GF2s have such crappy regulators, what can I do? [Archive]

View Full Version : GF2s have such crappy regulators, what can I do?

craig588

10-26-2005, 02:13 PM

There is a droop of about .5V It's disgusting. The little regulator just can't support all of the current required, without a mod the droop is about .3V, which is still awful.

While in 3D the Vcore bounces all over the place rather quickly between 2.00 and 2.09V, maybe a cap will help with that bit, no?

I wanted to hit some records with this little card, but it's just very sad. The regulator is getting a little warm, but I don't think it's hot enough to really affect the voltage that much. I think it just wasn't powerful enough in the first place.

It's a PCI card so I can use with my PCI-E FX rigs. I'm going to be hooking a cascade up to it, so that might help a bit with the current draw, but I don't think that's going to be enough.

The memory is fed straight from the 3.3 line so that doesn't give me any problems.

What else can I do to help this card?

SikaRippa

10-26-2005, 02:34 PM

What card (exactly) you have ? Are you talking about linear regulator, which model/code ?

craig588

10-26-2005, 02:50 PM

Here's the datasheet I used to make the mod for the card.

I don't have any exact details on the card, I know it's a 32MB one.

SikaRippa

10-26-2005, 03:11 PM

Here's the datasheet I used to make the mod for the card.

I don't have any exact details on the card, I know it's a 32MB one.

OK.. the drop maximum voltage over that regu is ~0.5 V according the datasheet. Is the drop voltage any problem, IF you have enough voltage at output ? What are the input/output voltage levels of regu ?

I wonder a good capacitor (enough capacitance, low ESR) at output would stabilize the variation during 3D.

SoddemFX

10-26-2005, 03:32 PM

A capacitor wont help droop no matter how perfect it is, the best option in my opinion would be to remove it and add in another regulator...

If GF2's can handle 2.6v core then maybe just a diode from the 3.3v rail with a ~220uF capacitor on the output would do? Maybe drop the 3.3v rail a little for a lower vcore? Just ideas...

Tom

brandinb

10-30-2005, 09:14 PM

hey soddemfx how does adding a diod to the 3.3 rail with a cap give you 2.6v i havent heard of this kind of method can you explain it just a little please?
________
GENETICALLY MODIFIED FOOD (http://gmfoods.info/)

SikaRippa

10-31-2005, 02:49 AM

hey soddemfx how does adding a diod to the 3.3 rail with a cap give you 2.6v i havent heard of this kind of method can you explain it just a little please?

Basic operation of a diode.. if a current at SPECIFIED range is feed through a diode the voltage across (VF) remains almost constant. Typical voltage is around 0.7 V, but also other values are available e.g. with schottky diodes.
So, if one put a diode (correct polarity !!, anode to higher voltage) between 3.3 V and Vmem (AND remove all old circuitry between those nodes !!!!) a voltage of ~ 2.6 V is obtained, in theory..

Some problems:

- VF varies between samples/diode types
- VF varies with load (=current)
- Vmem TRACKS input voltage (3.3 V) so it is not REGULATED at all i.e. all noise, ripple etc. appears to Vmem

If somebody is going to use this, please take care to have a diode with PROPER current handling capability (you should have some value of current consumption..)

10-31-2005, 08:09 AM

brandinb

10-31-2005, 09:22 AM

Basic operation of a diode.. if a current at SPECIFIED range is feed through a diode the voltage across (VF) remains almost constant. Typical voltage is around 0.7 V, but also other values are available e.g. with schottky diodes.
So, if one put a diode (correct polarity !!, anode to higher voltage) between 3.3 V and Vmem (AND remove all old circuitry between those nodes !!!!) a voltage of ~ 2.6 V is obtained, in theory..

Some problems:

- VF varies between samples/diode types
- VF varies with load (=current)
- Vmem TRACKS input voltage (3.3 V) so it is not REGULATED at all i.e. all noise, ripple etc. appears to Vmem

If somebody is going to use this, please take care to have a diode with PROPER current handling capability (you should have some value of current consumption..)

thanks for that explanation now i have somewhere to start im gonna do some more research thanks!
________
MEDICAL MARIJUANA CARE MUTUAL DEALS (http://dispensaries.org)

SoddemFX

11-01-2005, 04:19 PM

Its a very shoddy method, but if craig has 0.5v variation now...

Ive seen some old (think it might have been QDI) boards use this to regulate DDR voltage. Its not great, but as SikaRippa has said if you use appropriate diodes in the correct region then it can work.

Personally i wouldnt use it but you work with the resources you have, and the cost of failure is a GF2...

I'd guess at 4A - 6A.

Good luck with it anyway :)

Tom