@ Niccolò & Clint: Can you say something about the difference in power consumption ?
Printable View
@ Niccolò & Clint: Can you say something about the difference in power consumption ?
If I shade that part, is there any way to know what the change in voltage is without a voltmeter?
Are there any applications that can tell me?
Thanks
Mayk-Freak:
Can you post your BIOS settings ?
hej everyone.
sucsessfully vmodded my card, have it running @ 1039 MHz and my temps are at 55C, just wondering how far i can go. my volt-meter reads 1.710 during load. thats quite hi, ain't it? what is the max you think...
nobody really knows, maybe you can be the first!!! tell us when it dies and we will set ours .001 v below that :rofl: :ROTF: . seriously though I can do 999 with 1.62 maybe you should lower it some and see if it can still hit the 1039 as before. and what are you using to test?
I would buy one multimeter ifI were you, a cheap insurance;)
But to answer your question;
I have had a really hard time to shade it down to 1.0 kOhm with a 4B pencil that is very soft, and that has given me 1.415 volts 3D.
I now bench @ 958/1250 no problems, this core scales like a dream :up:
So if you go nuts on the shading I doubt you will get more than 1.45 volts out of it anyway, and that is pretty safe even on the stock coooler.
I can also confirm that the voltmod works. 3870 VCore now at 1.42 after using a #2 pencil. :up:
http://img249.imageshack.us/img249/6679/dscn2700xp3.jpg
This i guess ;)
no, that's for 3850 card, I want for 3870 :)
btw I can run Nature at 1428mhz on rams without any mod, stockcooled :D
but core only 900 :(
Here's my pencil mod results on my HIS 3850 512 MB IceQ3 Turbo.
__ohm Volt MaxMHz Temp
2.26K 1.31V 826 MHz 48
1.38K 1.36V 850 MHz 51
1.45K 1.39V 876 MHz 53
1.13K 1.44V 904 MHz 56
Penciling in 1.13Kohm was not easy.
Scales pretty linearly across the board. Stable in 3DMark06 and AtiTool artifact tester.
I also slapped on a different fan onto the original heatsink, an 80 CFM Delta 80mm @ 70%. Brought down the temps by about 7 degrees on average.
I've A GeCube 3850 512MB Turbo 3 Card with an AC S1 and dual 92mm fans
the ressistor is 2.44 kohm on my card default and i measured a vcore of 1.26V
this is a pic of my card
http://img155.imageshack.us/img155/6...0markedfc4.jpg
green marked are my measured vcore points
blue is the ressistor i penciled
if i penciled it down to arround 2.00 kohm or 1.8 kohm my measured vcore is lower than default
(around 2kohm it was 1.18volt)
do i measure the vcore on wrong points?
it seams that this is the wrong resistor but my layout is a little bit different what resistor i've to use?
what to do with this card help...
i get 780Mhz gpu clock my card need more vcore :(
http://img149.imageshack.us/img149/5293/pencil1ay2.jpg
i penciled the black marked resistor
it was 2.43 default
i penciled it to 2.22 what gave me 1.40 volt
when i pencil it to 2.19 it give me 1.44 volt
it seems a verry good ressistor for a huge voltage range
i get with my gecube 3850 turbo 3 850Mhz@1.40 volt
but i cant get 864 stable...is it the bios bug?
try overclocking the mem and see what happens
so the resistor to shade on HD3870 is the first below:
R1222
R685
R684
that have a black point in the middle?
thanks in advance for the info :)
even at default memclock (900mhz) i get artifacts with ati tool @864Mhz
when i set the GPU to 850Mhz (there is nothing possible between 850 and 864) i get no artifacts even if the mem is clocked at 1060
i mean 850 is not bad for a 3850 but default 3870 can reach it easiely but not more because of the bios bug
I cut the tip of the pencil flat, large enough to cover the resistor.
Then put pressure and dragged it across.
llll
llll --->
llll
I could put a bit of pressure since it was a soft pencil and it simply crushed if it was too hard. I tried it loads of times before i got it to 0.42
If it dose not work on the first 2-3 passes wipe it and try again.
I tried the circular 8 method but that tend to rub it off if you put just too much or too little pressure
(I managed to get 0.38 > 1.45v but was not getting stable expected clocks most likely due to uneven spread)
hope it helps