ATI 4850/4870 voltmod thread

[Caveman787]

I think you mean now, good job.

How much were you able to get before you pencil mod'ed?

Anyone know if there's a pencil mod for the 4870?

[gurusan]

yeah I meant now, thanks for spotting that

At stock vGPU I can do 710mhz.

And no I don't think there is a pencil mod for the 4870, only hardmods.

[Caveman787]

I find it strange that a 4850 with less power (1 6 pin whether then 2) can manage to get up to 800-900 core easier then the 4870's and with lesser measures.

Eh guess that's true

[gurusan]

power supply is not an issue. My PCI-E slot can supply 150w while the PCI-E plug can supply another 75W. That should be more than enough...

Also I am using around .05v more than a 4870 and most 4870s can do 820-880mhz with good cooling on stock voltage.

[JeffLo]

I'd wait and see how the T Rad2 comes out and see if you can fit fans on that.

Accelero s1 rev2 's fanless or gfxchilla because it can fit in crossfire I believe, otherwise dunno what would fit.

JeffLo I think I may kill you for your computer lol and why get 4870 crossfire if you've already got gtx280?
.

thanks for the info. omg im scared already need to hide

i just wanna try 4870 cf. its arriving on tuesday or wensday gonna give ati a shot again been a sort of nvidia fanboy since my bad experience with ati long time ago

might wait on the t-rad2 wish it would be super slim so i can add a good fan to it may a slim type of fan like the scythe one, 100mm x 100mm x 12mm http://www.scythe-usa.com/product/ac...12_detail.html hope this would fit

[Tomasz25]

Here you go gentlemen. The (in)famous 4870 XOC BIOS. 128-kb, exported with GPU-Z 0.2.6. I've zipped it so that I can put it here

By the way, I have finally managed to make a fully stable setup with this card. I am running at 850/1150 artifact free and very solid. Max GPU temps are around 65C. The volts are 1.349 core and 1.598 memory.

may I use this BIOS on Sapphire HD4870 ?

[ImmortalX]

The people in the other thread have not had much success with the XOC BIOS. Seems people prefer the ASUS TOP BIOS.

[Tomasz25]

thanks

supposed I want to use GPU Voltage Mod (soldering) on 4850/4870 should I firstly flash oryginal BIOS to moded one? or may I just solder 100K and so on..

[stingush]

okey, not in purpose i just disolderd all the 4 points circled in the green circle in the picture(the 2 of the pencil mod and the other 2 ones in the bottom of them)
now the card doesn't work... is there any way to fix it? maybe to get the VGPU back to it somehow?

[Katzenschleuder]

Ok, well, I am sort of leaning now that despite the reported temps it is the cooling on the VRM modules that causes the cards to fail at higher volts. I was benching/stress testing with Crysis and I noticed that with the fan at 90%-100% I can get stable 875/1185 but only if I am running the fans so high. Drop the fans to say 50% and the test freezes. The volts needed for that clock are ending up to be 1.45V core and about 1.64V memory. Either way, I am pretty much convinced now that the 4870 XOC BE does not clock any higher than say 835/1150 on stock volts. Indeed, to be generally stable you need not go past 825/1125. Pretty lousy if you ask me for a card that states should be able to OC to 950/1200.

Yeah, maybe you can do 950/1200 with LN2 and about 1.6V on the core and another 1.7 on the memory. With air-cooling the best one can hope for that is true rock-solid stability (no ATITool artifacts, any game continuous 24/7) is about 850/1150 with roughly 1.35V core and 1.6V memory. Of course, as you can tell volt mods are required. Damn lying Diamond. They may have as well put 2000MHz max core speed and 4000MHz max memory, it would be just as realistic as the current CCC OverDrive limits for the 4870 XOC BE.

I now officially give the (3rd) 4870 XOC BE I tried and the whole line a big fat

Oh and for Diamond and their lousy lying.

Kid, how did you get the ridicules idea that Diamond lied if you can't overclock the card to "950/1200"?!

[dejanh]

(1) You are a fool to assume that I am a kid as I'd be betting I'm older than you. Next time, save your semi-derogatory comments for your buddies. This is a public forum so I'd ask you to both reserve yourself and to keep your snot-nosed comments to yourself.
(2) I am not expecting the card to clock to 950/1200, the card is not stable at factory clocks of 800/1100. So far I have tested 6 cards on three different systems and none could complete 1h+ of ATITool or more than 20 minutes of FurMark without artifacts or crashing. The problem however is that Diamond sells this card as a "superior overclocker" using a "customized bios engineered to unlock higher speeds" but they conveniently forgot to do anything on the card other than package it with a customized bios that allows you to crank the OverDrive to 990/1200 and click apply. Do you see a problem here, 'cause I sure do

So "kid", go troll elsewhere

[xMrBunglex]

the kid has a point.

[ImmortalX]

okey, not in purpose i just disolderd all the 4 points circled in the green circle in the picture(the 2 of the pencil mod and the other 2 ones in the bottom of them)
now the card doesn't work... is there any way to fix it? maybe to get the VGPU back to it somehow?

If you don't know soldering, you shouldn't attempt these mods.

What do you mean by desoldered? Did you remove all the solder from the pads? Is there a hole left on those points (it could've been a via)?

[Katzenschleuder]

(1) You are a fool to assume that I am a kid as I'd be betting I'm older than you. Next time, save your semi-derogatory comments for your buddies. This is a public forum so I'd ask you to both reserve yourself and to keep your snot-nosed comments to yourself.
(2) I am not expecting the card to clock to 950/1200, the card is not stable at factory clocks of 800/1100. So far I have tested 6 cards on three different systems and none could complete 1h+ of ATITool or more than 20 minutes of FurMark without artifacts or crashing. The problem however is that Diamond sells this card as a "superior overclocker" using a "customized bios engineered to unlock higher speeds" but they conveniently forgot to do anything on the card other than package it with a customized bios that allows you to crank the OverDrive to 990/1200 and click apply. Do you see a problem here, 'cause I sure do

So "kid", go troll elsewhere

This grossly contradicts your previous statement, where you said that "a card that states [it] should be able to OC to 950/1200" and then blame Diamond for "lousy lying".

This IS infantile nonsense defamation!

[Viilutaja]

after rubbing off my pencil off my 4850 and running at 700mhz stock volts for awhile I redid the pencil mod for 1.05v idle 1.35v load....and am now stable for my 24/7 clock of 820/1150mhz

Hello!
Can you make a high quality picture from your mods? And what kind of pen did u use? How strong grafit? I used 2HB and it's impossible to draw lines with that on PCB, it just wont stick

[dejanh]

This grossly contradicts your previous statement, where you said that "a card that states [it] should be able to OC to 950/1200" and then blame Diamond for "lousy lying".

This IS infantile nonsense defamation!

I don't see the contradiction. Did Diamond state that the XOC is a "superior overclocker" using a "customized bios engineered to unlock higher speeds" (quoting Diamond's words)? Yes they did. Did they claim that the cards will be able to overclock to 950/1200? Yes they did (look it up on the net if you don't believe me). Did anything they claimed live up to the hype they built and the specifications they stated? No it did not. Therefore, Diamond simply lied. Either that or their engineers are too stupid to be able to tell what is required to push these cards to the stated levels of performance and thereby made an honest albeit ignorant mistake

Anyway, that's the last comment I will make directed at you. I am going back on topic of 4850/4870 voltmod.

How are people finding the temperatures of their VRMs after volt modding the 4870. I am finding that especially on the memory, the VRMs get quite hot, to a point where they were burning off the foam that covers the VRMs on the stock cooler, and this is only at 1.62V. I know for a fact that this memory is rated to maximum 2.0V. We should definitely be able to push it to at least 1.8, 1.85V without problems :s

I think my VRMs get to about 80C atm...

[W1zzard]

latest everest beta can show vrm temps, so can the next gpu-z version. gpu temp #1/#2/#3 in gpu-z are not the VRM temps

[gurusan]

latest everest beta can show vrm temps, so can the next gpu-z version. gpu temp #1/#2/#3 in gpu-z are not the VRM temps

So I guess us 4850 users can't see our vrm temps?

[MonsterMash]

Hi,
there's some pencil vmod for the hd4870 cards?

Thank you.

Bye

[dejanh]

latest everest beta can show vrm temps, so can the next gpu-z version. gpu temp #1/#2/#3 in gpu-z are not the VRM temps

Care to share the latest GPU-Z? I have the 0.2.6 and it does not show VRM temps. In terms of the #1/#2/#3 temps, I already ate my own words, hehe

I am using Everest now to show the VRM temps. They get pretty darn high on the 4870, as you will see later in this post...

@MonsterMash - there are no pencil vmods for the 4870, only hard mods.

Finally, an update on the latest state of affairs on my end of successes (and numerous failures) volt-modding and finally OCing the 4870 XOC BE.

I managed to stabilize the last 4870 XOC BE I got. I am currently running at 850/1150 with 1.350V on the core (core voltage obtained through volt-mod) and 1.501V on the memory (stock memory voltage). ATITool is artifact free, 3DMark06 can run all test, including features and batch tests for hours without problems. Crysis benchmark runs without a hitch as well. Still no go on the FurMark. I realized that the reason it crashes is because the VRMs get way to hot. Using Everest I read the VRM temps at the moment when crash occurs in FurMark and they are as hot as hell. In fact, they pass 125C mark when FurMark blows up, which is much too hot. Unfortunately, there is no obvious solution to me at the moment that would do a better job at cooling the VRM. I installed the HR-03 GT which comes with bigger heat sinks than those with the Zalman VF1000 and the added fan blows a bit onto the VRMs. I could mod a fan to blow specifically on the VRM heatsinks, but I do not really care. In games the VRMs never pass 95C, so I am okay with that. It seems that I will be settling on this clock unless someone can help out with the VRM temps. I can push things to 875/1150 with 1.43V on the core, but I do not see the 25MHz gain to be justifiable for added 0.08V. It's not 950/1200, but I guess it is not bad. Only thing that sucks is that no matter what a core volt-mod is required if you truly want a rock solid stable 4870 XOC BE card.

If anyone has any suggestions on what to do to lower the VRM temps some more, I'd gladly try them out. Currently, that is my bottleneck to going higher with volts on core and memory needed for higher clocks. On every cooler so far VRMs were not cooled adequately (VF1000, HR-03 GT, stock). On the stock cooler the VRMs were burning off the thermal foam that contacts them. This is the single biggest determent I think to being able to OC higher. My highest core temps never pass 62C with the HR-03 GT. The single reported core temp actually never passes about 57C.

So, any advice on improving VRM cooling to try out that does not require taping a fan to the back of the card?

[Barr3l Rid3r]

Care to share the latest GPU-Z? I have the 0.2.6 and it does not show VRM temps. In terms of the #1/#2/#3 temps, I already ate my own words, hehe

I am using Everest now to show the VRM temps. They get pretty darn high on the 4870, as you will see later in this post...

@MonsterMash - there are no pencil vmods for the 4870, only hard mods.

Finally, an update on the latest state of affairs on my end of successes (and numerous failures) volt-modding and finally OCing the 4870 XOC BE.

I managed to stabilize the last 4870 XOC BE I got. I am currently running at 850/1150 with 1.350V on the core (core voltage obtained through volt-mod) and 1.501V on the memory (stock memory voltage). ATITool is artifact free, 3DMark06 can run all test, including features and batch tests for hours without problems. Crysis benchmark runs without a hitch as well. Still no go on the FurMark. I realized that the reason it crashes is because the VRMs get way to hot. Using Everest I read the VRM temps at the moment when crash occurs in FurMark and they are as hot as hell. In fact, they pass 125C mark when FurMark blows up, which is much too hot. Unfortunately, there is no obvious solution to me at the moment that would do a better job at cooling the VRM. I installed the HR-03 GT which comes with bigger heat sinks than those with the Zalman VF1000 and the added fan blows a bit onto the VRMs. I could mod a fan to blow specifically on the VRM heatsinks, but I do not really care. In games the VRMs never pass 95C, so I am okay with that. It seems that I will be settling on this clock unless someone can help out with the VRM temps. I can push things to 875/1150 with 1.43V on the core, but I do not see the 25MHz gain to be justifiable for added 0.08V. It's not 950/1200, but I guess it is not bad. Only thing that sucks is that no matter what a core volt-mod is required if you truly want a rock solid stable 4870 XOC BE card.

If anyone has any suggestions on what to do to lower the VRM temps some more, I'd gladly try them out. Currently, that is my bottleneck to going higher with volts on core and memory needed for higher clocks. On every cooler so far VRMs were not cooled adequately (VF1000, HR-03 GT, stock). On the stock cooler the VRMs were burning off the thermal foam that contacts them. This is the single biggest determent I think to being able to OC higher. My highest core temps never pass 62C with the HR-03 GT. The single reported core temp actually never passes about 57C.

So, any advice on improving VRM cooling to try out that does not require taping a fan to the back of the card?

Try this.... It seems to help over here










Sinks originais voltam para os Mosfets

[xMrBunglex]

the soon to be released Accelero Twin Turbo might do a good job of cooling the VRM's. its dual fan design looks like it will force cool air over the entire video card. they were supposed to be released last month and no sign of them yet. maybe the delay in release has something to do with the hot VRM's on our 4870's. hopefully the people at AC are working on a good cooling solution. they specifically mention the 4870 on its list of supported cards.

[dejanh]

Try this.... It seems to help over here










Sinks originais voltam para os Mosfets

That's a good idea, but it unfortunately does not work for the 4870.

(1) The stock cooler does not have a specific VRM heatsink. It is just touching the VRMs with the frame of the whole stock cooler. Not much cooling going on there either as the stock stuff burns under the heat of the VRMs.
(2) It seems a heatsink alone is not enough on the 4870 VRMs. They just get too hot.

the soon to be released Accelero Twin Turbo might do a good job of cooling the VRM's. its dual fan design looks like it will force cool air over the entire video card. they were supposed to be released last month and no sign of them yet. maybe the delay in release has something to do with the hot VRM's on our 4870's. hopefully the people at AC are working on a good cooling solution. they specifically mention the 4870 on its list of supported cards.

Yeah, but I really do not want to be buying yet another aftermarket cooler. Especially since I am not even sure how it would perform. I have the HR-03 GT on it now, and it does a helluva job cooling the GPU so I'd rather not change it. However, the VRMs do need something. Maybe a larger sink would work. I was considering buying this http://www.thermalright.com/new_a_pa...50aHVzaWFzdA== and actually attaching it to the video card VRMs. It is meant for motherboards, but hey, I can't see how it can hurt I am only concerned about the mounted weight. I think if I used Arctic Silver Adhesive it will hold, but I'd rather not as it makes removing impossible. I do not see 5 pieces of tiny thermal tape holding this up. Then again, I saw it in the store, and it is pretty small. It seems a lot bigger on the site.

[mth91]

Ok, because of my crazy temps I may have come up with a solution. Lower the volts. Sounds crazy right? Well here's my logic. I know that people can get higher clocks than 750 mhz with a voltage that's less than 1.33, and because the voltage is obviously too much for my card, I could probably lower it to decrease temps, but at the same time increase the overclockability because of lower temps. Is this logic correct?

Oh, and the smart guy that said 1 2 and 3 are not vrm temps was w1zzard for those who were wondering about my previous post. Hey w1zzard, can you post up the latest gpu-z and everest?

[dejanh]

Ok, because of my crazy temps I may have come up with a solution. Lower the volts. Sounds crazy right? Well here's my logic. I know that people can get higher clocks than 750 mhz with a voltage that's less than 1.33, and because the voltage is obviously too much for my card, I could probably lower it to decrease temps, but at the same time increase the overclockability because of lower temps. Is this logic correct?

Oh, and the smart guy that said 1 2 and 3 are not vrm temps was w1zzard for those who were wondering about my previous post. Hey w1zzard, can you post up the latest gpu-z and everest?

Sort of, in a convoluted kind of way. Lower volts however will result in lower overclocks as a rule of thumb. After all, the whole point of increasing volts is to be able to achieve higher clocks. However, if your temps are killing you because of higher volts, then maybe you do need to lower them to achieve a stable overclock.

Oh and I'd like that GPU-Z as well