Can anyone figure out a way to cool MCP on 780i for me?

[lopri]

I just received a 780i board from EVGA Step-Up, and the overall feel of the board is solid. But knowing that NV chipsets tend to get hot, I had a Thermalright HR-05 SLI and Evercool Twinkling VGA ready for NB and SB, respectively. (which were used on my previous 680i setup)

The trouble that I didn't think of is, that the 780i has an extra chip for PCI-E 2.0 support and the stock heat pipe cooling covers both chips with one heat sink. There is no solution that I can think of to cover both chips..

http://www.techreport.com/articles.x/13790

But real trouble is actually the SB. Covered by thin aluminum which connects to NB heat sink via heat pipe, the SB temp is through the roof. (like 90C) So the thing is, I CAN use an after-market cooling for SB but that leaves me no solution for the NB and the PCI-E 2.0 chip. :frown:

Any suggestion? The high SB temperature is a real concern to me.

Thanks in advance,

 

[aigomorla]

Originally posted by: lopri
I just received a 780i board from EVGA Step-Up, and the overall feel of the board is solid. But knowing that NV chipsets tend to get hot, I had a Thermalright HR-05 SLI and Evercool Twinkling VGA ready for NB and SB, respectively. (which were used on my previous 680i setup)

The trouble that I didn't think of is, that the 780i has an extra chip for PCI-E 2.0 support and the stock heat pipe cooling covers both chips with one heat sink. There is no solution that I can think of to cover both chips..

http://www.techreport.com/articles.x/13790

But real trouble is actually the SB. Covered by thin aluminum which connects to NB heat sink via heat pipe, the SB temp is through the roof. (like 90C) So the thing is, I CAN use an after-market cooling for SB but that leaves me no solution for the NB and the PCI-E 2.0 chip. :frown:

Any suggestion? The high SB temperature is a real concern to me.

Thanks in advance,

:T

nothing for air. 1 solution for water.

I was thinkn maybe an Moded Enzo mini sink would fit on that little chip. Would need to use some time of adhesive tho. Thats the only downside.

Anyhow havent had time to test it or set it up.

 

[lopri]

Thanks for the suggestion but I'd reserve the water for the braves.

But the trouble is, as you can see, the mounting holes for SPP + NV200 heat sink are totally off-standard. I just touched the MCP heat sink, and it's scorching hot - so the contact must be there. Checked out EVGA forum and it looks like RMA is in order. The normal operating temperature looks like 50~60C for MCP. I have no idea why NV keep using this troublesome MCP55 (570 SLI). Ugh.

 

[DerwenArtos12]

Wow, they're getting PCI-e2.0 through an external chip, that's the second most retarded thing from nVidia I've ever seen. Only way I can figure to mount anything to those is going to either be with custom aluminum mounting brackets or thermal adhesive. Sorry.

 

[Fallen Kell]

Well, the 790i replaces the chipset with one that has PCI-E 2.0 built-in, which is part of the reason why most of the big name motherboard makers (ASUS, Gigabyte, ABIT) have not released 780i mothboards. The only issue is if DDR2 will be supported or not.

Back to topic. Yeah, your board seems bad. My temps are around 58C on my MCP (a little higher then I want, but since it is in an inverted case, I guess that is all I am going to get).

 

[krnmastersgt]

The nforce 200 was it? In any case it doesn't heat up much when I use it, wasn't it just placed there for the 2.0 and to help with the tri-sli? In any case water would probably be the way to go, I think the EVGA forums have some guys modding their boards for just that, might want to check em out.

 

[DerwenArtos12]

Originally posted by: krnmastersgt
The nforce 200 was it? In any case it doesn't heat up much when I use it, wasn't it just placed there for the 2.0 and to help with the tri-sli? In any case water would probably be the way to go, I think the EVGA forums have some guys modding their boards for just that, might want to check em out.

What they're doing, from the ones that I could find is just fabbing and aluminum or acetal or plexiglass bracket to mount the blocks to them mounts in the factory. IMHO the same could be done with an aftemarket air cooling solution.

 

[lopri]

After an hour of labor (taking off the foam and applying fresh layers of Ceramique, as well as lots of tightening) I'm down to 70~80C. Spoke to EVGA and as long as the board itself isn't damaged, I can do anything I want with the heat pipe or custom cooling, so I will see what I can do before I send it back.

Originally posted by: DerwenArtos12
Wow, they're getting PCI-e2.0 through an external chip, that's the second most retarded thing from nVidia I've ever seen. Only way I can figure to mount anything to those is going to either be with custom aluminum mounting brackets or thermal adhesive. Sorry.

What's the most retarded thing from NV you've seen?

BTW, while it may not be the most elegant solution to use an extra chip for PCI-E 2.0, but it is quite practical if you think about it. Today's north bridges are a lot more stressed than yesterday's. Users are demanding 400FSB+ as well as 4~8GB of RAM (meaning 4 sticks). This is a quite recent phenomenon, and adding communication of two or three GPUs in there surely wrecks havoc. By having an NV200 responsible for inter-GPU communication, the SPP (north bridge) gets a single signal from the NV200 for display information - a huge break for the already over-burdened chip.

Eventually they will come up with a better solution, but the current implementation isn't bad at all, IMO. I just wish they had considered the possibility of after-market cooling.

 

[DerwenArtos12]

Originally posted by: lopri
After an hour of labor (taking off the foam and applying fresh layers of Ceramique, as well as lots of tightening) I'm down to 70~80C. Spoke to EVGA and as long as the board itself isn't damaged, I can do anything I want with the heat pipe or custom cooling, so I will see what I can do before I send it back.

Originally posted by: DerwenArtos12
Wow, they're getting PCI-e2.0 through an external chip, that's the second most retarded thing from nVidia I've ever seen. Only way I can figure to mount anything to those is going to either be with custom aluminum mounting brackets or thermal adhesive. Sorry.

What's the most retarded thing from NV you've seen?

BTW, while it may not be the most elegant solution to use an extra chip for PCI-E 2.0, but it is quite practical if you think about it. Today's north bridges are a lot more stressed than yesterday's. Users are demanding 400FSB+ as well as 4~8GB of RAM (meaning 4 sticks). This is a quite recent phenomenon, and adding communication of two or three GPUs in there surely wrecks havoc. By having an NV200 responsible for inter-GPU communication, the SPP (north bridge) gets a single signal from the NV200 for display information - a huge break for the already over-burdened chip.

Eventually they will come up with a better solution, but the current implementation isn't bad at all, IMO. I just wish they had considered the possibility of after-market cooling.

I'm gonna have to give you a little bit of background for the retardedness to really sink in so bear with me, this is going to be yet another long winded post of mine.

A long time ago in a galaxy far far away there was a 3D graphics company founded by people who knew fundamentally little about 3d graphics but, plenty about IC design. Slowly but surely the company gained a serious foothold on the market with the advent of the PC gaming enthusiast. Along the way many of the founders dropped off to work on other projects they were more interested in and few stole away to other start-up 3D processor companies and they kind of lost their way. At the peak of their performance they decided to push the envelope a little further and put two GPU's on a single card, then four gpus on a single card. Coupled with astronomical ammounts of memory for the time these cards could not be beat. But, in their performance kingship they got complacant and slothly. And out of no where a competitor grew from not-so-humble beginings and started eating away at their performance lead in very big chunks. Before the King could take notice and make haste to defend himself it was too late, he was overthrown and there was much upheval throughout the land. The now former king had a great plan to take back his kingship but, it was already too late, the new king had been working on more than one plan and was a step ahead in ever direction. Not so long later the former king slowly drifted into a colapse and death.

Many years later the new king with all his new adversaries had a grand idea. What had once challenged them to become their current greatness was their best plan to stay at the top of their kingdom. So they dug back in the history books and found a name, what a grand name it was, scalable link interface. OOH, it sounds so technical and mysterious it must be great. But they thought to themselves, how can I take the former kings idea of having multiple GPU's without complicating the design process so I can roll it out at a much smaller cost, regaurdless of how much it would improve performance to do it any other way. What did they come up with? We'll just make people buy two whole cards! who cares if it only improves performance by 20-60% instead of 100% and clogs up the chipset with astronomical ammounts of data, it'll be cool!

And thats the most retarded thing nVidia ever did.

I'm not saying I dislike SLI, it has it's purposes and their more recent implimentations are very very well done but, the original implimentation and taking over the name from a different technology from a company they put out of business more than a decade ago, retarded!

 

[DerwenArtos12]

Originally posted by: lopri
BTW, while it may not be the most elegant solution to use an extra chip for PCI-E 2.0, but it is quite practical if you think about it. Today's north bridges are a lot more stressed than yesterday's. Users are demanding 400FSB+ as well as 4~8GB of RAM (meaning 4 sticks). This is a quite recent phenomenon, and adding communication of two or three GPUs in there surely wrecks havoc. By having an NV200 responsible for inter-GPU communication, the SPP (north bridge) gets a single signal from the NV200 for display information - a huge break for the already over-burdened chip.

Eventually they will come up with a better solution, but the current implementation isn't bad at all, IMO. I just wish they had considered the possibility of after-market cooling.

They already have a better solution on it's way out in the form of the 790i where the only change is teh inclusion of internal PCIe2.0. Read the whitepages on teh NV200, realistically it's only purpose is to take a single PCIe1.1 x16 slot and turn it into two PCIe2.0 x16 slots. It's not taking the northbridge out of the loop on communications at all, it's just funneling said communications to the northbridge.

 

[lopri]

And do you think that 790i launch will be smooth? I'd expect even bigger fiasco than 680i's debut when 790i comes out.

 

[DerwenArtos12]

Originally posted by: lopri
And do you think that 790i launch will be smooth? I'd expect even bigger fiasco than 680i's debut when 790i comes out.

I didn't say the launch was going to be smooth, or driver support would be existant, or that there would be no heat issues on the northbridge. I just said it was a better solution and it's in the works. I didn't even say it would work as well as the current set-up, let alone better. For all we know they're just integrating the silicon from the NV200 into the 790i northbridge and calling it a day.

 

[lopri]

Since EVGA said that it's OK to do pretty much anything as long as the board itself isn't physically damaged, I'm considering to cut off part of the heat pipe solution and leave the NB/VRM part as it is, but replace the SB heat sink with either Thermalright or Evercool.

http://img90.imageshack.us/img90/4018/evgasg4.jpg

I wonder if it's a good idea? The biggest concern after the mod would be whether it'd have a negative effect on NB cooling (heat pipes supposed to have 'liquid' in it) and before the mod is how I'm going to cut it. From the looks of it, the SB to NB path is made of 2 copper heat pipes covered by stainless steel. (or aluminum?)

 

[WoodButcher]

I can't tell you what metals are used on your board but the "copper" heatpipe system on my commando is all aluminum.

 

[DerwenArtos12]

Originally posted by: lopri
Since EVGA said that it's OK to do pretty much anything as long as the board itself isn't physically damaged, I'm considering to cut off part of the heat pipe solution and leave the NB/VRM part as it is, but replace the SB heat sink with either Thermalright or Evercool.

http://img90.imageshack.us/img90/4018/evgasg4.jpg

I wonder if it's a good idea? The biggest concern after the mod would be whether it'd have a negative effect on NB cooling (heat pipes supposed to have 'liquid' in it) and before the mod is how I'm going to cut it. From the looks of it, the SB to NB path is made of 2 copper heat pipes covered by stainless steel. (or aluminum?)

Be aware that as soon as you break the hermetic seal on your heatpipe it will be more or less completely useless. It's metal so it'll still conduct heat but, it will no longer be a superconductor. Sorry.

 

[VirtualLarry]

Originally posted by: lopri
Since EVGA said that it's OK to do pretty much anything as long as the board itself isn't physically damaged, I'm considering to cut off part of the heat pipe solution and leave the NB/VRM part as it is, but replace the SB heat sink with either Thermalright or Evercool.

http://img90.imageshack.us/img90/4018/evgasg4.jpg

I wonder if it's a good idea? The biggest concern after the mod would be whether it'd have a negative effect on NB cooling (heat pipes supposed to have 'liquid' in it) and before the mod is how I'm going to cut it. From the looks of it, the SB to NB path is made of 2 copper heat pipes covered by stainless steel. (or aluminum?)

If the assembly is all one (shared) heat-pipe system, then I wouldn't dare cut it open. If it's a bunch of seperate connecting heatpipes, then I'd say that you might be ok. I'd contact the mobo mfg and find out.

 

[aigomorla]

EDIT:

Nm.... it wont work... looks like im waiting for the EK solution.

 

[lopri]

Originally posted by: VirtualLarry
If the assembly is all one (shared) heat-pipe system, then I wouldn't dare cut it open. If it's a bunch of seperate connecting heatpipes, then I'd say that you might be ok. I'd contact the mobo mfg and find out.

It's two copper heat pipes going from the MCP to all the way up where MOSFETs are. Guess cutting can't be an option. In the mean time, I've conceived a ghetto-mod until I find a solution. This lowered the temps further down to 60C+.

http://img134.imageshack.us/im.../3200/ghettomodjn3.jpg

 

[aigomorla]

blah... the spacing on it wont even fit the normal 680i blocks.

im sending the board out to a custom builder and having him design some things up. When he starts posting i'll let you know.

Oh i think im gonna have him do the entire board mosfet and everything. Lopri remember that 780i isnt much different from the 680i. So its not a board that can take stress very lightly.

 

[lopri]

Thank you and keep me posted plz. I do know that 780i isn't much different from 680i, but 680i has been through many revisions, and so far it's been a lot better than 680i. One of the biggest concerns was disk/RAID controller, and that's what I am stressing right now. (2xRaptors RAID0 and 2x500GB RAID1)

 

[DerwenArtos12]

Originally posted by: aigomorla
blah... the spacing on it wont even fit the normal 680i blocks.

im sending the board out to a custom builder and having him design some things up. When he starts posting i'll let you know.

Oh i think im gonna have him do the entire board mosfet and everything. Lopri remember that 780i isnt much different from the 680i. So its not a board that can take stress very lightly.

Martin gonna do it for you? Using Aluminum? Acetal? Plexi?