buildzoid Posted June 19, 2015 Posted June 19, 2015 I'm trying to run my E-powered GTX 590 and the card doesn't want to start. To make things worse the POST code on my RIVE doesn't works only the debug LEDs do. So all I know is that I have a GPU problem of some kind. When I was soldering on the E-power I managed to drip solder on the PCI-e pins however I've tried to fix them and as far as my DMM is concerned all of them are fixed. Now they may not be making good contact but is it possible that I need to provide a PWR_GOOD signal when using the E-power? I ripped of both the VR controllers so I thought that I don't need to deal with any PWR-GOOD signals. Is that correct? Quote
Strong Island Posted June 19, 2015 Posted June 19, 2015 I'm trying to run my E-powered GTX 590 and the card doesn't want to start. To make things worse the POST code on my RIVE doesn't works only the debug LEDs do. So all I know is that I have a GPU problem of some kind. When I was soldering on the E-power I managed to drip solder on the PCI-e pins however I've tried to fix them and as far as my DMM is concerned all of them are fixed. Now they may not be making good contact but is it possible that I need to provide a PWR_GOOD signal when using the E-power? I ripped of both the VR controllers so I thought that I don't need to deal with any PWR-GOOD signals. Is that correct? Ya you shouldnt have to deal with the card checking for the stock vrm since you removed the controllers. You still have to plug in the gpu's power connectors for the auxillary controllers. Quote
buildzoid Posted June 19, 2015 Author Posted June 19, 2015 Ya you shouldnt have to deal with the card checking for the stock vrm since you removed the controllers. You still have to plug in the gpu's power connectors for the auxillary controllers. I know I have both 8 pins plugged in. Quote
der8auer Posted June 19, 2015 Posted June 19, 2015 Did you check if all voltages are present? Also NF200 voltage? Quote
buildzoid Posted June 21, 2015 Author Posted June 21, 2015 Did you check if all voltages are present? Also NF200 voltage? What's NF200? Quote
Guest TheMadDutchDude Posted June 22, 2015 Posted June 22, 2015 It's the bridge between the two cores, I believe. It needs power or the card will not boot. Quote
buildzoid Posted June 22, 2015 Author Posted June 22, 2015 It's the bridge between the two cores, I believe. It needs power or the card will not boot. Great I'll try again tomorrow. I'll add some wires for voltage reading to the card. Though I really wonder what the hell I will do if all the voltages are there and the card doesn't work. That would mean I'm screwed right? Quote
Guest TheMadDutchDude Posted June 22, 2015 Posted June 22, 2015 If you ask me, yes. However, I am nowhere near qualified to say what it will or won't do. I'm sure someone else like Roman can lend a hand though. Quote
K404 Posted June 24, 2015 Posted June 24, 2015 I'll add some wires for voltage reading to the card. Ummmm....you're zombie modding WITHOUT them? Quote
buildzoid Posted June 26, 2015 Author Posted June 26, 2015 Ummmm....you're zombie modding WITHOUT them? 1. This is my first zombie mod. 2. I figured that I would just read everything of the cap banks however now I realize that that was a stupid stupid idea. Quote
buildzoid Posted June 30, 2015 Author Posted June 30, 2015 Ok so I finally got around to running the card. The bridge chips is good but VMEM for the GPU next to the output connectors is 0.3V. Quote
steponz Posted June 30, 2015 Posted June 30, 2015 Start tracing what is messed up.. Any pics? is mem good on the other side? Quote
buildzoid Posted June 30, 2015 Author Posted June 30, 2015 Start tracing what is messed up.. Any pics? is mem good on the other side? Yep mem for the other core is fine running at 1.6V. I did no soldering in the vicinity of the VRAM controller or the VRAM VRM. Everything looks OK so I have no idea what could have broken it. Assuming that the still working VRAM VRM isn't utter garbage I'm thinking of just pulling 1or 2 AWG12 wires from VMEM2 to VMEM1. Quote
buildzoid Posted July 3, 2015 Author Posted July 3, 2015 Ok so I checked the VRAM2 VRM and it looks like it should handle the extra load that running the other VRAM chips will add. Is there any other potential issues that I should be worried about? Quote
buildzoid Posted July 6, 2015 Author Posted July 6, 2015 Okay so I tried the VRAM VRM bypass. Both sets of VRAM now get 1.55V. The card however still doesn't work. So it looks like I need to try fix the PCI-e gold fingers because those are the only things that went wrong when I was attaching the E-power and so the card is probably not getting PCI-e data properly. I'm going to buy a 1X to 16X PCI-e extension cable and see if that gets the card to start up because all of the PCI-e X1 fingers are fine the problem is with X8 and X16. Quote
steponz Posted July 6, 2015 Posted July 6, 2015 My guess from looking at your blog it could be alot of different things... did you check any resistance before starting to make sure there isn't any shorts? Did you pretest the card and test voltages at all? Are you using only 1 epower for gpu? I know when Tin did this, he used 2.. You should be using 2. Was there 2 separate controllers for gpus? Quote
buildzoid Posted July 6, 2015 Author Posted July 6, 2015 (edited) My guess from looking at your blog it could be alot of different things... did you check any resistance before starting to make sure there isn't any shorts? Did you pretest the card and test voltages at all? Are you using only 1 epower for gpu? I know when Tin did this, he used 2.. You should be using 2. Was there 2 separate controllers for gpus? I have tested the card before the e-power but with the CLCs. It worked just fine but didn't clock very well. I checked resistance vcc to gnd it's fine. IDK how powerful the old e-powers were but the current e-power is ok upto 900a according to Roman and I only need 600a. As for other shorts I've checked visually if anything looked luke it might be shorted I checked with DMM no shorts anywhere on the card. I also didn't break any smd resistor/caps The card has 5 vrms and 5 controllers. VRMs were vcore gpu1, vcore gpu 2, vmem gpu1, vmem gpu2, pcie bridge chip vrm. Edited July 6, 2015 by buildzoid Quote
steponz Posted July 6, 2015 Posted July 6, 2015 You obviously did something wrong... so you need to trace and see what it is. Here's the reason why I would use 2... each gpu will require more or less voltage. So it would be better to install 2..to control each. You have not listed any detail steps.. so it's very hard to help you. You haven't said any voltages per gpu or memory what they were stock. You are hitting a fault on mem for a reason. So you need to trace back. Did you remove the controllers.. please explain more so that you can be helped. When I do epowers I test all voltages and 1 step at a time remove and then test again. This is why I can make a card that exceeds the factory performance or at least is the same performance. Quote
buildzoid Posted July 7, 2015 Author Posted July 7, 2015 (edited) You obviously did something wrong... so you need to trace and see what it is. Here's the reason why I would use 2... each gpu will require more or less voltage. So it would be better to install 2..to control each. You have not listed any detail steps.. so it's very hard to help you. You haven't said any voltages per gpu or memory what they were stock. You are hitting a fault on mem for a reason. So you need to trace back. Did you remove the controllers.. please explain more so that you can be helped. When I do epowers I test all voltages and 1 step at a time remove and then test again. This is why I can make a card that exceeds the factory performance or at least is the same performance. I know I did something wrong but so far the only visible issue is the messed up PCI-e slot connector. As for using 2 E-powers. They are sold out ATM and even if they weren't I'm broke. History of the card: Pulled of stock cooler and replaced it with CLCs. Card worked and ran Heaven at stock clocks with CLCs core temps stayed in the 50s Pulled of CLCs and removed vcore gpu1 and vcore gpu2 inductors and controllers Soldered on E-power in process spilled solder on PCI-e slot connector connector was shorted Disconnected 2 PCI-e data pins when trying to remove solder from them Reconnected PCI-e pins Checked resistance from where the E-power connects to the card and GND. 0.5 ohms which is OK because the 2 GF110 cores are connected to the E-power in parallel. Set E-power to 1.2V Reattached CLCs and put small aluminum HS on bridge chip. Tried to run card. Card didn't start. Reconnected PCI-e pins fell of. Reattached the PCI-e pins a second time with glue and solder. Checked pins with DMM for shorts no shorts found. 1 other PCI-e pin comes of during soldering. Checked it with a DMM it's a GND pin and I didn't bother fixing it. Rattached CLCs and tried to run card. Card didn't work. Attached voltage read wires to card. Tried to run card found that Vmem GPU1 is at 0.3V not 1.55V like it should be. Cleaned up soldering job on reattached PCI-e pins to try help improve contact in PCI-e slot. Bypassed Vmem GPU1 VRM with 3 18AWG wires from Vmem GPU2 VRM after removing inductors of VMEM GPU1. Both sets of memory are now supplied 1.55V(checked with DMM) But card still doesn't boot. POST code on MB is still not working. Now I'm thinking of trying to pull of the VMEM GPU1 controller since it might have a PWR_GOOD pin and I also wanna try connect the card to a 1X PCI-e connection because then I avoid the damaged data pins. Edited July 7, 2015 by buildzoid Quote
TiN Posted July 13, 2015 Posted July 13, 2015 What is CLC? Try just pullup for memory controller ENABLE pin (10k to 3.3v). Sometimes there is no pullup on card on powergood from GPU PWM IC, in this case you would need to add it. It does not matter if you provide memory voltage, if GPU RESET is kept inactive due to power good wrong. Quote
buildzoid Posted July 13, 2015 Author Posted July 13, 2015 What is CLC? Try just pullup for memory controller ENABLE pin (10k to 3.3v). Sometimes there is no pullup on card on powergood from GPU PWM IC, in this case you would need to add it. It does not matter if you provide memory voltage, if GPU RESET is kept inactive due to power good wrong. CLC = closed loop liquid cooler. The GTX 590 doesn't support proper GPU water blocks so I ziptied some Coolermaster Seidon 120Vs to it. Core temps were good with this cooling setup when the card still worked. So I should find the pull up pin for the CHiL CHL8266 that controls Vcore or should I pull up the Vmem controller which I can't find any datasheets for? Quote
TiN Posted July 15, 2015 Posted July 15, 2015 it does not matter , usually it's same signal they connected to. I don't remember exact details, did GTX590 zombie long time ago, but pretty sure that's the case. Quote
buildzoid Posted August 1, 2015 Author Posted August 1, 2015 Ok so I've hunted down the data sheets for both of them. There's a 1.5V Power OK signal on the VRAM PWM(APW7088) but the datasheet for CHL8266 is in typical IR fashion borderline useless so I have no idea what the power good pin for it is and I have doubts about putting 3.3V on the Power OK pin of the VRAM PWM. Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.