Announcement

Collapse
No announcement yet.

XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

    Here is another eBay XFX $9 special video card with more blown Sacon FZ capacitors. This time, a GeForce 9600 GSO model PV-T960-SDFH V1.4, based on the G92 GPU core (96 shaders, 192-bit 768 MB of DDR2 RAM).

    Overview front and back:



    Label:


    The bad boys:


    Let's see who does what – a cap diagram:


    I think it should be pretty clear, but in case it's not:
    starting from the bottom (now we here? ) right corner with the 6-pin PCI-E power connector, we have 12V power directly from PSU feeding into two filter capacitors C214/C217 and C215/C216. On that note regarding the double-labeling: this is because the PCB can take either SMD or through-hole caps, with the option of 8 or 10 mm diameter – a nice feature of XFX boards, as always. Now if only XFX actually cared to use good caps in the first place… Anyways, these two caps filter 12V power that goes to the GPU V_core buck regulator (i.e. GPU core VRM).

    Meanwhile, the two cap spots at C202/C137 and C208/C138 get 12V power from the PCI-E connector on the motherboard. The caps feed into the RAM Vdd buck regulator. Even though these caps are also connected to the 12V rail, it's worth noting that putting good caps in spots C214/C217 and C215/C216 for the GPU buck regulator will *not* provide any filtering to the RAM buck regulator high side. So having at least one good cap in spots C202/C137 or C208/C138 is required.

    Moving on up, there are three caps that filter the GPU V_core rail (but spots for four caps). These are C224/C254, C225/C229, C253/C244, and empty spot C245 (SMD spot only). Only one blown Sacon FZ cap on this rail. Because of the two Chemicon PSC polymers, these 9600 GSO cards will still continue to work even with the blown FZ cap there.

    To the left of the GPU VRM section/caps are the RAM filter caps: two 8/10 mm spots that hold either SMD or through-hole parts again. These are CX31/CX32 and CX34. Both of these were Sacon FZ caps on my card… and blown.

    Finally, there's the RAM Vtt rail. This is derived from the RAM Vdd rail via linear regulation. Originally, this rail did not have an electrolytic filter capacitor (only the SMD ceramics on the board). But there was a spot there, so I filled it. Why? Because some “higher-end” 9600 GSO and 8800 GS models based on this PCB have a cap installed there. Thus, I figured it wouldn't hurt anything (and maybe even help with overclocking, if doing any – not that I would, though).

    This being said, it is worth noting here that not all XFX 9600 GSO cards come with Sacon FZ caps. Actually, most come with Panasonic ultra-low ESR caps (not sure which series, but probably FL, FJ, or FJS judging by the pictures I've seen). And also worth noting that XFX uses this same PCB for the GeForce 8800 “GS”, which is XFX's weird spin-off of the G92 core – i.e. probably another name for the 9600 GSO. I think even some XFX 8800 GT and 9800 GT use this PCB, though I'm not 100% sure. In any case, the “better” cards typically come with Chemicon PSC polymers everywhere. On occasion, however, I've also seen them come with the Panasonic caps mentioned above. Thus, it looks like XFX use whatever parts (GPUs, RAM, and caps) they have available at the time and just re-tier the card afterwards. But when it comes to the PCB, I don't think there are many differences between the various abovementioned models, save for some fan header options perhaps (my 9600 GSO only comes with a 2-pin fan, for example).

    On that note, let's talk about the cooling on this card: it's pretty LOUD. Not the most obnoxious I've heard, but annoying nevertheless. This is mainly due to the fan running at 100% all the time. On the positive side, at least this allowed the card to stay somewhat cool, even when the heatsink was fairly clogged with dust when I received it:


    Of course, the above heatsink is not enough to keep the card under 60°C under full 100% load at normal room temperatures. But when v-sync is enabled in certain games that the card can run very well (i.e. the GPU load being no higher than 70%), the card will stay in the high 50's degrees Celsius and rarely top 60°C. Not great results again, but somewhat more acceptable, especially considering the cooler takes only a single slot. Regarding the cooler, it's all-Copper (including the fins) and fairly heavy for what it is. However, IMO the cooler could have been made to have double the fins (and hence almost double the surface area), as the fan is a fairly high-pressure squirrel-cage type with 50 mm diameter – it would have pushed air through more dense fins just fine and thus allow the card to run cooler.

    And here is a shot of the heatsink from the bottom

    Not much to talk about here. Though I have to say I find it a bit amusing that there are 4 cooling pads for the RAM when there are only 3 sets of RAM chips on the top of the card. Again, this is probably because XFX uses the same cooler on other models of the card too, and perhaps some of them come with 4 sets of RAM chips on the top.

    Finally, a picture of the temporary test recap I did:
    https://www.badcaps.net/forum/attach...1&d=1543298909

    The card worked, so I revised the temporary recap a bit.

    I used two Chemicon PSA 4V, 820 uF polymers (10x13 mm) from an Xbox 360 motherboard for the RAM buck filter on this card (i.e. spots CX31/CX32 and CX34). Meanwhile, the GPU V_core rail only had its lone bulged Sacon FZ cap replaced with a Chemicon KZG, 6.3V, 3300 uF cap. As can be seen in the first temporary recap, there were actually two Chemicon KZG 6.3V, 3300 uF caps (one for the RAM output filter and the other same place as it is now). Reason I did this is because the RAM showed only 2-3 Ohms resistance when I got the card with the bad FZ caps. On most cards, I find that the RAM VRM output typically reads at least 15 Ohms. Thus, I though the bad Sacon FZ caps on the RAM buck output had killed the RAM. Luckily, that wasn't the case. When I saw the card working, I replaced the lone Chemicon KZG 6.3V, 3300 uF cap on the RAM with something a “tiny bit” more reliable: the two Chemicon PSA polymers, shown above.
    As for the RAM buck input (12V), I used a single Sanyo SEP, 16V, 330 uF cap from a dead (bad GPU) GeForce 7900 GS video card. And finally, the RAM Vtt rail received a lone Teapo SC, 6.3V, 1000 uF cap. Yes, I know , it's a Teapo, and SC series at that. But as I mentioned, this rail did not have a cap before, so XFX probably designed the card to work okay even without it. However, I just couldn't stand seeing that cap spot left empty. Also, I have a bag of these Teapo SC caps back from 2012 when I did a few ASUS P4SD motherboard recaps for some HP DC5000 computers. Since then, I've not yet had one of these Teapo caps fail on me, even in storage with no use. Their specs also still appear to be fairly okay on my ESR meter. So for non-critical use, I do re-use them. In this case, I was curious to see how the Teapo SC would hold up on a video card, given that it is only filtering a low-power linear rail with very little noise.

    Of course, if this card was going to be given to someone else, I would have recapped it properly. But being that I haven't even found a good-matching PC for it yet, it hardly gets any use. As such, I felt a bit experimental and that's why I used the Teapo SC and Chemicon KZG caps above.

    Anyways, that's all I got for this card. Another one saved from the Sacon FZ curse.
    I'll probably also do a cooler mod on this card at some point. But that will be later on in the future, when I find more use of the card and when I get some temperature graphs for it to see what cooler/fan combo will work best for it.
    Attached Files

    #2
    Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

    You noted in this post, over 3 years ago now, that you had just such a 1000uF 6.3V 8mm Teapo SC bulge in storage. Or did you forget about that?
    Last edited by Wester547; 11-27-2018, 12:41 AM.

    Comment


      #3
      Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

      Originally posted by Wester547 View Post
      You noted in this post, over 3 years ago now, that you had just such a 1000uF 6.3V 8mm Teapo SC bulge in storage. Or did you forget about that?
      Nope, I have not. In fact, I still keep that bag. Actually, I have those Teapos divided in two bags: one with the "good ones" and the other with the questionable ones (because many of them appeared to have come from really overheated boards). So far, I've reused several of those Teapo SC caps from the "good" bag and had no issues yet. Of course, most of these were all cherry-picked from P4SD boards that didn't appear overheated or with too many hours.

      As for the bulged SC cap... it came out of the questionable bag. In it, I also had some Teapo SZ 6.3V 1500 uF - all pulled from the same P4SD motherboards. And since then, I've had all (4) of the Teapo SZ bulge and leak too (not surprised at all, though - these are all very low ESR and quite unstable).

      Anyways, I'm still going to keep the lone Teapo SC 1000 uF cap on this 9600 GSO card. As mentioned, there was nothing in its spot on the video card, so surely it can't hurt things (unless it short-circuits, which I just haven't seen with small Teapo SC). When I find some long-term use for this video card, I might recap it then. Though in all honesty, I can't say what will fail first: that Teapo SC cap or the GPU chip, given that it's G92 core (not terrible, but can't handle heat too well, and the stock cooler is a little on inadequate side).
      Last edited by momaka; 11-29-2018, 10:46 PM.

      Comment


        #4
        Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

        Originally posted by momaka View Post
        Though in all honesty, I can't say what will fail first: that Teapo SC cap or the GPU chip, given that it's G92 core (not terrible, but can't handle heat too well, and the stock cooler is a little on inadequate side).
        Neither. The 3300uF 6.3V KZG you decided to use on that card will fail first. (unless it was manufactured after 2007)

        I would guess the caps in the good bag lasted longer than the caps in the questionable bag because 1) heat exponentially accelerates chemical reactions and 2) the long term ripple current probably caused the electrolyte to break down by way of electrolysis (due to insufficient oxidizers and depolarizers), causing the aforementioned caps to outgas faster even when sitting. But it could simply be due to Teapo's inconsistent QC rather than anything else. I'm less taken aback by XFX's choice of crap caps and more the fact that they decided to mix Chemi-con PSA and Sanyo OS-CON SVP polymers with rock-bottom Sacon FZs, and then other times use Panasonics. I guess they really do just use whatever's available in their inventory.

        Comment


          #5
          Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

          i'd say the qimonda-gate vram fails first on that card! lol!

          anyway, the 9600 gso (g92) is merely a rebrand of the 8800 gs. there is also another 9600 gso that uses the faulty g94 core with only 48 working cores instead called the 9600 gso 512. very confusing naming schemes to bin all their faulty chips huh, nvidia?

          despite being a rebrand, the 8800 gs has a tdp of 105w, while the 9600 gso (g92) has a tdp of 84w. i guess one must use the 65nm g92a chip while the other uses the 55nm g92b chip, hence the lower tdp despite being clocked the same.

          and yes that cooler is indeed inadequate for handling 84w of tdp. much less even the 105-125w tdp of the 8800 gt and 9800 gt. i think for sure the same crappy single slot blower cooler is used also on the 8800/9800 gt cards! i say this because of the four sets of thermal pads on the cooler to make up a 256 bit mem bus which is what is on the 8800/9800 gt cards. the 8800 gs and 9600 gso (g92) only have 192 bit mem buses so 1 set of ram chips is missing.

          because the pcb is double sided for soldering mem chips on both sides of the board, u can have only mem chips soldered on one side to make up either a 384mb 8800 gs/9600 gso (g92) or 512mb 8800/9800 gt. if u solder mem chips on both sides of the board, u end up with either 768 mb 8800gs/9600 gso (g92) or 1gb 8800gt/9800 gt. so xfx made a flexible pcb design that can be used with multiple video card models with varying vram versions! so thats one pcb to rule them all! to dump all your faulty g92 chips with only 112 cores or 96 cores, that is! lol!

          note that the g94 based 9600 gso 512 doesnt use this pcb. the pinout and bga ball pitch of the g94 would be different from the g92. so the 9600 gso 512 and 9600 gs would probably use the 9600 gt pcb instead.

          anyway, for 84w of tdp, something like the quad heatpipe zalman vf1000 or dual fan thermaltake duorb would be more appropriate. i dont think that xbox gpu or cpu cooler with that one thick heatpipe that curves up or sticks up would be appropriate as that struggles even to cool a 60w tdp 6800 xt.

          Comment


            #6
            Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

            Originally posted by Wester547 View Post
            Neither. The 3300uF 6.3V KZG you decided to use on that card will fail first. (unless it was manufactured after 2007)
            Ha, good catch!
            It's only one, though, and it's backed up by two PSA polymers. So even if it fails, the card will continue to work.

            Originally posted by Wester547 View Post
            I'm less taken aback by XFX's choice of crap caps and more the fact that they decided to mix Chemi-con PSA and Sanyo OS-CON SVP polymers with rock-bottom Sacon FZs, and then other times use Panasonics.
            Yeah, it makes no sense, really. They could have gone with cheapo CapXon polymers or similar, and it probably would have been all fine (or at least better).

            Originally posted by ChaosLegionnaire View Post
            i'd say the qimonda-gate vram fails first on that card! lol!
            Doubt it.

            Originally posted by ChaosLegionnaire View Post
            anyway, the 9600 gso (g92) is merely a rebrand of the 8800 gs. there is also another 9600 gso that uses the faulty g94 core with only 48 working cores instead called the 9600 gso 512. very confusing naming schemes to bin all their faulty chips huh, nvidia?
            I saw one like that (9600 GSO 512) the other day on eBay and wondered why it looked different and without a PCI-E connector. Now it makes sense.
            *EDIT*
            here is a picture of it (not mine, BTW):
            https://www.badcaps.net/forum/attach...1&d=1543894827

            Originally posted by ChaosLegionnaire View Post
            i think for sure the same crappy single slot blower cooler is used also on the 8800/9800 gt cards!
            Yes, it's very similar, if not the same.

            Strangely enough, this cooler by XFX appears to do better than some of the stock reference 8800 GT / 9800 GT coolers (there are two versions of the reference cooler, though). And I'm not even sure how, because this XFX one doesn't even have a heat pipe or coolant chamber - just plain copper base with copper fins.

            Originally posted by ChaosLegionnaire View Post
            because the pcb is double sided for soldering mem chips on both sides of the board, u can have only mem chips soldered on one side to make up either a 384mb 8800 gs/9600 gso (g92) or 512mb 8800/9800 gt. if u solder mem chips on both sides of the board, u end up with either 768 mb 8800gs/9600 gso (g92) or 1gb 8800gt/9800 gt. so xfx made a flexible pcb design that can be used with multiple video card models with varying vram versions! so thats one pcb to rule them all!
            Yup.
            It's pretty clever they did that, though. Rather than design multiple PCBs and then have different assembly lines to make different cards, they can just have one assembly line and make whatever card they got parts for that day/week/month with just a few tweaks on the assembly machines. No need for retooling at all. So I do have to give them credit for their PCB design - it's well-done. Too bad they didn't use better caps and cooler.

            Originally posted by ChaosLegionnaire View Post
            i dont think that xbox gpu or cpu cooler with that one thick heatpipe that curves up or sticks up would be appropriate as that struggles even to cool a 60w tdp 6800 xt.
            Eh, we will see about that.
            The Xbox 360 rev. 2 CPU cooler (the one with the copper heat pipe) can just about manage to cool an 85W TDP Pentium 4 Prescott CPU (100W peak) with a strong 70 mm fan. So I probably will test one of these coolers on this card or perhaps another g92 card of mine.
            Attached Files
            Last edited by momaka; 12-03-2018, 09:41 PM.

            Comment


              #7
              Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

              Hello, nice work.
              I'm on the same recap process.
              Is still working the card?
              Could you please post the list of final cap configuration like "cx337cx34: 820uf/4v"
              Thanks in advance.

              Comment


                #8
                Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

                Is this a 'repair it for the fun of it' or 'repair it to learn something' thread? I am guilty of lots of those so not complaining at all. I just wondered as these cards can be bought fully working for £25 or best offer - with free postage

                https://www.ebay.co.uk/itm/384MB-Gen...aa5a7150f8f2e6
                Follow me on YouTube
                ------------------
                Learn Electronics Repair
                https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                Comment


                  #9
                  Re: XFX GeForce 9600 GSO [PV-T960-SDFH V1.4] recap info

                  Originally posted by hamelkarth View Post
                  Hello, nice work.
                  I'm on the same recap process.
                  Is still working the card?
                  Thanks.

                  Yes, the card is still working. Though, admittedly, I haven't really used it much as I still haven't upgraded the cooler and the original cooler is not adequate, like on all other 9600 GSO and 8800/9800 GS/GT cards.

                  Originally posted by hamelkarth View Post
                  Could you please post the list of final cap configuration like "cx337cx34: 820uf/4v"
                  Sure thing.

                  C202/C137 ---> Sanyo SEP, 16V, 330 uF.
                  C208/C138 ---> nothing / empty
                  C224/C254 ---> United Chemicon PSA, 4V, 820 uF, 10x13 mm
                  C225/C229 ---> no change (left original Chemicon PSA 1200 uF cap)
                  C253/C244 ---> no change (left original Chemicon PSA 1200 uF cap)
                  empty SMD spot C245 ---> no change (still empty.)
                  CX31/CX32 ---> United Chemicon PSA, 4V, 820 uF, 10x13 mm
                  CX34 ---> United Chemicon PSA, 4V, 820 uF, 10x13 mm
                  C258/C259 ---> Rubycon ZLH, 6.3V, 820 uF, 8x13 mm

                  Originally posted by dicky96 View Post
                  Is this a 'repair it for the fun of it' or 'repair it to learn something' thread? I am guilty of lots of those so not complaining at all. I just wondered as these cards can be bought fully working for £25 or best offer - with free postage
                  Definitely the "repair for fun" category.

                  I don't repair computers for a living anymore (tried it only briefly at several repair shops, but it's not really worth it in this area.) Still, I do miss it. So now it's my hobby (and I occasionally do it for friends and relatives, if needed), as is "up-cycling" in general.

                  These cards aren't any good for modern gaming. But for older games, they still do alright. And the fact that I can grab a nice mid-range or high-end card from 10 years ago for less than $15 shipped to my door when I know they originally cost several $100... kind of cool! Bonus points if it's broken / deemed worthless and I can repair it for cheap.

                  As stated in the first post, I got the above card for $9 - that is, including shipping to my door. With a bit of time and spare caps, I got it running again. So why buy working when I can get a broken one and fix it.

                  And if anyone thinks this is a "worthless" repair, you should see some of the other things I have fixed.
                  I even got a whole thread on that.
                  https://www.badcaps.net/forum/showth...ight=worthless

                  Comment

                  Working...
                  X