Radeon HD 4850 Voltmods

[Velvet Wafer]

be careful with conductive liquids/paints/silver based stuff... they can short out your card in a sec.. silver paste killed my 4850... it got on the resistors surrounding the core.. it tried to boot in windows... CRACKLE.... card dead.

 

[cr1pz]

Daamn man ,that might hurt :S !!
so the best I can do is to patiently make some soldering once I have the cooling system ready... Is the safest then , I'll avoid the mod to fail because of false ''pencil'' contact and not to kill the card afterall.
Thanks for the advice

 

[Velvet Wafer]

also be careful with too much heat... the components hate that... safest would be pencil mod with a hb pencil...it doesnt crumble much

 

[theorw]

the manufacturer of your graphics card is surely able to read if an original bios is flashed. but i dont know if they do it...

btw
got a shortened 4850 in rma... its pretty common here too.

I wouldnt be so sure...
If u flash the original bios back i dont think they will notice anything!

 

[Velvet Wafer]

if its dead, you cant.

 

[cr1pz]

So , I bought a Staedler 2b , I should get an hb one then.
Thanks a lot Velvet ...

 

[Velvet Wafer]

lol you own a german pencil... staedler is like ferrari of pencils

btw.
MY 4850 TOXIC is back from RMA!!! lets see what it does!

 

[cr1pz]

Congrats man ,is the one you got killed by voltmod ?
About the pencil , as for me they're the best pencils ever , always i had to design or copy something's design at the Industrial High School I went I used Staedler

 

[Velvet Wafer]

no, i never failed at volts... but at silver paste on the transistors surrounding the core... i will never use that on gpu anymore..

EDIT: my card doesnt wants to clock to more than 10 mhz increase in memory... its at 1100 already, but the used samsungs are specced for 1200 mhz, and they did on my last card... today ram clocking went really bad after i changed to AC twin turbo... any ideas?

 

[Wile E]

no, i never failed at volts... but at silver paste on the transistors surrounding the core... i will never use that on gpu anymore..

EDIT: my card doesnt wants to clock to more than 10 mhz increase in memory... its at 1100 already, but the used samsungs are specced for 1200 mhz, and they did on my last card... today ram clocking went really bad after i changed to AC twin turbo... any ideas?

Do you have cooling on the memory?

And the memory may be either undervolted, or be running tighter timings, and that might be why you can't get any higher on the memory.

I'd focus more attention on the core speeds anyway, they'll show the most benefit.

 

[Velvet Wafer]

its a toxic.... vrm and rams with heatsinks.... mem is at 2.1v stock

i thinks its otherwise... the core is mighty enough... a 4870 is not much stronger in it.... but its rams.... are much faster.... so... until 1250mhz ram i had performance increases with my pre-rma model. even with 700 core. what do you say to that? what are your results`.

core fails at more than 745 with 1.19-1.20... but 745 is rock stable... more volts? how?

 

[Wile E]

its a toxic.... vrm and rams with heatsinks.... mem is at 2.1v stock

i thinks its otherwise... the core is mighty enough... a 4870 is not much stronger in it.... but its rams.... are much faster.... so... until 1250mhz ram i had performance increases with my pre-rma model. even with 700 core. what do you say to that? what are your results`.

core fails at more than 745 with 1.19-1.20... but 745 is rock stable... more volts? how?

I got increases by increasing my ram as well, but once I hit 1100MHz mem, I got larger increases by increasing the core speeds.

As far as more core volts, if your card has the digital voltera regulators, you can increase volts in BIOS or by using a program like ATI Tray Tools, GPU tool, or RivaTuner.

If it doesn't have regulators that can be controlled by software, the only thing you can do is a hard mod, or if you are lucky, there might be a pencil mod for your card.

 

[theorw]

I got increases by increasing my ram as well, but once I hit 1100MHz mem, I got larger increases by increasing the core speeds.

As far as more core volts, if your card has the digital voltera regulators, you can increase volts in BIOS or by using a program like ATI Tray Tools, GPU tool, or RivaTuner.

If it doesn't have regulators that can be controlled by software, the only thing you can do is a hard mod, or if you are lucky, there might be a pencil mod for your card.

4850s dont have Voltera s...
Theyd cost 40+ more bucks for this...
They only thing u can do is pencil or hard mod man!

Sapphire toxic 4850 has referrence PCB so u wont have problems finding the pencil or hard mod !

 

[Velvet Wafer]

u dont understand, i believe. i already did it to my last toxic (toxic is reference design,as said)... but, now, on the rma model the memory sux at proper voltage...and i dont know why... it runs even at 1230 with artifacts... but they only stop at nearly perfect 1100... i cant overclock even 10 mhz on that...WHY?

otherwise, how much volts can i put in a twin turbo, and what overclock should i reach until ive done the mod?

 

[rico82]

powercolor 4830, more Vgpu ... which resistor to pencil or alternative bios flash?

hi
i need help with the pencil mod or an alternative vgpu modification (up to 1.3V).

i cannot find the right resistor to "pencil" to 0.9 ohm.

ive already edit the bios with rbe to the max that was possible (1.158 V) -- and reached 770 / 1003.
with the 0.9 ohm (+DMM check) i want to pencil it to ... 1.25-1.3 V. (stock cooling solution , fan manually raised up to 100%)

or is there another solution without penciling it --- like a "unlocked 1.26V gpu" bios flash for my card, that works with my pcb ?
(the org bios is about 64k big)

its a
powercolor 4830 512mb
and the chip (vmod for gpu) says ...

uP6201BQ
VLP88N

pic // backside:
Powercolor 4830
512mb ddr3

maybe someone of you can help me out.

thanks in advance.
rico


[img=http://www.abload.de/thumb/img_2327lzsc.jpg]

additional pics of the card:
http://www.abload.de/image.php?img=img_2341tw4b.jpg
http://www.abload.de/image.php?img=img_2338rnx6.jpg
http://www.abload.de/image.php?img=img_2339h8da.jpg
http://www.abload.de/image.php?img=img_2336t8u9.jpg

 

[Wile E]

u dont understand, i believe. i already did it to my last toxic (toxic is reference design,as said)... but, now, on the rma model the memory sux at proper voltage...and i dont know why... it runs even at 1230 with artifacts... but they only stop at nearly perfect 1100... i cant overclock even 10 mhz on that...WHY?

Probably used lesser quality ram, or they tightened the timings on it.

otherwise, how much volts can i put in a twin turbo, and what overclock should i reach until ive done the mod?

Sorry, but the only way to figure that out, is to test for yourself. Every single card is different (as you have found out thru the ram not clocking).

 

[Velvet Wafer]

i meant the operating temperature... how much should i have maximum when i try to clock past 800+?

edit: remeasured the volts of the ram... under full load its 2.02-03... underspecificated.... could this be the problem?

theyre specced at 2.05 +- 0.5... meaning i can use 2.1 safely,am i right?

http://www.samsung.com/global/syste...oduct/2008/5/22/841580ds_k4j52324qh_rev10.pdf

i made the vcore mod yesterday... with 1.25 i get 770 stable, att goes until 784, then it finds an artifact

 

[Scrizz]

them mem doesn't make as much of a difference as the core.
Once you hit 800+ on the core u will see big gains.
the card can take 1.3v vgpu np
I had mine at 850/1100 @1.4v vgpu

 

[Velvet Wafer]

LOL.

Now the card is crap. after lowering resistance from 1380ohm to just 1150ohm, first no screen, after removing the pencil, nice bar like artifacts... the ram must be faulty from the beginning, it never got far from its minimal running spec, i suppose theyre binning is shit today...or i just had bad luck. i just penciled a QUARTER of the transistor, i was very careful... dead. only with artifacts. i would like to puke, and die from that.

besides from that, that somehow fucked with my bios.... he didnt recognized 2nd harddrive at first, only after 5 reboots, sometimes it appeared in bios, but not as bootable. system was horribly sluggish, so sluggish that it was nearly a crash (at 3.8 it got worser)

now everything is alright (besides the card) sapphire have an rma to await.

 

[Wile E]

i meant the operating temperature... how much should i have maximum when i try to clock past 800+?

edit: remeasured the volts of the ram... under full load its 2.02-03... underspecificated.... could this be the problem?

theyre specced at 2.05 +- 0.5... meaning i can use 2.1 safely,am i right?

http://www.samsung.com/global/syste...oduct/2008/5/22/841580ds_k4j52324qh_rev10.pdf

i made the vcore mod yesterday... with 1.25 i get 770 stable, att goes until 784, then it finds an artifact

1.25V isn't even as high as a 4870 yet.

These things can handle insane core temps. It's the vregs you must watch, keep them as cool as possible. At least under 90C, preferably under 80C, imo. (I'd say the same for the core as well.) A twin turbo should be ok all the way to 1.4v, so long as you keep the vregs cooled well.

As far as the mem, 2.1 should be ok, if that's what their specs are.

 

[Wile E]

LOL.

Now the card is crap. after lowering resistance from 1380ohm to just 1150ohm, first no screen, after removing the pencil, nice bar like artifacts... the ram must be faulty from the beginning, it never got far from its minimal running spec, i suppose theyre binning is shit today...or i just had bad luck. i just penciled a QUARTER of the transistor, i was very careful... dead. only with artifacts. i would like to puke, and die from that.

besides from that, that somehow fucked with my bios.... he didnt recognized 2nd harddrive at first, only after 5 reboots, sometimes it appeared in bios, but not as bootable. system was horribly sluggish, so sluggish that it was nearly a crash (at 3.8 it got worser)

now everything is alright (besides the card) sapphire have an rma to await.

Just saw this. A dying card won't do that to your mobo. Check your psu. A dying one will cause problems like that. It will even cause the graphical artifacts.

 

[Scrizz]

PSUs are the last thing u want going bad in your system.
One bad PSU can equal one dead rig (ie mobo fried, gpu, mem, hdd, etc..)

 

[Velvet Wafer]

my psu is a month old. all values except 5v are above theyre minimum level.

bios settings somehow work when i dont load saved profiles.
drive is found, but i had to run chkdsk from windows 7 to got many errors repaired that somehow appeared. now wcg will work again....

UPDATE: my computer froze, sound looped, but a few sec later it came back to life,suddenly... the computer freezes a few sec and then works again. it did 5 times during writing this.

what now?

i never encountered such strange behaviour... after reseating cables, drive got recognized normally
im somehow desperate... should i rma the card? whats with my psu?

*VELVET WAFER TAKES A PILLOW AND CRIES MADLY*

 

[Wile E]

Anything can cause that. From bad drivers, all the way to a failing mobo. Either way, something isn't right. I'd back up my important data now, if I were you.

And the psu being a month old doesn't mean it is good, nor does software giving good voltage readings. You have to see what it does under heavy load with a multi-meter.

First thing I would do is run the memtest 86+ boot disc, and test my ram.

Next thing I would do is back up my data, then try a clean install of Windows to see if my problems stop. If not, I would start testing the rest of my hardware.

 

[Velvet Wafer]

update.... opening of my RIGHT sight panel fixed the drive and sluggy probs somehow... now i will try the card.. i believe its dead anyway...

i have to say that the main cable string is most times stuck between a not so unsharp edge and the panel... do you think this can cause problems, even when the material coating of the cables isnt damaged?

memtest is integrated in my mobo... at least it looks the same... until now, the random crashes disappeared... but i feared to put my side panel on again