I did it again – I had to fulfill my inner desire to fix yet another old piece of hardware. This one is actually kind of worthwhile, too – it’s an EVGA e-GeForce 7600 GS AGP video card, model P/N: 256-A8-N542-T2. With the popularity of “retro” PCs from the Win9x to XP era going up, and the diminishing supply of decent AGP cards, it is expected their prices will go up. Or is it? I bought the this video card rather on the cheap side (~$9 USD shipped to my door), because the eBay seller listed it for parts or repair (and correctly noted the bad capacitors.)
So, here is the GPU, as I received it:
Sacon FZ capacitors at their finest again!
I was feeling a bit uneasy, because way back I acquired two other EVGA GeForce 7600 GT cards (PCI-E versions) with bad Sacon FZ caps, thinking they would be easy repairs. They were anything but that. In fact, I never succeeded to fix them, because both of them blew their RAM chips when the caps went bad. I tried replacing those too, and got the cards to sort-of work (with artifacts, at least.) However, this experience has made me always be a bit cautious when dealing with EVGA cards and completely busted Sacon FZ caps. In particular, when Sacon FZ caps go bad, their capacity drops like no other. The Sacon FZ on the 7600 GS card above were no exception – they all measured in the 600-1000 pF (picoFarad) range!!!
Of course, at under $10 and no one else interested in the above 7600 GS, I figured I might as well try to save it from its doomed fate otherwise. Like always, I started by cleaning the card, noting down which caps filter each rail, and then taking high-res pictures, in case something goes up in smoke (after the repair).
https://www.badcaps.net/forum/attach...1&d=1591483007
https://www.badcaps.net/forum/attach...1&d=1591483007
There are 5 major rails on this video card, and I marked the caps for each:
1) 12V rail for GPU VRM high-side (2x 16V, 470 uF caps with a single blue dot)
2) 5V rail for RAM VRM high-side (1x 16V, 470 uF cap with two blue dots)
3) GPU V_core (3x 6.3V, 1500 uF and 1x 6.3V, 1000 uF caps with a single blue stripe)
4) RAM V_dd (2x 6.3V, 1000 uF caps with two blue stripes)
5) GPU V_tt (1x 6.3V, 1500 uF cap with three blue stripes)
I did NOT try powering the video card with the bad caps in this state – not even for a brief test. Instead, I went straight to removing them.
The above pictures should also reveal how the caps are connected.
C29 and C34(?) - input 12V rail (GPU VRM high-side)
C137 - input 5V rail (RAM VRM high-side)
C76, C50, C77, and C71 - GPU V_core (GPU VRM low-side)
C100 and C135 - RAM V_dd (RAM VRM low-side)
C59 - GPU V_tt
As usual, not knowing if the card would work or not, I didn’t put brand new caps on it. Instead, I recapped it with whatever caps I had on hand (all used pulls & tested, except for the single new Rubycon ZLH cap.)
1) 12V rail ---> 1x United Chemicon KZG 16, 1500 uF (in place of C29)
2) 5V rail ---> 1x Rubycon ZLH 6.3V, 820 uF (in place of C137)
3) GPU V_core ---> 2x Nichicon LF 2.5V, 820 uF polymers (C71 & C77) and 1x Nichicon HZ 6.3V, 2200 uF (C50)
4) RAM V_dd ---> 2x Nichicon LF 2.5V, 820 uF polymers (C100 & C135)
5) GPU V_tt --> 1x Nichicon HZ 6.3V, 2200 uF (C59)
Not going to talk about the ESR of my replacement caps, because obviously they are light-years ahead of the Sacon FZ caps (even when they were new.) Capacitance-wise, I did drop the GPU V_core slightly from 5500 uF total before (3x 1500 + 1x 1000) to 3840 uF total now (a little over 30% drop.) But that shouldn’t be an issue, especially for testing. For OC-ing, it would probably be better to have slightly higher capacitance, since this card’s GPU V_core VRM is single-phase only. I’ll probably end up gutting the Nichicon HZ caps, since they have a 2005 date code (known to be affected by the issue Nichicon had with their HM, HN, and HZ series between 2001-2004 and a few from 2005.) Will likely replace them with 2700 uF Rubycon MFZ… or rather, add a second Ruby MFZs on the GPU V_core and put something more regular in the range of 1500-1800 uF on the GPU V_tt, since it’s a linear rail and doesn’t care about ESR as much. That will be for another revision, though.
Now the more interesting question at hand – did the video card work?
Well…
Yes it did.
And all of the caps and VRMs run super-cool.
On the other hand, the GPU chip core is not so cool: 56°C idle with an open case in a warm 28°C (83°F) room temperature. Putting any load on the card for more than 2 seconds easily brought the GPU core to above 60°C, despite the stock fan running full blast all the time (and being very loud, too.) For a GPU chip based on nVidia’s 7000-series, bumpgate damage is guaranteed if the card is used with any significant load. So I did not test it further, aside from a 5-second 3D load test to check for artifacts. The next step would be a heatsink / cooler mod – something I do/show in this thread:
https://www.badcaps.net/forum/showthread.php?t=70502
So, this one is… to be continued.
So, here is the GPU, as I received it:
Sacon FZ capacitors at their finest again!
I was feeling a bit uneasy, because way back I acquired two other EVGA GeForce 7600 GT cards (PCI-E versions) with bad Sacon FZ caps, thinking they would be easy repairs. They were anything but that. In fact, I never succeeded to fix them, because both of them blew their RAM chips when the caps went bad. I tried replacing those too, and got the cards to sort-of work (with artifacts, at least.) However, this experience has made me always be a bit cautious when dealing with EVGA cards and completely busted Sacon FZ caps. In particular, when Sacon FZ caps go bad, their capacity drops like no other. The Sacon FZ on the 7600 GS card above were no exception – they all measured in the 600-1000 pF (picoFarad) range!!!
Of course, at under $10 and no one else interested in the above 7600 GS, I figured I might as well try to save it from its doomed fate otherwise. Like always, I started by cleaning the card, noting down which caps filter each rail, and then taking high-res pictures, in case something goes up in smoke (after the repair).
https://www.badcaps.net/forum/attach...1&d=1591483007
https://www.badcaps.net/forum/attach...1&d=1591483007
There are 5 major rails on this video card, and I marked the caps for each:
1) 12V rail for GPU VRM high-side (2x 16V, 470 uF caps with a single blue dot)
2) 5V rail for RAM VRM high-side (1x 16V, 470 uF cap with two blue dots)
3) GPU V_core (3x 6.3V, 1500 uF and 1x 6.3V, 1000 uF caps with a single blue stripe)
4) RAM V_dd (2x 6.3V, 1000 uF caps with two blue stripes)
5) GPU V_tt (1x 6.3V, 1500 uF cap with three blue stripes)
I did NOT try powering the video card with the bad caps in this state – not even for a brief test. Instead, I went straight to removing them.
The above pictures should also reveal how the caps are connected.
C29 and C34(?) - input 12V rail (GPU VRM high-side)
C137 - input 5V rail (RAM VRM high-side)
C76, C50, C77, and C71 - GPU V_core (GPU VRM low-side)
C100 and C135 - RAM V_dd (RAM VRM low-side)
C59 - GPU V_tt
As usual, not knowing if the card would work or not, I didn’t put brand new caps on it. Instead, I recapped it with whatever caps I had on hand (all used pulls & tested, except for the single new Rubycon ZLH cap.)
1) 12V rail ---> 1x United Chemicon KZG 16, 1500 uF (in place of C29)
2) 5V rail ---> 1x Rubycon ZLH 6.3V, 820 uF (in place of C137)
3) GPU V_core ---> 2x Nichicon LF 2.5V, 820 uF polymers (C71 & C77) and 1x Nichicon HZ 6.3V, 2200 uF (C50)
4) RAM V_dd ---> 2x Nichicon LF 2.5V, 820 uF polymers (C100 & C135)
5) GPU V_tt --> 1x Nichicon HZ 6.3V, 2200 uF (C59)
Not going to talk about the ESR of my replacement caps, because obviously they are light-years ahead of the Sacon FZ caps (even when they were new.) Capacitance-wise, I did drop the GPU V_core slightly from 5500 uF total before (3x 1500 + 1x 1000) to 3840 uF total now (a little over 30% drop.) But that shouldn’t be an issue, especially for testing. For OC-ing, it would probably be better to have slightly higher capacitance, since this card’s GPU V_core VRM is single-phase only. I’ll probably end up gutting the Nichicon HZ caps, since they have a 2005 date code (known to be affected by the issue Nichicon had with their HM, HN, and HZ series between 2001-2004 and a few from 2005.) Will likely replace them with 2700 uF Rubycon MFZ… or rather, add a second Ruby MFZs on the GPU V_core and put something more regular in the range of 1500-1800 uF on the GPU V_tt, since it’s a linear rail and doesn’t care about ESR as much. That will be for another revision, though.
Now the more interesting question at hand – did the video card work?
Well…
Yes it did.
And all of the caps and VRMs run super-cool.
On the other hand, the GPU chip core is not so cool: 56°C idle with an open case in a warm 28°C (83°F) room temperature. Putting any load on the card for more than 2 seconds easily brought the GPU core to above 60°C, despite the stock fan running full blast all the time (and being very loud, too.) For a GPU chip based on nVidia’s 7000-series, bumpgate damage is guaranteed if the card is used with any significant load. So I did not test it further, aside from a 5-second 3D load test to check for artifacts. The next step would be a heatsink / cooler mod – something I do/show in this thread:
https://www.badcaps.net/forum/showthread.php?t=70502
So, this one is… to be continued.
Comment