• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

HD4870 Voltage regulators

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

El_Poochino

Registered
Joined
Aug 24, 2008
Hi all. First I'd like to say thanks for the info. I've lurked around here and the stuff helped me with my OC.

But I have a question.

How much stick can the voltage regulators on a HD4870 cop?

I put on a Zalman VF-1000 and have been trying to cool the voltage regulators. First I started with some chopped up copper RAM sinks, but they fell off (which is where I picked up this electrical buzz in ATI tool). Then I hacked a bit off the stock cooler, glued a lot of copper RAM sinks to it and replaced the thermal pad. I sanded off the paint and cleaned up the area when I stuck something on.

GPU-Z is reporting ~110C under load.

I've had a quick look around but can't really seem to find any info about limits and good cooling solutions. I don't want an OC to end in a puff of smoke and a light show.
 
when you replaced the stock thermal pads you prob lost some hight and are no longer making contact with the vregs mosfetts.

oh ya and :welcome: to OCF
 
Take some pics of the card so we can see what you're talking about. I also seriously hope you didn't just "glue" metal together. If you used some good thermal epoxy, that's fine, but no glue!
 
Used some thermal paste to make contact and a bit o super glue around the edges (after sticking it on) to stop it from falling off. Others are held on with the supplied thermal tape.

I'll get some pics now.

Edit:

P1010342copy.jpg

It is a Macgyver inspired heatsink.
 
Last edited:
Those big grey ones that are a bit under the edge of the GPU sink are part of the power circuits. I think your sawed-off stock sink is a bit short :) Stick a few RAM sinks on top of those and it should fix your problem.
 
the power delivery systems the 4850 and 4870 uses are vastly different. The 4850 uses the more old school way of power delivery where as the 4870 uses a three phase Vitec 59PR9853 chip (the big one). This chip can pump up to 120A sustained through it and has a fault tollerance between -40C and 125C and most likely a larger range with less Amps going through it. I dont want to get to technical here but it can easily handle the amps and temps it can sustain on the 4870 with no additional cooling. If you look at the stock heatsinks the 4850 heatsink covers all voltage regulators where as the 4870 doesnt cover the Vitec and pulse chips. As for the pulse chips those dont need to be actively cooled either. These chips are very similar to the ones used on the x1900 series which didnt need to be cooled actively either.

The only part of the 4870 power delivery system that needs active cooling are the smaller vrms. The large inductors on the 4850 are cheaper and less efficent and thus active cooling of them along with the vrms is all but required

I still maintain that when he replaced the thermal pad on the stock cooler part it didnt have enough height to maintain contact with the vrms. The heatsink that he made is good enough to cool the 4870 vrms provided that it actually touches them
 
I don't see how different the power supply scheme is? Same basic DC-DC conversion. Why should the large inductors on the 4850 be crap? They are larger probably because the chopping rate controlling the FETs is lower. Maybe that is all what is required. Many companies are saying more phases means more efficiency. I think that is bull****.
 
I don't see how different the power supply scheme is? Same basic DC-DC conversion.

thats like saying a cheap psu is the same thing as a 80+ psu because they do the same basic process.

Why should the large inductors on the 4850 be crap?

Im not saying they are crap. However like most midrange and lower products they are built to do the job they are certified for as cheaply as possible. The power system the 4850 uses works fine but like a cheaper psu isnt to effecient and as such puts out more heat in the power regulaton part of the card.

Many companies are saying more phases means more efficiency. I think that is bull****.

more phases does mean better efficiency if all factors are the same however in its current state on mobos phases mean little between manufatures. Sure more phases on a gigabyte board vs another gigabyte board with the same basic layout otherwise is more efficent but not nessisarily so between a "16 phase" gigabyte board, a "12 phase" asus board, and a 4 phase dfi board. Usually people just count the squares around the cpu and assume one per phase which is not always the case. I often see people saying one manufatures board is better then anothers baised on how many squares they count which is horribly flawed so in this case I agree with you hole handedly

I was not in any way trying to compair the phases between the 4870 and 4850 due to how different their power delievery is it wouldnt be an apple to apple compairison; instead I focused on the components used. The only time I even mentioned the word phase was when I mentioned how many phases the Vitec 59PR9853 chip has and certainly didnt compair it to the phases on a 4850
 
Last edited:
More phases keep a cleaner voltage than just a single phase. Boss is right, no active cooling is needed and I have to agree with him that he probably sanded to much off and or maybe unevenly and it's not making good contact with the vrms.
 
I'll try a remount with some thermal paste on the regulators and a new strip of themal pad. It's only no name jaycar stuff with god only knows thermal condutivity.
 
More phases actually lowers efficiency. Turn ON/OFF losses are higher. The 16 phase etc from mobo makers is an eyewash. Buck converter only support 6 or 8 phase maximum. In HD4850 v/s 4870, the power draw is different but VRM scheme is the same. Too expensive to use different regulators.
 
Last edited:
Back