4870x2 VRM "2" - Which is it ? Can you help ?

[{SAS}TB]

I have a watercooled 4970x2 (well 2) and both have EK Full Cover blocks on

They work fine (despite ATIs drivers) but the VRM temp of VRM "4" (I assume this is VRM "2" on a single card) gets WAY hotter than the rest very quickly.

The memory and rest of the temps are fine (memory 34 33 40 40 idle) the VRMs idle are 37 32 45 56 - As you can see VRM 4 is already a bit higher.

Running Furmark (or any GPU straining benchmark) the VRM temps get to around 58 58 68 120 !!! - And teh 4th VRM shoots there very quickly indeed :(

I have reseated the card, checked the termal pads, made sure there is good contact - but im not 100% sure which set of VRMs i need to concentrate on - can anyone help ?

http://img128.imageshack.us/img128/7104/dsc01923rv7.jpg

Heres the card with thermal pads applied (pre TIM) so im not sure if VRM 2 is the set above the central chip or to the right hand side next to the memory ?

Or am i completely wrong thinking its either ?

Can anyone help? It would me much appreciated

Cheers

TB

[spajdr]

Write about it to user Eddy_EK (he is very helpful guy directly from EK). Im sure that you need thicker thermal pad, but Eddy will tell you where to put it exactly.

[{SAS}TB]

Hi

A thicker pad is what I had in mind, just not sure where im putting them !

Ill give Eddy_EK a shout

Cheers

TB

[kiselk]

I had the same problem with triple-fan Accelero Xtreme cooler. VRM2 was hitting 80-90C under load. After touching the GPU with my fingers I knew VRM2 is situated between the GPUs above them - it was the hottest place, I couldn't hold my fingers there for more than 1-2 secs. On the pic it is marked with red. Green is VRM1 with OK temps 35-60.

I tried to apply thicker pads, adding some from reference cooler to VRM first, then to neighbor chips since they stopped touching the sink after I applied additional pad to VRM. This had no effect. The problem is Accelero cooler which has 3 sinks - 1 small, 1 middle size and 1 huge for GPUs. Alum sink for VRM2 is the smallest and it just can't give away all the heat it gets from VRM.

I am using my 4870x2 still with this problem, everyone who got Accelero cooler tells me it is the standard problem, guys with waterblocks also have this problem. I could not find a solution. VRM working temp is up to 120, 80-90 is not a problem as stated in ATI VRM supplier's whitepaper. No proof link though was months ago, I lost it. If you need some more info, try to google - there are a lot of guys trying to cool down 4870x2's VRMs.

It is very strange that your waterblock is not able to cool it. Probably thin pads. Is your waterblock hot there?
PM me if you find a solution, I want my 4870x2 to stay cool too until I get 5870x2.

[{SAS}TB]

Hi

First off thanks for your reply and info !

Odly enough the problem was worse with an older non-Q version of the card that I had (which has now been sold!) - I tried a number of different thickness pads, with and without TIM. The same block on other cards was OK. I had a second block which was also tried on two different cards - one had great VRM temps (a Gainward) the other was shocking (Sapphire).

I have had four x2s - two have been RMAd (one is still off being RMAd) and the VRM temps have varied drastically. When this final x2 returns I am selling that also.

These multi GPU cards do seem to have a limited lifespan - currently im using my wifes x600pro until something (possibly the 5870) grabs my attention :D

[kiselk]

Oh, do they die so quick?

There was one thing I noticed - the original pads were some sort of "hardened" ones for VRMs. Mem pads were like chewing gum and VRM pads were like dried glue, maybe even dried thermal grease, with fishing-line inside. That made me think I need something special for these VRMs. But the odd thing is that my second VRM is good.

As you tell me different manufacturers have different temp with your block I start thinking it is either pad thickness or the chips theirselves. Mine is Sapphire.

I am thinking about selling mine too in a couple of months, but who needs it after 5870 hits the market?

[jaredpace]

Nice to see you guys finding a solution, and thanks for that picture, kiselk. Personally, I would rather have two 4870 1gb cards vs. a 4870x2, likewise I would prefer two 5870s over a single 5870x2. Yes there will be more noise and space taken up, but you have more user control and headroom. Two sets of vrms on two pcbs with two fans and two large copper HS's. The overclocking potential is greater this way, and scaling is usually the same going either path. Some reviews even showed crossfire 4870 1gbs to be 2-6% faster than a 4870x2. Two 5870's in crossfire does take away the potential for 4-way CFX, unless you have 4 spacious pci-e slots or the accelero s1 trick on the 3rd card, but imo 2-way and 3-way are the sweet spots in crossfire.

[{SAS}TB]

The VRM pads are the same as the rest, its just they are buggered by the excessive heat from the VRM and degraded to a hardened mess ;)

[kiselk]

Nice to see you guys finding a solution, and thanks for that picture, kiselk. Personally, I would rather have two 4870 1gb cards vs. a 4870x2, likewise I would prefer two 5870s over a single 5870x2. Yes there will be more noise and space taken up, but you have more user control and headroom. Two sets of vrms on two pcbs with two fans and two large copper HS's. The overclocking potential is greater this way, and scaling is usually the same going either path. Some reviews even showed crossfire 4870 1gbs to be 2-6% faster than a 4870x2. Two 5870's in crossfire does take away the potential for 4-way CFX, unless you have 4 spacious pci-e slots or the accelero s1 trick on the 3rd card, but imo 2-way and 3-way are the sweet spots in crossfire.

I got 4870x2 for triple screens surround gaming setup to render all on 4870x2 and have third display via the second 4350 with SoftTH. No way I can have 2 4870 and hook up 3 screens as a single directX array. 5870 and 5870x2 will solve this problem, so I would rather have 2x5870 instead of 5870x2, BUT What if I get 3x2560 and want to play Crysis maxxed out? :))


The VRM pads are the same as the rest, its just they are buggered by the excessive heat from the VRM and degraded to a hardened mess ;)

I was actually talking about the original pads from the stock cooler. They were different for mem and VRMs. When I put the new cooler with all pds the same type, noticed the huge VRM2 temp and tried to pull off some pads for VRM from the stock cooler to make them thicker I couldn't do that because they fell apart and what I had is a tiny fishing-line net with some tiny bits of the rest of the dried grease. So I took the ones from mem from old cooler and put them along with the new ones on the VRMs after shaping them a little.
So all the pads were the same on the new cooler. On the old cooler pads on mem were like all on the new one, but pads on the VRMs were special.

[{SAS}TB]

I was actually talking about the original pads from the stock cooler. They were different for mem and VRMs. When I put the new cooler with all pds the same type, noticed the huge VRM2 temp and tried to pull off some pads for VRM from the stock cooler to make them thicker I couldn't do that because they fell apart and what I had is a tiny fishing-line net with some tiny bits of the rest of the dried grease. So I took the ones from mem from old cooler and put them along with the new ones on the VRMs after shaping them a little.
So all the pads were the same on the new cooler. On the old cooler pads on mem were like all on the new one, but pads on the VRMs were special.

I know - I was being serious - look at the rest of the stock pads - They are the same with the "mesh" embedded in them. The reason the VRM ones are hard and flakey (as they were with all of my x2s) is due to degrading because of the VRM heat. I had a less than month old Sapphire I took apart to watercool and even after such a short time the degradation had started - not as bad as some of the other cards, but it was starting all the same.

Another indicator of just how hot the VRMs get :D

[kiselk]

Take a look at the pic this is stock cooler off the 5870/5850 not sure.
4870x2 pads looked exactly the same even color when I took the stock cooler off.
All the pads have chips imprinted and the VRM is simply squashed. You can see some pieces don't fall off because they are held by a net of some kind of a fishing line which is a base for the whole pad

[kiselk]

Oh my apologies its not the VRM, its the neighbour small chips. VRMs have simple pads.

[largon]

Those "neighbouring small chips" ARE THE VRMS. The big grey blocky ceramic parts (inductors) marked in post #4 do not need any cooling.
No wonder stuff overheat when people have no idea what's what.
:rolleyes:

[kiselk]

OMG that means I was sucking the wrong balls all the time!!!


Thanks for statement, Largon. The thing that made me think those ceramic are VRMs were the VRM manufacturer PDF with a large photo of this long chip. The doc was about VRMs so I thought the pic is also related to VRM. Bummer. Now I know what I will be doing today coming home after my work - reassembling the cooler and removing the additional pad from the big one to make small ones touch the sink better.

Thank you!

pic taken from
http://en.expreview.com/img/2009/08/01/HD4890_powerchip_needcool.jpg

[kiselk]

This guy has the same accelero cooler working even worse :( 115 VRM load temp
http://forum.lowyat.net/topic/1139026

[texasreefer]

What application are you guys using for the VRM readings?

[{SAS}TB]

I use GPUZ http://www.techpowerup.com/gpuz/

[kiselk]

Uaaauuauaua, at last!

As is said in my previous comment I made third or maybe even a fourth attempt to remount my Accelero Xtreme cooler to my 4870x2. This time I removed the old grease from the GPUs, and removed additional pad from the chip I've mistakenly marked with red X as VRM. It is now in such a way that it barely touches the sink, but the whole pressure it put upon the VRMs between GPUs through two pads (one from original cooler one from accelero cooler). I did not touch the second VRM which is situated aside of the printboard.

My temps are 45 idle, 65 load for BOTH VRMs!!! They are finally identical.

I was playing Crysis, Demigod for 30 mins 1920x1200 8xAA @max settings and it made VRMs to heat only 65C. I was stunned. I was using Everest to monitor. For some reason a new version of RivaTuner does not show VRMs for my 4870x2. I will try to make some recorded graphs not to miss some peaks while I Alt+Tab and post them later.

Thanks to all of you for your help!

[largon]

You can get Rivatuner to display VRM temps by opening "Hardware monitoring" -tab -> click "Setup" in bottom right of the window -> click "Plugins" at bottom left and tick "VT1103.dll" from the list.

[kiselk]

You can get Rivatuner to display VRM temps by opening "Hardware monitoring" -tab -> click "Setup" in bottom right of the window -> click "Plugins" at bottom left and tick "VT1103.dll" from the list.

Oh, thanks, I lost my mind staring at that plugins list.
I will try VT1103 today.

[kiselk]

Well, heres the shot of the temps idle and under load. Everest was bull:banana::banana::banana::banana:ting me about 65C. It's almost 80 :(

[LiquidReactor]

Well, heres the shot of the temps idle and under load. Everest was bull:banana::banana::banana::banana:ting me about 65C. It's almost 80 :(

Wow...where exactly did you remove the extra thermal tape from?

[InCredible]

Well, heres the shot of the temps idle and under load. Everest was bull:banana::banana::banana::banana:ting me about 65C. It's almost 80 :(

80 seems fine..during benchmarking ive gotten temps in the 120's before and usually a crash shortly after...id think anything under 100c would be fine..i just ordered a waterblock for my x2 hoping to get in this week to control vrm temps better then this modded air heat sink.