Yet another eBay special I couldn't resist - an XFX GeForce 8600 GT with 256 MB GDDR3 RAM for $8 total (won the auction for $1).
Why could I not resist, you wonder? Simple answer: more busted Sacon FZ caps.
Now, I learned my lesson long before to be very careful with hardware with busted Sacon FZ that is described as not working (more specifics on that for another thread/day). But the seller of this auction claimed the card was tested and working. He, too, noted that the card needed a recap. So I decided to bite and bought the card.
Here are some pictures of how the card looked like when it came.
And the back:
https://www.badcaps.net/forum/attach...1&d=1460251222
So, usual first order of business: remove heatsink and remove all Sacon FZ caps.
https://www.badcaps.net/forum/attach...1&d=1460251222
Then, I reached in my bag with good caps for testing purposes. And this is what the resultant test recap looked like:
Close-up shots:
https://www.badcaps.net/forum/attach...1&d=1460251222
https://www.badcaps.net/forum/attach...1&d=1460251222
A bit crude? Yes, it is (especially that leaning Fujitsu poly cap, on which I had to add a “prosthetic” leg ). But nonetheless, everything was calculated to meet or exceed specs – even though I only intended this to be a temporary recap.
Now, if there is one thing I like about these older XFX cards is that they provide you with plenty of options regarding what caps you can use. This particular 8600 card has the capacitor spots silkscreened for both through-hole caps (both 8 mm and 10 mm diameter) as well as SMD. Neat! That means we have a lot of choices when it comes to choosing replacement capacitors. And that's the next order of business.
Replacement caps for Sacon FZ
On paper, Sacon FZ series are supposedly equivalent to Nichicon HM and Rubycon MBZ (more or less). But let's be honest: Sacon / Evercon / Elcon / GSC caps are terrible. Therefore, don't let that stop you from using other brands and series of caps, even if they have slightly inferior ESR and ripple current (RC) specs. In my opinion, the following would be absolutely acceptable choices: Panasonic FM, as well as Rubycon ZLG and ZLJ.
Of course, if you have no means to obtain equivalent caps, then even general purpose caps from the Japanese manufacturers may be okay. I measured a few busted FZ caps with my cheapo ESR meter, and all of them read way out of spec (some even open-circuit!). If the card worked with those (at least according to seller, I didn't test it as I didn't want to risk killing the RAM), then just about any functional cap will work. Don't believe me? Here is working proof of this concept:
https://www.badcaps.net/forum/showthread.php?t=43404
On the other hand, if you want to recap this card all with polymers, then be my guest. It should be more than happy .
Anyways, here is some useful information regarding the caps on this 8600 video card and what function they perform:
---- 12V rail (on the card)
This rail feeds both the GPU core and VRAM voltage regulator modules (VRMs). It has five capacitor spots:
CX15/CX42 (before toroid filter coil)
CX10/CX41
CX11/CX40
CX12/CX43
CX44/CX9
Originally, these were 5x Sacon FZ 16 V, 470 uF, 8x12 mm (1150 mA ripple current and 36 mOhms ESR). For my recap, I used only three caps to replace these, but they were adequate for the task (even more than needed, perhaps ).
CX15/CX42 --> Sanyo SEPC polymer (SMD), 16 V, 150 uF, 8x7 mm (3320 mA RC, 22 mOhms ESR)
CX11/CX40 --> Panasonic FL, 16 V, 1500 uF, 10x20 mm (similar specs to Rubycon MCZ below)
CX44/CX9 --> Rubycon MCZ, 16 V, 1500 uF, 10x20 mm (2770 mA RC, 11 mOhms ESR)
---- GPU core rail (a.k.a. GPU V_core or GPU Vcc)
This rail provides power to the GPU chip. IIRC, stock it is at 1.3-1.4 Volts (forgot to measure mine), so for those of you who are poly-modding, you can use capacitors with a voltage rating as low as 2.5V.
The GPU core rail has two capacitors spots that came with the following caps:
CX20/CX18: Chemicon PXA polymer (SMD), 4 V, 1200 uF, 10x12 mm (5500 mA RC, 10 mOhms ESR)
CX25/CX22: Sacon FZ 6.3 V, 1500 uF, 10x12 mm (1800 mA RC, 22 mOhms ESR)
The Chemicon PXA polymer above is probably what kept the card running, despite the other busted FZ caps. Thus I only replaced the busted FZ.
CX25/CX22 --> Nichicon HZ 6.3 V, 2200 uF, 10x20 mm (3770 mA RC, 7 mOhms ESR)
---- VRAM Vdd/Vddq rail
This rail provides power primarily to the RAM chips. Normal operating voltage should be around 1.8 to 2 Volts. Because of this, replacement capacitors with a voltage rating of 4V or more is recommended (2.5V caps will likely work as well, but 4V will keep a better voltage safety margin.)
Originally, this rail came with two FZ 6.3 V, 1500 uF, 10x12 mm caps at spots CX21/CX19 and CX46/CX45. I replaced them with the following:
CX21/CX19 --> Nichicon HZ 6.3 V, 2200 uF, 10x20 mm (3770 mA RC, 7 mOhms ESR)
CX46/CX45 --> Fujitsu FPCAP RE 4 V, 820 uF, 10x12.5 mm (6100 mA RC, 7 mOhms ESR)
---- GPU secondary / PCI-E comm. Rail
Though not 100% sure, I believe this rail is used for communication between PCI-E bus and GPU chip, though I'm not sure. It is usually 1.2V or thereabouts on most nVidia PCI-E cards up to (and including) the 9 series. Thus, any capacitor with a voltage rating of 2.5V or more can be used here. This rail is also linearly derived from the GPU core rail, so the ESR and ripple current of the replacement cap are not that critical – any low-ESR cap will likely work fine. The capacitance, on the other hand, is advisable to be matched within ±30%.
The rail on this card had only one cap at spot CX6/CX1. Originally, it was a Sacon FZ 6.3 V, 1000 uF, 8x12 mm (1150 mA ripple current and 36 mOhms ESR). I used a Nichicon SMD cap here (not sure what series, as it came from dead PS3 board), rated for 4V and 1000 uF. I've done this before on a GeForce 6200 (128-P2-N361) and it hasn't caused any problems.
Results
Unfortunately, after all of this work (looking up cap specs sheets and the recap itself), the video card did not work. And here is why:
See that bottom-right corner? That, folks, is a cracked core! And it shouldn't have been a surprise: after all, the video card came through my mailbox in a single padded mail envelope. No hard box or any other packaging. So my guess would be it broke in the mail. Of course, I can't blame the postal office for this, because it was the seller who should have known better and packed the item more appropriately. I mean, I know I got it for cheap. But that's no excuse to just stick a postal stamp on something and dump it in a mailbox.
Oh well, at least I got a working fan out of this one. The fan (made by ARX) is the same as the one used on the XFX 6800 Xtrem and also an MSI GeForce 7600 GS of mine (CeraDyna series). So at least this wasn't a total rip-off. Anyways, I hope this recap info is still useful to someone on the internets, even though the GeForce 8 series of video cards are quite dated at the time of this post.
Why could I not resist, you wonder? Simple answer: more busted Sacon FZ caps.
Now, I learned my lesson long before to be very careful with hardware with busted Sacon FZ that is described as not working (more specifics on that for another thread/day). But the seller of this auction claimed the card was tested and working. He, too, noted that the card needed a recap. So I decided to bite and bought the card.
Here are some pictures of how the card looked like when it came.
And the back:
https://www.badcaps.net/forum/attach...1&d=1460251222
So, usual first order of business: remove heatsink and remove all Sacon FZ caps.
https://www.badcaps.net/forum/attach...1&d=1460251222
Then, I reached in my bag with good caps for testing purposes. And this is what the resultant test recap looked like:
Close-up shots:
https://www.badcaps.net/forum/attach...1&d=1460251222
https://www.badcaps.net/forum/attach...1&d=1460251222
A bit crude? Yes, it is (especially that leaning Fujitsu poly cap, on which I had to add a “prosthetic” leg ). But nonetheless, everything was calculated to meet or exceed specs – even though I only intended this to be a temporary recap.
Now, if there is one thing I like about these older XFX cards is that they provide you with plenty of options regarding what caps you can use. This particular 8600 card has the capacitor spots silkscreened for both through-hole caps (both 8 mm and 10 mm diameter) as well as SMD. Neat! That means we have a lot of choices when it comes to choosing replacement capacitors. And that's the next order of business.
Replacement caps for Sacon FZ
On paper, Sacon FZ series are supposedly equivalent to Nichicon HM and Rubycon MBZ (more or less). But let's be honest: Sacon / Evercon / Elcon / GSC caps are terrible. Therefore, don't let that stop you from using other brands and series of caps, even if they have slightly inferior ESR and ripple current (RC) specs. In my opinion, the following would be absolutely acceptable choices: Panasonic FM, as well as Rubycon ZLG and ZLJ.
Of course, if you have no means to obtain equivalent caps, then even general purpose caps from the Japanese manufacturers may be okay. I measured a few busted FZ caps with my cheapo ESR meter, and all of them read way out of spec (some even open-circuit!). If the card worked with those (at least according to seller, I didn't test it as I didn't want to risk killing the RAM), then just about any functional cap will work. Don't believe me? Here is working proof of this concept:
https://www.badcaps.net/forum/showthread.php?t=43404
On the other hand, if you want to recap this card all with polymers, then be my guest. It should be more than happy .
Anyways, here is some useful information regarding the caps on this 8600 video card and what function they perform:
---- 12V rail (on the card)
This rail feeds both the GPU core and VRAM voltage regulator modules (VRMs). It has five capacitor spots:
CX15/CX42 (before toroid filter coil)
CX10/CX41
CX11/CX40
CX12/CX43
CX44/CX9
Originally, these were 5x Sacon FZ 16 V, 470 uF, 8x12 mm (1150 mA ripple current and 36 mOhms ESR). For my recap, I used only three caps to replace these, but they were adequate for the task (even more than needed, perhaps ).
CX15/CX42 --> Sanyo SEPC polymer (SMD), 16 V, 150 uF, 8x7 mm (3320 mA RC, 22 mOhms ESR)
CX11/CX40 --> Panasonic FL, 16 V, 1500 uF, 10x20 mm (similar specs to Rubycon MCZ below)
CX44/CX9 --> Rubycon MCZ, 16 V, 1500 uF, 10x20 mm (2770 mA RC, 11 mOhms ESR)
---- GPU core rail (a.k.a. GPU V_core or GPU Vcc)
This rail provides power to the GPU chip. IIRC, stock it is at 1.3-1.4 Volts (forgot to measure mine), so for those of you who are poly-modding, you can use capacitors with a voltage rating as low as 2.5V.
The GPU core rail has two capacitors spots that came with the following caps:
CX20/CX18: Chemicon PXA polymer (SMD), 4 V, 1200 uF, 10x12 mm (5500 mA RC, 10 mOhms ESR)
CX25/CX22: Sacon FZ 6.3 V, 1500 uF, 10x12 mm (1800 mA RC, 22 mOhms ESR)
The Chemicon PXA polymer above is probably what kept the card running, despite the other busted FZ caps. Thus I only replaced the busted FZ.
CX25/CX22 --> Nichicon HZ 6.3 V, 2200 uF, 10x20 mm (3770 mA RC, 7 mOhms ESR)
---- VRAM Vdd/Vddq rail
This rail provides power primarily to the RAM chips. Normal operating voltage should be around 1.8 to 2 Volts. Because of this, replacement capacitors with a voltage rating of 4V or more is recommended (2.5V caps will likely work as well, but 4V will keep a better voltage safety margin.)
Originally, this rail came with two FZ 6.3 V, 1500 uF, 10x12 mm caps at spots CX21/CX19 and CX46/CX45. I replaced them with the following:
CX21/CX19 --> Nichicon HZ 6.3 V, 2200 uF, 10x20 mm (3770 mA RC, 7 mOhms ESR)
CX46/CX45 --> Fujitsu FPCAP RE 4 V, 820 uF, 10x12.5 mm (6100 mA RC, 7 mOhms ESR)
---- GPU secondary / PCI-E comm. Rail
Though not 100% sure, I believe this rail is used for communication between PCI-E bus and GPU chip, though I'm not sure. It is usually 1.2V or thereabouts on most nVidia PCI-E cards up to (and including) the 9 series. Thus, any capacitor with a voltage rating of 2.5V or more can be used here. This rail is also linearly derived from the GPU core rail, so the ESR and ripple current of the replacement cap are not that critical – any low-ESR cap will likely work fine. The capacitance, on the other hand, is advisable to be matched within ±30%.
The rail on this card had only one cap at spot CX6/CX1. Originally, it was a Sacon FZ 6.3 V, 1000 uF, 8x12 mm (1150 mA ripple current and 36 mOhms ESR). I used a Nichicon SMD cap here (not sure what series, as it came from dead PS3 board), rated for 4V and 1000 uF. I've done this before on a GeForce 6200 (128-P2-N361) and it hasn't caused any problems.
Results
Unfortunately, after all of this work (looking up cap specs sheets and the recap itself), the video card did not work. And here is why:
See that bottom-right corner? That, folks, is a cracked core! And it shouldn't have been a surprise: after all, the video card came through my mailbox in a single padded mail envelope. No hard box or any other packaging. So my guess would be it broke in the mail. Of course, I can't blame the postal office for this, because it was the seller who should have known better and packed the item more appropriately. I mean, I know I got it for cheap. But that's no excuse to just stick a postal stamp on something and dump it in a mailbox.
Oh well, at least I got a working fan out of this one. The fan (made by ARX) is the same as the one used on the XFX 6800 Xtrem and also an MSI GeForce 7600 GS of mine (CeraDyna series). So at least this wasn't a total rip-off. Anyways, I hope this recap info is still useful to someone on the internets, even though the GeForce 8 series of video cards are quite dated at the time of this post.
Comment