GPU-Z Perfcap with crippled VCore phase

[didi79]

I have a GTX 970 with 6 phases for VCore. I rebuilt the VRM's, but had to use different MOSFET's. Also 2 of the 6 inductors were replaced: should be R33, but I only had R22.
The card works. Each of the 6 phases work (tested with oscilloscope).
But still GPU-Z shows the card being capped for Perfcap reason Pwr at ~60%.

1. Could that mean that the VRM is still somehow not delivering entirely what is needed ?
2. I can't imagine the "smaller" inductors (less uH) would be the cause. Or am I totally wrong ?

[andreynew1]

Re: GPU-Z Perfcap with crippled VCore phase

Importantly. The inductance just cannot be changed. Many parameters are frequency, resistance, saturation.
XL = 2pfL
R33 = 330 nH 3 mΩ here the current will be low
R22 = 220 nH 0.9 mΩ - here the current will be hight!!!

[didi79]

Re: GPU-Z Perfcap with crippled VCore phase

Importantly. The inductance just cannot be changed. Many parameters are frequency, resistance, saturation.
XL = 2pfL
R33 = 330 nH 3 mΩ here the current will be low
R22 = 220 nH 0.9 mΩ - here the current will be hight!!!

Ok I had no idea, so I will try finding the right parts.
Thank you very much !

[dicky96]

Re: GPU-Z Perfcap with crippled VCore phase

What was wrong with the original R33 inductors? It is really odd for inductors to fail.

[didi79]

Re: GPU-Z Perfcap with crippled VCore phase

I had a hard time getting them off (with my poor soldering skills) and they got damaged.
I wanted to separate the phases. That was before I got my oscilloscope...

Price to pay when learning something new, I guess...

[didi79]

Re: GPU-Z Perfcap with crippled VCore phase

Ok I found similar inductors on a donor card and put them on.
Result is slightly better: still hitting PerfCap Pwr, but now around 70% TDP.

But I think the main problem is something else: Looking at the replacement MOSFET's datasheet again - altough the specs looked similar at first sight - I found out the pulsed drain current of the original MOSFET's was far higher than the ones I'm using now.

So ... more desoldering and soldering to do.

[didi79]

Re: GPU-Z Perfcap with crippled VCore phase

Oddly, that had almost no effect.
In GPU-Z I notice the power draw on the 8-pin connector does not get higher than 60W (card only has one 8-pin). This should be higher, since 4 (or 5? ) phases get their 12V from it.
When stressed, the reported voltage does not seem to drop very much neither. So the power is available, but the card doesn't draw it.

Makes me wonder: could the shunt resistor be bad and as such limiting the current the cards draw ?

[didi79]

Re: GPU-Z Perfcap with crippled VCore phase

Well now the problem has in fact changed: Measuring for PWM signals once more, I noticed I now have a doubler (IR3598) that's gone bad. Maybe because of the heat of all the MJ (MOSFET Jockying :p).

That of course takes out 2 phases 4 out of 6 phases or 4/6 = 66%
And that's about the TDP I get!

[RJARRRPCGP]

Re: GPU-Z Perfcap with crippled VCore phase

Maybe because of the heat of all the MJ (MOSFET Jockying :p).

LOL!

This may be "phantom power throttling syndrome", which I saw before, was caused by a wrong/corrupted VBIOS. I know that can happen with Kepler, the gen right before Maxwell, IIRC.
Had that VBIOS issue with an eVGA GeForce GTX 660 Ti back in 2016. I found out that it had an Asus VBIOS! Then not too long after, it was to TechPowerUp I go! Found a proper eVGA VBIOS.