ATI 4850/4870 voltmod thread

[Mr.Guvernment]

i know i will be coming back to this thread once i get WC for my 4870!

[ceemic]

OK, Vgpu mod done and working
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_01.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_02.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_03.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_04.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_05.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_06.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_07.jpg
Maybe I'll do mem mods also...

[SNiiPE_DoGG]

OK, Vgpu mod done and working
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_01.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_02.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_03.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_04.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_05.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_06.jpg
http://ceemic.pri.ee/hardware/hd4870/hd4870mod_07.jpg
Maybe I'll do mem mods also...

looks very nice , where did you pick up that circuit board and switch? my local radioshack has bullocks for usable parts

[Barr3l Rid3r]

sweet but that switch shoul have loss of current on it.

[ceemic]

Barr3l Rid3r, what do You mean with that exactly? Ofcourse it will add a little resistance but in the end it sums up with VR and doesn't really mather...

[Barr3l Rid3r]

Barr3l Rid3r, what do You mean with that exactly? Ofcourse it will add a little resistance but in the end it sums up with VR and doesn't really mather...

I wasn`t talking about resistance and decreased voltage, I`m talking about current.

[Vatos_locos]

sweet but that switch shoul have loss of current on it.

it's not so bad the current at the core voltage is not huge to create a problem with the switch

[Barr3l Rid3r]

it's not so bad the current at the core voltage is not huge to create a problem with the switch

I`ve learned that since I`ve modded mine MSI K8N-Neo2 w/ Directwire 3.3v mod to feed mine BH-5. I`ve used an ordinary switch first and computer wasn`t stable and I`ve seem little to none OC gain...

When I changed for a bigger and more professional switch, all the good was born... stability and much higher BH-5 OC then before...from 230Mhz Cas 2 5-2-2 to 280Mhz x2 Cas2 5-2-2 1T.

Ok, older DDR chips must probably consume more current and power then todays GPU... but it might interfere anyway

[ceemic]

I have used these switches many times before in motherboards and graphics Vmoding. No instability or something like that so far
Also Jaan uses them (he is truly one of the best Vmoder ever ).

[jaredpace]

hey guys, biosmod possible for vgpu on 4850/4870?

[SNiiPE_DoGG]

hey guys, biosmod possible for vgpu on 4850/4870?

certainly not on the 4870

[G H Z]

I`ve learned that since I`ve modded mine MSI K8N-Neo2 w/ Directwire 3.3v mod to feed mine BH-5.

There's a big difference between feeding a power circut & altering output voltage with a FB mod

[Barr3l Rid3r]

There's a big difference between feeding a power circut & altering output voltage with a FB mod

I got it now, thanks preatty clarifying

[Vatos_locos]

Barr3l Rid3r 3.3 with 1.x Voltages is a big difference in current all tho the switch is not a problem at all it is a simple switch wen it is on it has no resistance to hold the current and the Ram's current is mush more than on one vga's core and everything that G H Z said it is a different think

[NBF]

Memory Vmod is working on my 4850, running at 1165 so far.

But I still can't figure out VGPU mod by myself

I would appreciate any help.

[ceemic]

Has someone done Vddc and Vddq mod for 4870? How much You gain in overclock?
Or is it pretty pointless to do that?

[8800's@over]

Has someone done Vddc and Vddq mod for 4870? How much You gain in overclock?
Or is it pretty pointless to do that?

I also wanted to know.


Guys, I am taking Black Screen with the Core at more than 815MHz, it's just more work when the memories of over 1065MHz, because it occurs? It is stable under FURMARK.


And if I did the Mod vGPU to try to go to 870MHz, even then I would have problems with BSOD?

[Death909]

Right, finally managed to obtain a trimpot of reasonably correct resistance. Jaycar kept selling me 5k trimpots when asked for 100k. I did the memory voltmod a week ago however.

Going from 2.05v to 2.13v (with a 2k trimpot at max resistance this happens) doesn't really allow much of a higher overclock. It's as stable at 1160 as it is at 1140, but it still artifacts after 30min. I have to say, I'm not impressed with the thermalright hr-03 gt severely impeding the cooling of 3 ram ICs.

Since I only did the Vgpu voltmod yesterday night (at 2am, only woke up like half an hour ago), I don't have as much of a conclusion yet. Boosting the voltage by 40mv or so appears to net stability at 780mhz instead of 750mhz, but I haven't really pushed it (that and it's summer in Australia now, and I'm look at ~47 degrees right now on one of the gpu temp sensors, instead of the previous ~38 degrees I was seeing like two weeks ago).

Someone requested a visual description of the voltmod? The added blue line shows where I soldered (onto one side of what I think is a capacitor). The added light green dots show what I think might have been some other possible solder points, but I think there's one missing, and I'm only really sure about the top left one, since I attempted to solder there.



I used a 100k trimpot instead of a 200k trimpot, and I'm getting the impression that maybe I should have gone with 50k. The NCP5388 advertised a different form of error correction (for one thing, the circuit diagrams in it's datasheet don't have a resistance to FB from ground and from Vgpu, instead they connect through resisters and capacitors to pin 20 and pin 18). Regardless, switching on the voltmod with the trimpot at 100k results in a 10mv jump, and a lot of trimpot turning was required for 40mv.

(Excuse the lack of concrete voltage measurements, I was using a different ground point and the measured Vgpu under load was different from my previous measurements, so I instead recorded the difference between the voltmod being turned off and on.)

[GamerBR]

Some interesting news,the new CCC has control over the voltage of the gpu.With the 8.9 I was getting 1.38V with the pencil mod,but I wanted to lower it a little bit,so I flashed the bios with a lower voltage,and I could get the voltage lower.Now after formatting my PC and installing 8.10 I noticed the voltage wouldn't go down from 1.38V,tried flashing it with the lowest possible option from RBE and it still wouldn't work.Then I tried disabling CCC startup,the voltage would get lower.After trying some different things,I discovered that only if Overdrive was enabled that would happen,so what it does is put the maximum possible voltage while overclocking(pretty much useless,it already max the voltage while on load,I don't get the point here).It would be nice if someone was able to hack this to include voltage selecion somehow while in Windows

Now I'm going to have to redo the pencil mod because of that

[jdrom]

Well I attempted doing the pencil mod via TechPowerUp's instructions for the HD4850 and it did not go well at all

I've since erased and cleaned with alcohol the pencil traces in hopes of reviving my GPU but that didn't help either.

Basically I have total screen corruption. Some random characters with a few that are readable along with some funky colours. It's also extremely low resolution (or high?) that my LCD cannot display it at times. It doesn't work at all using pure DVI, but I can get the above stuff using DVI-to-VGA.

I'm assuming there's no way to fix this, but if there is a way, I'm willing to try. I'm not exactly expecting a response from Diamond anytime soon.

[GamerBR]

Well I attempted doing the pencil mod via TechPowerUp's instructions for the HD4850 and it did not go well at all

I've since erased and cleaned with alcohol the pencil traces in hopes of reviving my GPU but that didn't help either.

Basically I have total screen corruption. Some random characters with a few that are readable along with some funky colours. It's also extremely low resolution (or high?) that my LCD cannot display it at times. It doesn't work at all using pure DVI, but I can get the above stuff using DVI-to-VGA.

I'm assuming there's no way to fix this, but if there is a way, I'm willing to try. I'm not exactly expecting a response from Diamond anytime soon.

I bet you did it by eye without a multimeter to measure the before and after resistance,you probably make it go to low,damaging the gpu by excess of voltage.

[jdrom]

I bet you did it by eye without a multimeter to measure the before and after resistance,you probably make it go to low,damaging the gpu by excess of voltage.

Yeah =/ I've definitely learned my lesson though.

The way TechPowerUp made it sound was that the maximum it would do was 1.4V so I thought it wouldn't do any harm but I was surely wrong assuming that.

[DemonEyez]

I have 1 of these cards. I've just doen the standard hardmod with 100k pot and it works (well card boots into windows OK). Waiting for glue to dry now before I do anything else with it.

Edit. I'm sure I sdet the resistor to 100k beforer soldering the mod, but I've got a core voltage of 1.4v now, which is fine for me but I gather its what people consider to be the upper limit for these cards.

Which one do you have? The EAH4850 512M with 3 phase pcb or the EAH4850 TOP with 4phase pcb?

Is there a known pencilmod/hardmod for the ASUS 4850 with Glaciator fan?

Just found out that pcb is a non reference model


Edit: Forgot to mention, pcb is the 3 phase design, same used in some Asus 4850 TOP.

I was under the impression that the Asus EAH4850 TOP uses a 4 phase pcb design. Can anyone confirm this? These pictures are from the Asus website and show 4 phase pcb for the TOP and the 1024MB cards, only the 512MB uses 3 phase:







I planning on buying a second HD4850 for running crossfire with my first Asus Reference HD4850, which has been voltmodded and runs at 850/1150.
I didn't read this topic (and others) for a few weeks and now to my surprise its realy hard to find those reference pcb single slot HD4850's

I'm going to voltmod the card so i'm looking for a card that has a working voltmod for it. So with all of this in mind , I thought I had the found the right card for the job in the EH4850 TOP

[largon]

I was under the impression that the Asus EAH4850 TOP uses a 4 phase pcb design. Can anyone confirm this? These pictures are from the Asus website and show 4 phase pcb for the TOP and the 1024MB cards, only the 512MB uses 3 phase: (...)

Asus EAH4850/HTDI/512M has 3 phases.
Both EAH4850/TOP/HTDI/512M and EAH4850/HTDI/1G have just 2 phases, not 4 (inductors ≠ phases).

Well I attempted doing the pencil mod via TechPowerUp's instructions for the HD4850 and it did not go well at all
(...)
I'm assuming there's no way to fix this, but if there is a way, I'm willing to try. I'm not exactly expecting a response from Diamond anytime soon.

What did you ask from Diamond? To help with a vmod?

[NBF]

Asus EAH4850/HTDI/512M has 3 phases.
Both EAH4850/TOP/HTDI/512M and EAH4850/HTDI/1G have just 2 phases, not 4 (inductors ≠ phases).

Don't know why the top would have less phases


I can't mod the 3 phase so far, maybe would be better sell it.