Announcement

Collapse
No announcement yet.

3dfx Voodoo5 glitches

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

    #41
    Re: 3dfx Voodoo5 glitches

    Originally posted by Justin1091 View Post
    Some '472' resistor networks read 4.50. No idea if that's within the tolerance limit of the resistor networks. That is the last thing I can think of, case closed for now .
    If that reading is in-circuit, that's usually normal. Most resistors will/should read lower resistance in-circuit due to possible other parallel resistances with them. If the resistance reads on the high-side in-circuit (that is, over 5%, or whatever tolerance the resistor specifies), then it's probably bad.

    Out of circuit readings, though, the resistor resistance should be close to the stated value (that is, within its %1/2%/5% tolerance).

    So your resistors reading lower is not necessarily a concern. But if you want to be sure, remove them from the card and measure them again. Most likely, they are fine.

    Originally posted by Justin1091 View Post
    Cards I put Into the oven last a couple of months before I have to 'bake' them again (Nvidia 9800GT and AMD 6870).
    If those are GeForce 9800 GT cards with single slot coolers, it's no surprise they fail again. To date, I haven't found ONE 9800 GT card with a single-slot cooler that runs cool. Most just bake themselves to death, even in idle mode (typically often starting at 56-58°C). Under load, it's not uncommon to see them go up in the 80-90°C range, and that is absolutely unacceptable. So again, no wonder they fail.

    Your best bet is to get one that still works and hasn't failed or been reflow yet, and replace that shitty single-slot cooler with something better. Better yet, see if you can undervolt the GPU chip a little like they do with the 9800 GT Green/Eco/Environmentally-Friendly editions (they are basically same as the regular 9800 GT cards, but just slightly underclocked and undervolted, which brings their overall TDP in the 60-70W range - probably still too high for the stock single-slot cooler, though). If you put a nice big cooler on one of those and keep them under 55°C at all times, they will last a long time.

    Same goes for the AMD cards (I think), or at least the older ones like HD4k and older. I'm not sure about the HD5k cards, but HD6k, 7k, and 8k are definitely not as reliable and need to run even cooler. The x1k and x800/x700/x600 series are also iffy.
    Last edited by momaka; 06-06-2018, 11:14 AM.

    Comment


      #42
      Re: 3dfx Voodoo5 glitches

      Originally posted by Per Hansson View Post
      You can use the Voodoo5 6000 VSA chips on a 5500 or 4500.
      That's what's in the eBay link below, but the aliexpress link is for the regular VSA-100 chips used on the 5500.

      https://www.aliexpress.com/item/100-...935691345.html

      https://www.ebay.com/itm/A-TRAY-OF-2...0/273257152276

      If you need help to solder it you can try contacting member Th3_uN1Qu3 and see if he'd be willing to do it for you.
      Of course there is always the possibility that the issue is not with the VSA-100 chips...
      Thanks, I might do that. I put it in the freezer and it didn't work. Also tried Myth 1, it froze and I saw vertical lines and blocks all over the screen. With GBC off the game played but had artifacts near water ingame and the HUD. I played the game a couple of years ago with the good Voodoo5 and had no problems. Will need to try it again with the same driver to be sure though.

      So, suppose it's not the VSA-100 chips, below is a list of some of the components on the card. (Credits to PCBONEZ @ vogons). Other than the resistor networks, resistors and small smd capacitors (ceramic?) these are the only ones that could be bad. I tested the Q1-Q4's and Sparkey said they're probably fine. Maybe someone can tell me what other components **could** be bad? Probably asking for something that's not possible . If those components are probably ok, then I think we can assume it's the second VSA-100 chip. Not sure yet, but if it's not too expensive (taxes, shipping from China) I might buy one or two of them.

      Comment


        #43
        Re: 3dfx Voodoo5 glitches

        Originally posted by momaka View Post
        If that reading is in-circuit, that's usually normal. Most resistors will/should read lower resistance in-circuit due to possible other parallel resistances with them. If the resistance reads on the high-side in-circuit (that is, over 5%, or whatever tolerance the resistor specifies), then it's probably bad.
        Yeah it was in-circuit, but compared to the good voodoo5 a couple of them are lower. Good Voodoo5 reads 4.68-4.74 for all of them, bad one has a couple of them that read 4.50.
        Out of circuit readings, though, the resistor resistance should be close to the stated value (that is, within its %1/2%/5% tolerance).

        So your resistors reading lower is not necessarily a concern. But if you want to be sure, remove them from the card and measure them again. Most likely, they are fine.
        I could do that, but I tried desoldering one on a dead sound card to practice and I couldn't do it without burning it's pins (even with lowest temperature on my soldering iron ~200C).
        If those are GeForce 9800 GT cards with single slot coolers, it's no surprise they fail again. To date, I haven't found ONE 9800 GT card with a single-slot cooler that runs cool. Most just bake themselves to death, even in idle mode (typically often starting at 56-58°C). Under load, it's not uncommon to see them go up in the 80-90°C range, and that is absolutely unacceptable. So again, no wonder they fail.

        Your best bet is to get one that still works and hasn't failed or been reflow yet, and replace that shitty single-slot cooler with something better. Better yet, see if you can undervolt the GPU chip a little like they do with the 9800 GT Green/Eco/Environmentally-Friendly editions (they are basically same as the regular 9800 GT cards, but just slightly underclocked and undervolted, which brings their overall TDP in the 60-70W range - probably still too high for the stock single-slot cooler, though). If you put a nice big cooler on one of those and keep them under 55°C at all times, they will last a long time.

        Same goes for the AMD cards (I think), or at least the older ones like HD4k and older. I'm not sure about the HD5k cards, but HD6k, 7k, and 8k are definitely not as reliable and need to run even cooler. The x1k and x800/x700/x600 series are also iffy.
        Yep, a single slot cooler. It did last ~7 years and always ran damn hot (85-90 degrees C while gaming). BTW it starts to throttle at precisely 105 degrees C, stupid right? I never put attention to it because nVidea said it's fine and designed to run that hot. After replacing the thermal paste it got to 70 degrees C while gaming but failed after a couple of months again. Rebaked it, it ran for another year before stopping. Windows wouldn't detect it anymore and baking didn't work anymore.

        The AMD 6870 still runs after baking it a year ago, thanks for the tip keeping it cooler. My brother uses it right now and I'll see if I can increase the fan speed or apply a different thermal paste.

        My own GTX 970' fans only start at 60 degrees C so I use MSI Afterburner to make them start earlier, really hard to keep it under 70C though. Guess the heatsinks/fans they use are shitty, I don't understand why they program it to only spin at 60 degrees C, it hardly makes any noise.
        Hmm, this makes me think of replacing the good Voodoo5's ancient heatsinks. Garbage epoxy they used in 2000 probably doesn't conduct well anymore, quite a challenge to take them off, maybe I'll try the freezer to pop them off. There are two holes in the PCB so it's possible to put a good heatsink on it. Anyway that'll have to wait, not done with this bad one yet!
        Last edited by Justin1091; 06-06-2018, 04:18 PM.

        Comment


          #44
          Re: 3dfx Voodoo5 glitches

          Originally posted by Justin1091 View Post
          So, suppose it's not the VSA-100 chips, below is a list of some of the components on the card. (Credits to PCBONEZ @ vogons). Other than the resistor networks, resistors and small smd capacitors (ceramic?) these are the only ones that could be bad. I tested the Q1-Q4's and Sparkey said they're probably fine. Maybe someone can tell me what other components **could** be bad?
          Q1-Q5, U4, U12, and AR1 (358 op-amp) are all power-related components. If any of these are bad, in most cases your video card wouldn't work at all and also probably trip the power supply short-circuit protection, as that's usually what happens when power components go bad.

          As for the other parts...
          U1 likely has something to do with just driving the monitor output. The artifacts you're seeing cannot come from this chip, because they only appear to come in specific scenarios (mainly in 3D mode / games). So that's not it.
          U6... definitely not, as that is probably used for TTL logic between various chips on the board, not 2D/3D data.
          U7... maybe. Depends what IC(s) are connected to it... but I'm doubtful about this being the problem too. Perhaps if this chip has something to do with transferring data between the two GPU chips, then it's possible it may be corrupting some data. Otherwise, probably not.
          U8 - definitely not. This is your BIOS chip. If it was bad, the video card wouldn't boot. Only exception is if the BIOS is corrupted and parameters regarding the RAM timing or GPU/RAM clocks are not right.

          So with that said, there's just very little chance any of these are causing the artifacts. Again, artifacts appear usually when texture/geometry data becomes corrupted in the buffer or GPU. Thus, I think the problem is more than likely either faulty GPU chip or faulty RAM chip(s) corrupting data somewhere. There's also a small chance the ceramic caps on the board may be getting too old and loosing their capacitance (especially at elevated temperatures), but I doubt it.

          By the way, good to hear PCBONEZ is still out there posting stuff. He used to be here and often posted lots of good info.

          Comment


            #45
            Re: 3dfx Voodoo5 glitches

            Originally posted by Justin1091 View Post
            Yeah it was in-circuit, but compared to the good voodoo5 a couple of them are lower. Good Voodoo5 reads 4.68-4.74 for all of them, bad one has a couple of them that read 4.50.
            I've seen much worse variation on in-circuit readings on same video cards before (albeit newer cards, but still). To compare the two cards truly, they must use the same RAM chips (same manufacturer, type, and speed) and made during similar periods. Reason I say this is because I often see RAM chips read different resistance. Sometimes, even a newer vs. older batch of the same RAM type can have different readings. And things like that tend to make in-circuit readings different.

            Originally posted by Justin1091 View Post
            I could do that, but I tried desoldering one on a dead sound card to practice and I couldn't do it without burning it's pins (even with lowest temperature on my soldering iron ~200C).
            I usually use my unregulated 35 Watt wall iron for all small SMD components like that, and I've never had a problem. The key is to use a proper-shaped tip and lots of flux. Tip should be either chisel/bevel or spoon -shaped. Also add fresh solder to the ends of the SMD components before trying to remove them. Practice on a junk board, and you will get the hang of it. I sometimes even prefer doing SMD work over regular through-hole - that's how much I'm used to it now.

            Originally posted by Justin1091 View Post
            Yep, a single slot cooler. It did last ~7 years and always ran damn hot (85-90 degrees C while gaming). BTW it starts to throttle at precisely 105 degrees C, stupid right? I never put attention to it because nVidea said it's fine and designed to run that hot.
            Yeah, nVidia and ATI started spreading lies way back that it's fine for their chips to run at high temperature. The truth is, their chips can only withstand those temperatures... but for how long, of course they don't tell you, because it's well-known fact that BGA fails after prolonged use at higher temperatures.

            The fact that your 9800GT lasted close to 7 years is pretty amazing though, considering the high temperatures. But still, it's a shame, because a chip like that could have lasted much longer if cooled properly.

            Originally posted by Justin1091 View Post
            After replacing the thermal paste it got to 70 degrees C while gaming but failed after a couple of months again. Rebaked it, it ran for another year before stopping. Windows wouldn't detect it anymore and baking didn't work anymore.
            Yup, that's how they always end.

            Originally posted by Justin1091 View Post
            My own GTX 970' fans only start at 60 degrees C so I use MSI Afterburner to make them start earlier, really hard to keep it under 70C though. Guess the heatsinks/fans they use are shitty
            They are not exactly shitty. Just not enough to cool that particular video card / TDP. Put the same heatsink on a GTX 960 or a 950, and watch it perform just fine.

            Originally posted by Justin1091 View Post
            I don't understand why they program it to only spin at 60 degrees C, it hardly makes any noise.
            In the last few years, manufacturers have started to use crap fans that don't last quite frequently. So to prevent them from failing sooner, they turn them off at "lower" temperatures.

            That said, if your fan(s) are Delta or Nidec, or any other brand with ball-bearings, I wouldn't spare them at all - crank them up. They should be fine.

            Originally posted by Justin1091 View Post
            Hmm, this makes me think of replacing the good Voodoo5's ancient heatsinks. Garbage epoxy they used in 2000 probably doesn't conduct well anymore, quite a challenge to take them off, maybe I'll try the freezer to pop them off.
            If it's really thermal epoxy, it should still be good. But if it's the yellow thermal "gunk" that was often used on older video cards from that time... yeah, replace it for sure.

            By the way, is the heatsink held with screws also (in addition to the thermal epoxy/glue/compound)?

            Originally posted by Justin1091 View Post
            There are two holes in the PCB so it's possible to put a good heatsink on it. Anyway that'll have to wait, not done with this bad one yet!
            Well, you can use the bad card as a "test dummy" on which to try your new heatsink. Additionally, use can use the screw holes to "overtighten" the heatsink a bit more, which could, in some cases, become a "temporary" fix for the artifacts.
            Last edited by momaka; 06-07-2018, 07:06 PM.

            Comment


              #46
              Re: 3dfx Voodoo5 glitches

              Originally posted by momaka View Post
              Q1-Q5, U4, U12, and AR1 (358 op-amp) are all power-related components. If any of these are bad, in most cases your video card wouldn't work at all and also probably trip the power supply short-circuit protection, as that's usually what happens when power components go bad.

              As for the other parts...
              U1 likely has something to do with just driving the monitor output. The artifacts you're seeing cannot come from this chip, because they only appear to come in specific scenarios (mainly in 3D mode / games). So that's not it.
              U6... definitely not, as that is probably used for TTL logic between various chips on the board, not 2D/3D data.
              U7... maybe. Depends what IC(s) are connected to it... but I'm doubtful about this being the problem too. Perhaps if this chip has something to do with transferring data between the two GPU chips, then it's possible it may be corrupting some data. Otherwise, probably not.
              U8 - definitely not. This is your BIOS chip. If it was bad, the video card wouldn't boot. Only exception is if the BIOS is corrupted and parameters regarding the RAM timing or GPU/RAM clocks are not right.

              So with that said, there's just very little chance any of these are causing the artifacts. Again, artifacts appear usually when texture/geometry data becomes corrupted in the buffer or GPU. Thus, I think the problem is more than likely either faulty GPU chip or faulty RAM chip(s) corrupting data somewhere. There's also a small chance the ceramic caps on the board may be getting too old and loosing their capacitance (especially at elevated temperatures), but I doubt it.

              By the way, good to hear PCBONEZ is still out there posting stuff. He used to be here and often posted lots of good info.
              Yeah, unfortunately he isn't active right now. I've found out that the artifacts also appeared with the good card and updating the game fixed it. So, the card has zero artifacts.

              Now, chips can run in 3d but not with everything set to high quality and/or GBC. Either a chip is bad then indeed or that U7 one. I can find out with continuity checks where it is connected to. I ran out of time but did take it's voltage : same both cards. Interesting take on U7 btw, I had no idea what could be the purpose of it.

              Comment


                #47
                Re: 3dfx Voodoo5 glitches

                Originally posted by momaka View Post
                I've seen much worse variation on in-circuit readings on same video cards before (albeit newer cards, but still). To compare the two cards truly, they must use the same RAM chips (same manufacturer, type, and speed) and made during similar periods. Reason I say this is because I often see RAM chips read different resistance. Sometimes, even a newer vs. older batch of the same RAM type can have different readings. And things like that tend to make in-circuit readings different.
                Both cards use the same memory chips. The cards are however from a different revision, that explained why there are some differences in resistors used on the cards.

                I usually use my unregulated 35 Watt wall iron for all small SMD components like that, and I've never had a problem. The key is to use a proper-shaped tip and lots of flux. Tip should be either chisel/bevel or spoon -shaped. Also add fresh solder to the ends of the SMD components before trying to remove them. Practice on a junk board, and you will get the hang of it. I sometimes even prefer doing SMD work over regular through-hole - that's how much I'm used to it now.
                Think I'll need another tip yeah, mine is like a pencil. When trying to desolder something with legs, the legs get too warm and melt. Soldering the small smd caps also took quite a bit of practice but now it's "easy".

                Yeah, nVidia and ATI started spreading lies way back that it's fine for their chips to run at high temperature. The truth is, their chips can only withstand those temperatures... but for how long, of course they don't tell you, because it's well-known fact that BGA fails after prolonged use at higher temperatures.
                And that helps in selling more cards.. I'm going to keep mine way cooler than the thermal profile msi put it on now.

                If it's really thermal epoxy, it should still be good. But if it's the yellow thermal "gunk" that was often used on older video cards from that time... yeah, replace it for sure.

                By the way, is the heatsink held with screws also (in addition to the thermal epoxy/glue/compound)?
                I don't know if it's thermal, it is yellow and feels like stone. It's probably the "gunk". I have no idea how warm the chips get.

                Well, you can use the bad card as a "test dummy" on which to try your new heatsink. Additionally, use can use the screw holes to "overtighten" the heatsink a bit more, which could, in some cases, become a "temporary" fix for the artifacts.
                Or the freezes perhaps, too bad I can't do this because the heatsink is attached with the glue. Screw holes aren't used. But I can try this if I buy new heatsinks and use screws.
                Last edited by Justin1091; 06-09-2018, 01:19 PM.

                Comment


                  #48
                  Re: 3dfx Voodoo5 glitches

                  I looked at all you old posts on all the old sites where you asked for help and posted pictures, you replaced a cap with a crap cap that wasn't even close to what it should have been, you screwed the GPU, putting in the proper cap after screwing the GPU up won't fix the GPU. That's why I said before you destroyed the card. If you want to fix it replace the bad GPU or use it as is.
                  Last edited by brethin; 06-09-2018, 04:14 PM.

                  Comment


                    #49
                    Re: 3dfx Voodoo5 glitches

                    Originally posted by brethin View Post
                    I looked at all you old posts on all the old sites where you asked for help and posted pictures, you replaced a cap with a crap cap that wasn't even close to what it should have been, you screwed the GPU, putting in the proper cap after screwing the GPU up won't fix the GPU. That's why I said before you destroyed the card. If you want to fix it replace the bad GPU or use it as is.
                    Look, ~10 years ago I got the card. It didn't work properly so I put it away, a couple of months ago I replaced all of the caps. I did put one that was incorrect on it but corrected it. Nothing changed, it doesn't work better or worse compared to 10 years ago when it had all the original caps on it. This means that having that cap on for 2 minutes power on time (literally, just a quick test to see if it was working) had no effect on this case.

                    Did you even read the posts on this topic? Again, nothing has changed! It still has the same problem! 10 years ago: freeze. Crap cap 2 minutes on it: freeze. Good caps on it: freeze. Freezer trick: freeze. Different PC/PSU: freeze.

                    If it's the same as before replacing the cap, how can one say replacing that particular cap destroyed the card. So the cap replaced last month is causing the problems that occurred 10 years ago (both chips are still functioning the same as before)?

                    By the way, you asked me for photographs about after I recapped the card, I made them for you. Why did you ask about them if you just repeat your previous message?

                    Which old sites are you talking about? What do you mean with 'old'? I have asked years ago on some sites (not sure if they're 'old') but I never replaced anything on the card until last month. Which, by the way, didn't change anything. It works the same as 10 years ago.

                    Comment


                      #50
                      Re: 3dfx Voodoo5 glitches

                      Originally posted by Justin1091 View Post
                      Look, ~10 years ago I got the card. It didn't work properly so I put it away, a couple of months ago I replaced all of the caps. I did put one that was incorrect on it but corrected it. Nothing changed, it doesn't work better or worse compared to 10 years ago when it had all the original caps on it. This means that having that cap on for 2 minutes power on time (literally, just a quick test to see if it was working) had no effect on this case.

                      Did you even read the posts on this topic? Again, nothing has changed! It still has the same problem! 10 years ago: freeze. Crap cap 2 minutes on it: freeze. Good caps on it: freeze. Freezer trick: freeze. Different PC/PSU: freeze.

                      If it's the same as before replacing the cap, how can one say replacing that particular cap destroyed the card. So the cap replaced last month is causing the problems that occurred 10 years ago (both chips are still functioning the same as before)?

                      By the way, you asked me for photographs about after I recapped the card, I made them for you. Why did you ask about them if you just repeat your previous message?

                      Which old sites are you talking about? What do you mean with 'old'? I have asked years ago on some sites (not sure if they're 'old') but I never replaced anything on the card until last month. Which, by the way, didn't change anything. It works the same as 10 years ago.
                      I did put one that was incorrect on it but corrected it.

                      That's where you messed up, you cant put in a wrong cap then put in the right cap later after you have done the damage. That's why I said you killed the card.

                      Comment


                        #51
                        Re: 3dfx Voodoo5 glitches

                        Originally posted by brethin View Post
                        That's where you messed up, you cant put in a wrong cap then put in the right cap later after you have done the damage. That's why I said you killed the card.
                        A few wrong caps won't kill the card like that, because that's not how VRMs work at all - be it linear or buck type.

                        With linear, you typically need a tiny bit of capacitance for the regulator to work. Most cards have that on their linear rails just from the ceramic caps in circuit.

                        Buck-type regulators are more demanding and can cause damage if there is something wrong with the output caps. But when buck regulators go bad, you're not going to get artifacts like that. You will get majorly fried chips (i.e. not just partial loss of functionality but full loss of the chip), possibly the upper/lower MOSFETs for that buck circuit, and maybe even the coil too, if it's on the wimpy side. How I know? I had to repair hardware like that before. The eVGA GeForce 7600 GT video cards with their Sacon FZ caps were notorious for this, as I found out the hard way.

                        So I seriously doubt that whatever bad recap job the O/P may have done caused this.

                        Comment

                        Working...
                        X