slot when V5-6000 is in full load 8xFSAA.

SIMPLE DESIGN AGP3.3 ONLY FOR TESTING.

SCORE IN FASTEST MODE

I was worried since I was not sure if in full load the PCIx slot would be enough to feed the V5-6000 8xFSAA and yes, card was in looping mode for more than 30 minutes to 8xFSAA and not lock up, not heat.

Best regards,
Oscar.

Back to top

3Dfx Site - Museo 3Dfx - 3Dfx Collection.&&

IP Logged

	Welcome, Guest. Please Login