Gigabyte 5870 SOC

[massman]

raised it to 1.25, but no difference from stock.

[[XC] gomeler]

Got my retail sample today. Anyone have an ES card and willing to take pictures of the PCB and upload the BIOS somewhere? Want to figure out what's keeping the OC Tuner from feeding more than +.1v.


Testing the card now. Passes 3D06 SM3 at 975/1350 with stock volts. Not too pleased at the moment. We'll see what it does with +.1v.

[[XC] gomeler]

Got some dipswitches so I can do the VID mod. Looking over the VID mod I see the low table has a duplicate high table right above it. The resistors are 10kOhm and to set low to 1 you have a 10kOhm resistor between that VID pin and the gpu vddc. Does anyone know what the high table pins are being pulled to through that 10kOhm resistor? I'm about to solder two 8 pin dipswitches and I plan on only soldering to the ADP4100 pins and then a trace to vddc and whatever the high logic requires. With a 10kOhm resistor between the two logic voltages I should be able to adjust VID with ease and only solder to the card 18 times. Lastly, do I need to do the OCP mod for 1400MHz gpu, if the GPU is capable of it?

Cliffnotes: Anyone know what the logic high pins are pulled to? Is the OCP mod necessary?

[BenchZowner]

Voltage < = 0.3V --> logic low ( 0 )
Voltage = > 0.8V --> logic high ( 1 )

I'd leave OCP, just VID mod the card and test.
If you get shutdowns at a high frequency regardless of the temp & voltage applied ( if you don't have an AC/DC clamp A-meter ) then do the OCP mod.

[[XC] gomeler]

zomg. Just actually read the ADP4100 datasheet The pins are pulled to GND so I just need to pull the various logic high pins to vddc through the 10kOhm resistor. Awesome, now I just have to solder 9 things.

Thanks for the response. I'll check for OCP when/if it crashes

edit: I wonder if pulling pin 47, PWRGD, to 3.3v or whatever it is normally set to would prevent OCP crashes instead of tricking the controller?

double edit: Hrm. Says only if the output voltage it outside of the safe range. So, could pulling the PWRGD pin high permit the use of the FB mod and enable voltages outside of the OVP limit?

[miahallen]

Says only if the output voltage it outside of the safe range

VID mode overrides OVP ( Over-Voltage Protection ).

OCP needs to be done regardless of how you mod your card ( if you get it to kick in, don't know if it's necessary )

Sure you're not looking at OVP Chris...

[[XC] gomeler]

From my understanding PWRGD is a signal used to signify the state of the PWM. If the fets get too hot, the output voltage exceeds the limitation, or the current exceeds the limitation then the PWM controller trips the PWRGD signal to the GPU and the GPU then tells the controller to power down. I know on another card if you kept pwrgd pulled high then the card would run if you tripped OCP/OVP.

VID modding the controller manually changes the VID for the controller. Since this controller supports on-the-fly VID adjustments it'll change its output voltage to whatever the table says. The card should still trip if you then use a FB mod past the OVP limit. But perhaps if you pull that pin to the "good" state then you can do whatever you want and it won't shut down.

I may give it a shot later. Need to wrap up an essay and then I'm doing the VID mod and benching. Once I've got some results I may then play with the controller to see if PWRGD does what I think it does. Note: I haven't looked at the controller block diagram. OVP could trigger something internally to shut down, regardless of PWRGD state.

[Raja@ASUS]

The trouble with playing around with FBRTN is you will fundamentally change the transient response of the circuit at HF because error amp gain is set by FB RTN (the frequency dependant gain and phase margin is important for the stability of VCC especially with the sensitivity of modern die architectures to absolute VCC at negative temperature thresholds). What you really need is an external PS, although those have implications of their own..

[[XC] gomeler]

Hope I didn't just kill my card. Think I got some flux conducting voltages and messing up the pins and I'm seeing +0.2v on the core. Giving the end of the card an alcohol bath for a few hours and then going to dry it out and pray for the best.

Give it a shot tomorrow, hopefully I can get the card to work. Knew this mod was going to be a nightmare.. soldering is not my forte.

[[XC] gomeler]

Was able to resuscitate the card this morning. VID modded it to 1.40v and it worked for a little over 2 hours before dying again. Was testing 3D03, 1275/1350 with 1.45vgpu, 1.22vtt, 1.64vdimm at -70C. Fun card bringing it back to life but I think it is now dead PWM is doing the same +0.2v thing. I think I'll hook up the core to an external powersupply in a few weeks.

[miahallen]

Sorry to hear about that Chris GL getting it running again

[Jor3lBR]

First Ln2 test today this card has potential:



I only did the VID 1.40vgpu mod and it will probably do more on AQ3 and Vantage...

...............

[Hok]

the 5870 SOC Tuner beta download link doesn't work. Can anyone forward me a working link ? TIA

[wanky]

How would I do the FB mod if i wanted to LOWER the GPU voltage?