HD5970 cooling

[Painman]

What temp do the VRMs get to during gaming and furmark? Have you tried overvolting/overclocking at all and testing stability/temps? I'm curious to know as I will probably get a 5850 later on and do something similar to you.

1000/1250 is my OC. Final stable core voltage is 1.3125... more than I'd hoped, but temp management is under control. I'm guessing my core ended up as a 5850 due to leakage. It'll get up and dance, but yeah, I need to shock it...

My FurMark VRM temps are around 75º C with my Ramsink/fan arrangement. Core never really breaks 53º C while gaming, and it's in the same loop as my Q6600. The VDDC slave torture temp is pretty acceptable... general gaming VDDC temp is in the 60s. More than decent.

If you want a pic of my heatsink/fan arrangement for the VRM area, I'll snap one for you. It's not really complex. Stick-on RAMsinks and a bolt-on fan wired to a fan controller.

 

[thilanliyan]

If you want a pic of my heatsink/fan arrangement for the VRM area, I'll snap one for you. It's not really complex. Stick-on RAMsinks and a bolt-on fan wired to a fan controller.

Thanks for the info and YES I would definitely like to see a pic...I was considering modding the Zalman VRM heatsink I currently have to fit on the 5850 (I THINK it's possible with a bit of modding as the mounting holes seem similar but I'd have to cut away some pieces) BUT I would rather do what you describe and leave the backplate on.

 

[thilanliyan]

Just this dinky little thing.

If I don't run the fan and run OCCT they get super hot. With the fan, it's toasty but not extreme.

I don't know...maybe it was me running the fan at only about 40% but like I said even without the heat of the GPU around...the VRMs were still getting very hot. It was not impossible to run but I'm a bit of a cooling freak so I didn't like to see the VRMs break into the 80C range, which is why I bought the Zalman sink.

 

[Schmide]

VRMs are not like regular chips, they can typically depending on the part run 100c no problem. 80c for a VRM under load would be well within spec, in fact for a high wattage card, temperate.

 

[thilanliyan]

VRMs are not like regular chips, they can typically depending on the part run 100c no problem. 80c for a VRM under load would be well within spec, in fact for a high wattage card, temperate.

I know, I think the reference cards have VRMs that can go up to 120C but I just like to keep it all as cool as I possibly can...I don't think I can run them any cooler than they are now with aircooling which is why I stopped trying for better at that VRM heatsink.

When I was running the backplate, if I ran furmark they would get above 90C fairly quickly and I had to manually shut down the program because it would just keep getting hotter and hotter.

 

[cbn]

Internal complexity doesn't play too much into price. All full cover blocks are going to be in the $90-150 range. Here's a comparison of VRM coverage on all 5970 blocks - but note that the final version Dangerden block has been revised to run over the VRMs (no internal shots of the retail block yet)

It seems these fullcover waterblocks are either getting the middle VRMs or the right sided VRMs....but not both.

How much worse will VRMs do if not receiving water? Could a regular finned air heatsink end up be a better idea in cases where the waterblock itself (rather than the water passage) ends up covering up the VRM?

Some of these more simple waterblock arrangements covering just the GPU core supplemented with ramsinks/VRM heatsinks (for other areas of the video card) actually look rather efficient.

 

[cbn]

and those EK temps are probably not representative.

Are you talking about the VRM temps? Are the temps for those listed under GPU-Z as VDDC voltage?

 

[mmnno]

It seems these fullcover waterblocks are either getting the middle VRMs or the right sided VRMs....but not both.

How much worse will VRMs do if not receiving water? Could a regular finned air heatsink end up be a better idea in cases where the waterblock itself (rather than the water passage) ends up covering up the VRM?

Some of these more simple waterblock arrangements covering just the GPU core supplemented with ramsinks/VRM heatsinks (for other areas of the video card) actually look rather efficient.

The water doesn't really need to go right over the VRMs. All of those blocks have the important groups covered just fine, and from those pictures the Koolance seems does the best job (EK would dispute that I'm sure.)

Ramsinks are probably not going to cut it on the 5970, just because of the sheer power load of the card. At 5870 speeds (which is what everyone wants to run it at) it is way out of the ATX spec. The original version of the DD block had VRMs at 58 deg C with 20C water. That's a +38C delta under water. Sinks might do better than the ATI cooler, but that's far from ideal.

The 5970 is probably the first card that requires a full cover block for great overclocking, though since the stock cooler is (just barely) adequate for moderate OC anything more is gravy.

Are you talking about the VRM temps? Are the temps for those listed under GPU-Z as VDDC voltage?

That's right.

 

[Painman]

Thanks for the info and YES I would definitely like to see a pic...I was considering modding the Zalman VRM heatsink I currently have to fit on the 5850 (I THINK it's possible with a bit of modding as the mounting holes seem similar but I'd have to cut away some pieces) BUT I would rather do what you describe and leave the backplate on.

Hey, sorry for the delay and the blurry pics, but here you go...

Pic 1

Pic 2

Pic 2 came out nasty, but you can see the row of 5 RAMsinks and I pulled the fan aside to show more of its mount... kinda. Just a long shank #4/40 screw through one corner, secured with nuts and fiber washers through an isolated mount hole.

Hope those help.

 

[thilanliyan]

Hey, sorry for the delay and the blurry pics, but here you go...

Pic 1

Pic 2

Pic 2 came out nasty, but you can see the row of 5 RAMsinks and I pulled the fan aside to show more of its mount... kinda. Just a long shank #4/40 screw through one corner, secured with nuts and fiber washers through an isolated mount hole.

Hope those help.

Thanks very much for the pics...so did you actually cut out the part of the baseplate that contacted the VRMs and then stick the heatsinks on? It's a bit hard to see if the heatsinks are in direct contact with the VRMs.

 

[Painman]

The heatsinks are attached to a copper bar that's bonded to the baseplate over top of the VRM area. I didn't do any cutting... baseplate is still stock.