Vgpu mod for 5900XT:
http://www.akiba-pc.com/e107_images/...gpu_5900xt.jpg
Connect that pin to GND via VR. 3.76kOhm gives 1.6V on my card.
Printable View
Vgpu mod for 5900XT:
http://www.akiba-pc.com/e107_images/...gpu_5900xt.jpg
Connect that pin to GND via VR. 3.76kOhm gives 1.6V on my card.
I tried the Vmem mod too from 2.86V to 3.3V and it didn't help anything at all btw (its the second ISL pin 5 to GND).
Hmmmm I havent tried that yet....but thats mainly because my mem only runs at 770 but my core is good for 500mhz stock vgpu n' stock cooler.....Macci I was wondering If you couldnt post a pic of the vmem mod? I would greatly appreciate it...I think that you problem is caused by the mem speed on the 5950 cards, because the run at 850mhz dont they ? have you tried clocking the mem down ?
5950s run at 950 mem IIRC the 5900ultra runs 850. Ive tried both of those BIOSes and neither of em works. I just get some 'device cannot find resources' error in winxp control panel and the desktop screen is useing 4bit colors :D Flash back to albatron 5900XT and it works normally.
Rivatuner wont see the card at all so no way to lower the clockspeeds.
I've read that people w/ LEadtek 5900XT (it has 2.8ns hynix just like this card) have succesfully flashed to 5950 and are able to run the RAM at way above 900MHz.
My card has 2.8NS sammy mem...maybe the leadtek cards have more relaxed timings ? btw could you be so kind and post a pic of the mem vmod ?
Man, I was almost certain that there's something foooking wrong with that... got the mod too, but almost no improvement at all....Quote:
Originally posted by macci
I tried the Vmem mod too from 2.86V to 3.3V and it didn't help anything at all btw (its the second ISL pin 5 to GND).
But then again, where do you measure the Vdimm? Cuz the closest I can get with that Gigabyte is around 2.53V... with the mod I'm getting it up to 2.81V.... and that helps for around 20MHz or so... :rolleyes:
Wanna give me some info on that mod ? :P
Btw I have completely forget how to measure the capacitators...should I measure on both legs or should I measure the + pin and ground ?
I have 2 Point of View cards, one with a red pcb does 530/777 on stock cooling (it has samsung 2.8ns memory) and the other one has a black pcb and it does 530/945 stock (it has 2.2ns Hynix memory).
I also have the one with a red pcb...which unfortunately seems to be sucky :P
That's the "lazy man's vmod". Over on the ocforums, IIRC, someone tested the voltage before and after BIOS flashes.Quote:
I've read that people w/ LEadtek 5900XT (it has 2.8ns hynix just like this card) have succesfully flashed to 5950 and are able to run the RAM at way above 900MHz.
Apparently, the 5950 BIOS does provide about .2v more to core and mem. (Harder to test, but evident by initial lower scores at the same clock and higher possible clocks, is that the 5950 BIOS *also* reduces timings on the memory. Both things that lead to much-higher-than-anticipated memory clocks.)
The thing about these cards is that the voltage is 'software' controlled. For 'safe', '2d', and '3d' modes, it offers 3 different voltage levels - and, as noted - a BIOS flash can change them.
It seems pretty obvious someone could write a program to access that directly from Windows, if they were so inclined.
Pin 7 on that chip is ground, right?Quote:
Connect that pin to GND via VR. 3.76kOhm gives 1.6V on my card.
What resistance did you use for this? And what voltage is 'stock'?Quote:
I tried the Vmem mod too from 2.86V to 3.3V and it didn't help anything at all btw (its the second ISL pin 5 to GND).
These are the pins I tried to use! According to this document, for the ISL chip, pin 5 is feedback (as you are proving) and pin 7 is ground.
Yet, stock (unmodded), I measured .73 kOhm resistance between these two points! Shouldn't the FB pin to ground pin be 0.0 resistance?
Have you tried editing the BIOS's at all to change vendor ID's, or other such info macci? Hardclocking the card back down to XT speeds might fix something.
I was thinking....doesnt the 5950 have 256mb of mem ? and the FX5900Xt only have 128mb couldnt this be the problem ? because when ppl with the 9800pros flash to an 9800XT bios they get alot of problems because of that ?
No, the BIOS on the FX cards doesn't control how much memory the card is looking for. It's almost PC-like: on bootup, it detects how much ram is 'installed' and uses that for the card.Quote:
doesnt the 5950 have 256mb of mem ? and the FX5900Xt only have 128mb couldnt this be the problem
Oh allright....btw when measuring the Vgpu on the card...should i only hook one of the wires on my multimeter up to the capacitator and one to ground ?
I tried bios editing but the only tool Ive got (NV BIOS editor) cannot mod the clocks speeds on FX5900 series. Anyone got links to better bios editing tools/modded FX5900 bios files?
XanderF, Yes pin 7 is GND there. My card was at 2.86V stock for Vmem. I used a 15kOhm VR for starters and lowered the value until it was at 3.3V Vmem - which didn't help the RAM oc at all.
The original resistance from pin 5 to GND is around 0.7kOhm and once we solder a VR in parallel w/ that the resistance gets lower and voltage goes up.
Man, you didn't say anything about the Vdimm reading point?? :confused:Quote:
Originally posted by macci
My card was at 2.86V stock for Vmem.
Basically any SMD cap near the memchips
Didn't think of it that way... hmmmmm...
Just for the record - those two are showing some voltage change when I'm lowerin' the resistance on the second ISL chip.... can you check them too? Just for an opinion....
I'll check the SMDs right away....
Damn, I forgot the pic:
Oh allright....btw when measuring the Vgpu on the card...should i only hook one of the wires on my multimeter up to the capacitator and one to ground ?
Yes... one of the to ground (black cable on the MOLEX conectors) and the other in the desired spot...
Alright thx....could you post a pic of the vmem mod ?
Right, so I was doing the right thing.Quote:
XanderF, Yes pin 7 is GND there. My card was at 2.86V stock for Vmem. I used a 15kOhm VR for starters and lowered the value until it was at 3.3V Vmem - which didn't help the RAM oc at all.
The original resistance from pin 5 to GND is around 0.7kOhm and once we solder a VR in parallel w/ that the resistance gets lower and voltage goes up.
Probably my resistor was too high, though. I'm use to ATI cards that seem to take 13kOhm and up, so I 'made' a 20kOhm pots by soldering together a 10kOhm pots and a 10kOhm resistor.
Downside, of course, is that I can't do less than 10kOhm that way.
What resistance are you using on it now? If less than 10, I'll just use the 10kOhm pots on its own for both vMods. Easy enough to do.