Jump to content
HWBOT Community Forums

nachtfalke

Members
  • Posts

    859
  • Joined

  • Days Won

    8

Everything posted by nachtfalke

  1. ocp is simply just lowering in half the ressistance between ILIM and gnd! , but like i said OCP mod will leave only 50-80mv before OVP will trigger , i know these on semi pwm like my pants! , only way is VID mod ... i bench 460 soc 9 moths ago ...what i can said is i won´t bench that card anymore nachtfalke`s Unigine Heaven - Xtreme Preset score: 1378.66 DX11 Marks with a GeForce GTX 460 (256bit)
  2. thanks! ofcourse:D because she used ncp 5395 pwm , but the main problem isn´t OCP , it will be OVP! my card use hynix , they scale with volts , for 1100mhz , need atleast 1.68 -1.7v...it will be a problem! for the gpu i configure the vid mod for ADP4100 ...that should off de OVP , also think that will need hard mod for CBB aswell!!
  3. Does anyone have a cold slow bios for gtx470 SOC and gtx570 SOC?
  4. here the mods, be carfully when you will turn the trim pot for vdd and vddq ...they are very sensitive and rise the voltage spontanly, so rotated very very slowly! on gpu , it hint NO ocp or ovp !! enjoy!
  5. gpu-z it had a issue reading the gpu rev on G92 , read at all A2!! you should check on the chip die what rev it is !! DK is a custom implementation so that was on market after reference , it use for sure B1 rev chip! need to look after some reference card from first batch with reference cooler design! , from evga , zotac etc.. but if the card is A2 with Samsung ram´s it´s for nothing , should be A2 with hynix!!! that are very hard to catch one!
  6. you can voltmode gtx285 classified simply like other card aswell ,i did that few years ago! or software with comand line from rivatunner , because it use volterra VT1165mf pwm chip , it is a true digital pwm ...
  7. 9800gtx+ CS bios didn´t work:D tested on two reference card!! still shut down at -45 /-48. that are nothing to do with witch card implementation are , i tested over time more than 14 -16 G92 cards (B1 /55nm) doesnt matter if card is reference or custom still had this issue , epowered or not issue will be the same , think that are to do with thermaly protection from chip itself , because all card tested shut down at the same degree value ( +/- 3 degree). G92/ 65nm love cold better than 55nm , but had a strange cold slow , if you get the card that you can mange the CS can be benched at 1150mhz+ on the core! you should search 9800GTX (without+) A2 rev chip!! 65nm
  8. i sugest you to make VID mod with 2 polig dip-switch , it is very useful besides you will don´t have problem with OVP trigger anymore! so you will play from switches with these "default" voltages from the VID and with iRog chip soft and nice up the voltages from touches
  9. remove 3.3v from pin62(psi) and try again. also make strap on the right down corner( back side pcb) for disable OCP and pwm freq
  10. pin 51 IT IS conected from pwm , but his trace are underneath corespondent on the pcb back side , if you don´t find it i make for you a pic without EN pin pulled hight ( highter than 1.1v) you can´t enable the pwm:D pin 52 are VTT witch i foget to say today morning( europe) should be pulled hight(highter than 0.85v enable the controller ) aswell to enable pwm! cut the card like i did , is a bit dificult but you can for sure, i help you to make the gDDR vrm functionally aswell , when you are ready i give you advices for gDDR remember use 3.3v LDO that can generate at least 1.5 A , on the card are many led´s , specially on the iRog controller , i recomanded to use LM317 or LM1085
  11. sure:) pin 48 = +5VCC ( add +5vcc LDO to generate power supply for pwm) pin 51 = EN -enable chip ( detect whether MOSFET driver power supply is ready) pin 52 = VTT - enable input of the controller(this pin monitoring if any 12v voltage are presence in the pci-e slot to activated the chip) pin 62 = PSI ( power state indicator input for load ) here should be always 3.3v to enable ALL power phase , if 3.3v missing will be activated ONLY one phase. pin 46= SDA ( data for I2C interface) pin 47 = SCL ( clock for I2C interface) also in the right pcb corner you had some measured points , be sure were 3.3V ( this voltage was directly from pci-e slot ) point are should be alwasys 3.3v presence to enable iROG controller ...enjoy!!
  12. primarion PX3540 it has same configuration pin out like PX3544 , so p_good/vr_ready is same pin 15. VDD(3.3v) is pin36 GND is pin 25 vr_ready pin15 OCP per phase: phase4 pin 28( NC on px3540 and 8800GTS 640/320mb) phase3 pin 31 phase2 pin 34 phase1 pin 38 PX3544 can be used in DC-DC converters up to 130A and 1 to 4 phase pwm PX3540 can be used in DC/DC converters up to 200A up to six phase digital multiphase controller. also PX3544 it has more features like oldest PX3540 -Multiple Internal and External Temperature Limits -Configurable latched fault or autonomous recovery shutdown -Single +3.3V supply operation
  13. VID mode , it´s only way to pass over voltage protection (OVP) witch on many controllers it´s a really big problem , FB mod offer you controll at mv , both combined are best way. even controll via I2C can´t go pass over OVP ...so for me best way is VID mod , ofcourse that is not so simple like FB mod . vsense pins from pwm is to callibrate the LLC´s for GPU , they read voltages directly from gpu , because many card implementation even they had a strong VRM the powerplane from VRM output to gpu it has a voltage droop ( in pcb powerplane)in gpu load , vsense +/- will regulate this droop to be more accurate ! , actually the voltage that you set in software or that you read with DMM from output VRM are not the real voltage that the gpu take! , exemple MSI gtx570 TF from output vrm to back gpu MLCC´s cap are almost 100mv droop! , that means the powerplane on this card are very weak ..even that is one custom card... i made 4 x 8800gtx zombie VRM all working( ram vrm aswell) on my cards , for example on my gts 250 : nachtfalke`s 3DMark03 score: 71597 marks with a GeForce GTS 250 the 580 matrix zombie is one of best VRM , very good calibrate and very powerfull VRM peak 500A , also a very good option to replace epower or any other custom VRM from any vendors!!! , you can made that itself..
  14. today i made a very powerfull Powercard from a dead gtx 580 MATRIX! - add +5.0 vcc to pwm supply - add +3.3v to enable pwm ,chip and power state indicator input for load! - add +3.3v to power iROG microcontroller with I2C/SMBus that will allow you to increse vgpu on the fly +250mv from the touch ( +/-) ( depends what firmware are written ) - add +3.3v to power leds from platform stand-by next days i will made directly VID mod to 1.4v
  15. you soldered the wire directly to the pwm VTT ( enable INPUT for pwm)pin? because shorted to any VID pins is only way( think) , besides VTT resistor are near PGOOD resistor witch is pulled up also with 3.3v ..so near that are no VID resistor( or vid traces) to make any short! VTT pin monitors 12v presence from pci-e slot , and "ask" if any 12v presence are into pci-e slot , if is than pwm are enabled , if not are send to sleep! , BUT to enable the hole pwm you need find the second pin responsable to chip enable! you are sure that you dont foget another pin to enable the pwm:D ?
  16. yes , tried solid copper strips on my others power card( epower rev1 and 2 , apower aswell) but isn´t very good for pcb layer , especially when need to be removed they are very rigid and inflexible can pell off the pcb copper layer , also need a powerfull soldering iron , that used much amount of heat thru pcb layer can also pell off in the future and ruined the powercard! when powercard is mounted permanently on the target vga ...copper strips is best way! speaking about this homemade powercard , i don´t have very much joint points on the front side +VCC( front side they are 7 , and back side from +VCC on the solid caps 8 joints) , if you look used +VCC from pos capacitors and MLCC´s , joint all together with soldering ball is a very good/powerfull +VCC points . in total they are 15 joints for +VCC ( excluse joints directly from the chokes) if i solder from all joints wires 2.5mm they are 37.5mm² ( AWG1.5) i don´t find any issue using wire(1.5mm²(awg15) and 2.5mm²( awg13)) if they are very good soldered at the ends using enought amount of flux + solder, ofcourse manny wires are always better(added depending what gpu you wanted to be powered... in my first case are a G92 don´t need very munch wires ) GND joint are directly for the backplate holes witch are very good , i use always for GND´s 6 (awg9)and 8 mm² (awg8)wires. you mean +Vsense and -Vsense? add to the load near gpu? to regulate LLC´s? i made directly VID mode , because the controller is full of protection ( ocp , ovp and uvp) so used 3 polig Dip switch can be output setting to +100mv (VID4), +200mv(VID3) , +400mv(VID2) , using all switches ON add +600mv! as a tip for others , you should pay attention to the power state indicator input for LOAD , because active phase will be only one when load goes low ~20 -25A , and the rest of phases will be automaticly desactivated , so that pin should be pulled up to activated all phases!
  17. hi all , today i made dual output powercard from a dead gtx580 DC2 , but with functional VRM. on GPU VRM we have 8 power phase and 32 mosfet transistors NXP 7030AL ( N-channel TrenchMOS logic level FET) witch can go up to 50A at 80-100 degree , pwm used are ASP0907( remarked) , output filter has 10x 820uF solid cap rated 3V , 4x470uF POS cap and 20x MLLC DDR VRM we have 2 phase and use as pwm uPI6205A and 8x NXP 2530AL up to 25A , output filter 4 x820uF solid cap and on back side the yellow tantalium cap! - cutt the PCB and make sure that are NO short in pcb layer!! - add +5v ( home made DC-DC converter with L7805 - TO220 package) - add +3.3v ( i use a AMS1117 DC-DC converter in SOT package) - add +12V to DDR rail - gpu vid mod to 1.4v by default( fluke DMM) - gpu ocp overall - gpu per phase ocp mod. - ddr ocp mod need only add a trim pot to FB pin on both pwm ( gpu /ddr) and is ready:D
  18. if all exemplars had same behavior and that are not solved with bios from nvidia , seems to be a hardware problem not at software level , 1.25volt limit is from chip itself , technological chip process and nvidia can limit the core voltage at 1.25v and that is not good at all for OC community
  19. you need configure VID mod to go hight on core voltages , it´s only way to pass over voltage protection (OVP) witch will trigger above 1.3-1.35vgpu with NCP´s buck controllers!
  20. the mod from gtx260/280 " about OCP " has nothing to do with OCP;) , as we read the datasheet has to do with R_Ref( pin30 reference resistor use to calibrate) and R_Sel ( pin31- the range select pin) the pins that has to do with UVP and OCP/OVP is pin 24 and pin 25 , you should lowered the resistance R2003 from 10k to 8K or 5K( but i don´t recomanded if you don´t change the coil inductors and cooled very well the mosfets) that has to do with OVP , and R2001 for UVP!
  21. Items location are germany Gesendet von meinem SM-G930F mit Tapatalk
  22. add massive full cover waterblock compatible with asus GTX580 MATRIX in the same price! -tested -no leaks!
  23. look at the GT2, GT3 ,GT4 that are a bugged score so need to be removed!
×
×
  • Create New...