PDA

View Full Version : GF FX 5800 Ultra vmod...(incomplete)



gokickrocks
04-26-2003, 12:07 PM
a few months ago i had a 5800 ultra with the intentions on vmodding...the card died on me ways back, not due to vmods however as i didnt have it on at the time...

i was going to publish a vmod guide for the 5800, but for the life of me cant find the diagrams that i made (most likely i threw it away since i drew it on some scratch paper)...


edit:...skip to the 3rd post...

berkut
04-26-2003, 02:02 PM
When i saw this post i almos had an or... :D

2 days ago ive gought a fx ;] And i hoped for a full step by step walk through but unforunetely :(

Ive already found few resistors wich should be "hacked" but ive dont have the guts and dont know what resistors should i use :/

Guess ill have to wait...

Btw. is the NV30 powered @ 1.25V ? Actualy it has a 2 phase power suplyy near the molex connector and i belive its for the vpu ;]

The second mosfet near the dsub connector (1cm away from that square chip) has a 3.3V output and i dont have any idea what it could be for :confused:

gokickrocks
04-27-2003, 12:25 PM
i browsed the forums and found a pic of the 5800 and found the regulators that need to be vmodded, but i dont have access to a 5800 to measure voltages...


the vddr is an isl 6225ca chip (circled in yellow)...need someone to measure the voltage on pin16...

14 13 12 11 10 09 08 07 06 05 04 03 02 01 (indentation near pin1)
.................................................. ................o.
.................................................. ...................
.................................................. ...................
15 16 17 18 19 20 21 22 23 24 25 26 27 28


the vcore is an isl 6529 chip, (its circled in orange in the pic), i need someone to measure the voltage on pin 5 and 9 (it may be hard to measure the pin, so you can measure the trace or resistor coming out of it...if measuring from the closest resistor, please give values for both sides)

i believe it goes (it may be upside down, so make sure to look for the indentation...indentation should be between pin 1 and 16)

.o.16 15 14 13...
01...................12
02...................11
03...................10
04...................09
...05 06 07 08...

dont you just love my sexy diagrams? :D :banana:

http://image1ex.villagephotos.com/2357846.jpg

Hejatz
04-28-2003, 02:42 AM
look for the Ti 4800 mod posted in this forum! it is the same controllerchip! have fun :D

berkut
04-28-2003, 09:39 AM
[i]the vddr is an isl 6225ca chip (circled in yellow)...need someone to measure the voltage on pin16...

Its 1.25-1.29V..


pin 5 and 9

pin 5 is 0.84V and 9 is 0.81V...

QuadDamage
04-30-2003, 03:14 PM
kinda low... are you sure about pins? Would be nice if can figure it out. there's some hidden potential hidden in this card...

gokickrocks
04-30-2003, 07:07 PM
these voltages were expected...

i may need berkut to measure some more pins (these were not the points for measuring actual voltage output, these were just feedback voltages)

the voltages are low because they are divided by a resistor that goes to a feedback to reference voltages going out, if the feedback senses a lack of voltage it tells the chip to send out more...

hopefully i could get a guide up on the weekend and someone can be a guinea pig :D

berkut
05-01-2003, 12:39 AM
Im always here to help with that vmod ;]

I believe that the Vcore of the FX is 1.25V (a 2 phase power supply and most of the capaticors is used for it), there are also some 2.6V mosfets, 3.5V (:O) and few others...

saaya
05-10-2003, 06:16 AM
1.25v? woa, quite surprising for a gpu that puts out more heat than a low end xp..

berkut
05-10-2003, 10:09 AM
I would say that the VPU isnt the biggest heat producer...

Im finishing my phase-change/water/pelt chiller for my fx.. currently on a V7+ i can hit 590mhz core/1020-1100 mem....

But ive got a very faulty card... The contact between the core and the IHS was actually on 20% of the surface (ripped the IHS off and the temps went down from 130-140C to 40C @ 400mhz/ 48C @ 590mhz...)... Fixed it... Then ive noticed that my back ramsinks are actually VERY unevenly touching the ram so only 2 "inner" chips were cooled, the rest touched the sink with a 1mm gap.. So heres the ram's big difference in o/c

Ive made already the back ram block (very similar to the original back plate but its a water block), front ramsinks are in produce, the VPU ramsink with the 120W pelt are almost ready also so im like sitting on needles waiting for those Vmods :>


gokickrocks - and what about the third square like controller near the 4 mosfets suplyying voltage for the VPU ? Isnt that acutally the Vcore regulator with the one below it for the rams and the second square like for the Vref lets say ?

saaya
05-10-2003, 02:43 PM
respect dude...i wouldnt risk too loose a 500€ card by doing any mods. well were at xtremesystems.org... :D

berkut
05-10-2003, 10:59 PM
^^

too bad ill finish the o/c bar ^^ anyone knows how to increase the clock above 600mhz...

saaya
05-10-2003, 11:06 PM
hows performence? i read at vrzone that ocing decreases performence somehow? they got slower benches with 590/1150 then with 500/1000. does the memory have to run exactly the double speed of the gpu?

berkut
05-11-2003, 03:48 AM
The problem is the fact that with normal cooling etc. the card declock's itsefl to 300/600 or 400/800 (mine) or smth. like that... So it works @ its. "astronomical" 600/1200 speed for few sec's...
The best way to measure the clock's on the card (wich works for 100%) is to measue its fill rate in 3dmark 2k1... ~1200 means its heavily declocked, ~1500 is @ 500mhz and almost 1800 is @ 590...

STEvil
05-11-2003, 05:19 PM
Its got thermal protection like P4's so it slows down when it gets too hot...

berkut
05-12-2003, 02:48 AM
The problem is not only that it has that thermal throttling but also when the VPU generates some errors it instantly downclock's it self

saaya
05-12-2003, 08:33 AM
be carefull when hacking the thermal protection...

berkut
05-22-2003, 03:27 AM
Yust found the first Vmod.. its done with a pencil, its for the 2.65V in my case, currently lifted it by 0.05V and the mems instantly went from max 1020 to 1030, considering a 10mhz o/c from a 0.05V vmod thats nice, think the Vcore mod can be done similary, ill post some results ltr

macci
05-24-2003, 05:47 AM
I think I measured the GPU voltage on this card once and noticed that its indeed like 1.25V but only in 2D mode. Once it goes to 3D mode it increased to like 1.5V I think. Can you confirm this?

berkut
05-24-2003, 12:10 PM
w8 ill see...

berkut
05-24-2003, 12:10 PM
damn im runnin on my r8500 now.. il have to mount my water block on the fx first :/

berkut
05-24-2003, 12:34 PM
1.34V... could you check it 4 sure was it 1.5V ? What is the max speed you could archieve on your fx..

macci
05-25-2003, 03:36 AM
Im not sure if it was 1.5V..point was that its higher in 3D mode than in 2D mode. And that obviously is the case.

Got mine up to 630GPU/1100mem w/ dryice on GPU. no ram cooling or vmods. It was a §§§§ty card thou did only 530/1070 stock and 545GPU w/ P3 heatsink and 7k delta on GPU.

berkut
05-25-2003, 04:35 AM
Guess mine is a monster if it comes to gpu overclocking... if i remember well FUGGER clocked it @ 590mhz with cooler water and a mcw50 with a pelt... im only using a Volcano 7+.... ive gotto get my hands on a bios from a ultra... Or a special bios, maybe nvidia made it so we can change voltages :D

berkut
05-25-2003, 11:29 AM
Easyly did 630mhz with a 120W pelt @ 12V... Maybe more but with some much colder water (was @ 35C) and the pelt @ 18-20V...

Or if my new project with a car injector for a expansion valve works out ill probably go direct die with ~ -40C evap temps...