Anyone got a 4870x2 on water - NOT FC?

[CedricFP]

As per title.

If so, which blocks did you use, and what cooling did you use for the bridge chip?

[RickCain]

Not a good idea as the mosfets need some active cooling.

[DeathWalking]

As per the regular 4870, the mosfet controller also needs a fair bit of cooling. Cool the mosfets to improve hardware life; cool the controller to prevent crashes. ;)

[CedricFP]

Yes I understand this. That's why I'm looking at other people's solutions.

FC blocks don't fit in with my long term financial plans. Being able to reuse independent blocks is a far more enticing idea.

[dasa]

im sure i have seen people using two mcw60 and just leaving the original heatsinks in place with a bit of airflow to cool the vrm and memory but you need to buy some new bolts for the block to fit

edit take a look here
http://www.xtremesystems.org/forums/showthread.php?t=198501

[Scubar]

You could maybe get away with two MCW60s and then use some heatsinks on the bridge and mosfets.

[Blkout]

im sure i have seen people using two mcw60 and just leaving the original heatsinks in place with a bit of airflow to cool the vrm and memory but you need to buy some new bolts for the block to fit

edit take a look here
http://www.xtremesystems.org/forums/showthread.php?t=198501


A couple of notes...

First and most important, Gabe only tested this setup using Furmark or ATT since he mentioned "furry cube" but then shows a screenshot of Furmark. I figure he got mixed up or he only tested one or the other. Either way, he was testing using a windowed application, not a full screen application, which means the x2 only uses one core which is hardly stressful to the core or the voltage regulators. He never once mentioned testing this in full screen with any games or other benchmarks for stability.

Second, in order to make the MCW60's fit, you have to cut the stock heatsink which means your warranty is voided and you've potentially and most likely crippled your resale value. Most people are not going to be interested in this option.

To sum everything up, no real stress testing was done and in order to even attempt this, you have to hack up your stock heatsink. I just can't see this being a very good option when FC blocks are available for around $100-$140. Again, I've yet to see anyone pull off a GPU only block setup on an x2. Once this happens, I'll take a look and re-evaluate.

[dasa]

isnt overclocking and changing cooling supposed to void your warranty anyway
and why wouldnt it work under full load? those heatsinks are all the stock card has as long as there is some airflow over it
on the up side gpu temps should be slightly better and the heat from the rest of the card isnt being dumped into the water loop

[NaeKuh]

u really want the FC block.

I really recomend the DD Block as well.

The FC Block does keep the card in uber check. Here are my full load temps on the card.
http://i125.photobucket.com/albums/p73/aigomorla/Loadtemps.jpg

ur spending 600 on the gpu and your gonna short it the extra 80 or so on the block? :shrug:

[fornowagain]

Get the FC block the card deserves it. Here's mine furmark loaded (800/915) with an EK block. Hardly gets over the loop temp.

http://img214.imageshack.us/img214/1628/96906030oh3.jpg (http://imageshack.us)

[DeathWalking]

Render Fur in fullscreen mode next time, windowed only loads one GPU.

[fornowagain]

That was fullscreen you can see that both core/memory clocks increased. Regardless, the temps hardly rise in game, the FC block is worth the money.

[DeathWalking]

Naw, you had Fur in window rendering. Do you know how I can tell? Look at the little graph for GPU load. One core was loaded fully, then went idle, while the other never got loaded. Not to mention the temps are a wee bit different between the two cores.

[fornowagain]

Naw, you had Fur in window rendering. Do you know how I can tell? Look at the little graph for GPU load. One core was loaded fully, then went idle, while the other never got loaded. Not to mention the temps are a wee bit different between the two cores.

Naw, that was fullscreen. Do you know how I can tell? Cos window rendering doesn't even change the 2d clocks.


http://img378.imageshack.us/img378/1056/23745261mq9.jpg (http://imageshack.us)



The core temps are always slightly different and so are the two stock Vgpu, one is +2C at least even at idle. Here's another just done, same again, doesn't 'appear' to load both cores (Vista64 8.53's).



http://img246.imageshack.us/img246/3169/54106301fi2.jpg (http://imageshack.us)


And yet another with the latest build, still think its windowed?


http://img99.imageshack.us/img99/1448/26329015bs4.jpg (http://imageshack.us)


Crysis does load properly, like I said it still doesn't get very warm.

http://img502.imageshack.us/img502/6448/87015573hd0.jpg (http://imageshack.us)


But hey thanks for calling me a liar.

[h20 frag-monger]

http://www.mattreeves.com/photo.jpg

[NaeKuh]

how come your VDDC current is 9.5A and my current is 23.7A

:shrug:

OH Nevermind.

The pics i showed was the card on full load, while yours is on idle. :rofl:

[evil-98]

the dtek looks awesome very nice though the angle thing is kinda weird for tubing routing

[Giannis86]

indeed the d-teks look nice there..however i would rather see a black unisink on that specific card as it has a black pcb.. :D

[DeathWalking]

Naw, that was fullscreen. Do you know how I can tell? Cos window rendering doesn't even change the 2d clocks.
The core temps are always slightly different and so are the two stock Vgpu, one is +2C at least even at idle. Here's another just done, same again, doesn't 'appear' to load both cores (Vista64 8.53's).
And yet another with the latest build, still think its windowed?
Crysis does load properly, like I said it still doesn't get very warm.
But hey thanks for calling me a liar.Many apologies dude...that's a weird bug. Again, sorry for doubting you, was expecting GPUZ to function correctly.

[NaeKuh]

forn wtf is your ambients?

im seeing temps on your card at 18C means your block must be around 15C-16C which means your coolant should be anywhere from 13-14C, which im guessing then would mean your RAD would need ambients at 11-13C. :O

@ those ambients i dont think you even need h2o cooling. :rofl:

Dayam thats cold.

My ambients were taken at 26.4C incase anyone is wondering, and held coolant at 29-30C. MCR220 /w low speed yates.

Forn, also i think you got a bad mount My card always has even temps on both gpu's

Which is why i stand behind my DangerDen recomendation. :D

[fornowagain]

Many apologies dude...that's a weird bug. Again, sorry for doubting you, was expecting GPUZ to function correctly.
Thats cool, no problem.

forn wtf is your ambients?

im seeing temps on your card at 18C means your block must be around 15C-16C which means your coolant should be anywhere from 13-14C, which im guessing then would mean your RAD would need ambients at 11-13C. :O

@ those ambients i dont think you even need h2o cooling. :rofl:

Dayam thats cold.

My ambients were taken at 26.4C incase anyone is wondering, and held coolant at 29-30C. MCR220 /w low speed yates.

Forn, also i think you got a bad mount My card always has even temps on both gpu's

Which is why i stand behind my DangerDen recomendation. :D
38C is that your idle?

No where near 11C, around 18C I should think. I wouldn't take max values from that 21C idle, they're a bit higher back in the trace, 3 loops of Crysis don't do much to my loop temps and the load temps drop off very quickly. Also I'm running quite a bit more capacity than you might think, 2xPA120.3 and 2xUltra pumps, 6x temp-dependent high cfm fans in an external radbox, good for 1.4kW. I have calibrated sensors for the rad air in and loop temps. Being a bit more methodical, why not. Just checking now room ambient is 21C, loop temp is 23.4C. GPU1=24C GPU2=22C so the second gpu sensor is under reading. I don't think its the block I've remount to check contact and it was perfect. Going by the higher one it's idling within 0.5-1C of the loop temp, I'll heat soak it for 10mins see what the temps are then.

EDIT: Ok what we got:- 10mins Furmark fullscreen. Ambient 21C, loop 27C GPU1max=33C (hold the mouse over the trace), went to 28/29C right off load. Settled back to the above idle 24/25C after 7mins.

http://img440.imageshack.us/img440/5976/93353707ev5.jpg (http://imageshack.us)

[NaeKuh]

heh you have a borked sensor 1!

:rofl:

I have a borked sensor 2!

Your other screenie has your temps dipping in way way cold territory. 16C min value on real temp, and 18C on the GPU. Thats why i was going OMG! Even with overpowered radiator, your cpu core temp can only go so low. :P

And my system is overpowered as well. :T
My CPU loop is major overkill. :T [RD-30] :D
I just chose to cheapo the gpu loop because its better then air, and a lot quieter. :D

[fornowagain]

heh you have a borked sensor 1!

:rofl:

I have a borked sensor 2!

Your other screenie has your temps dipping in way way cold territory. 16C min value on real temp, and 18C on the GPU. Thats why i was going OMG! Even with overpowered radiator, your cpu core temp can only go so low. :P

And my system is overpowered as well. :T
My CPU loop is major overkill. :T [RD-30] :D
I just chose to cheapo the gpu loop because its better then air, and a lot quieter. :D
Yeah you're looking at two under-reading sensors there. Realtemp values can be all over the place with my QX, those core DTS are mad. I go by the one that's highest to be safe (its does move though), 3&4 are within a few degrees of real world when I last setup Realtemp. Tbh its not very linear at the low range. Even taking that into account the cpu and gpu (higher one) idle a little above the loop. The radbox is pretty quiet, the fans hardly ever spin up in game. I couldn't stomach the stock 4870X2 fan for long, I edited the bios profile to keep reasonable temps, couldn't live with it.

[pur]

http://www.mattreeves.com/photo.jpg

Is this custom? I would love to get one of those heatsinks...

[NaeKuh]

no those are the d-tek's unisinks.

Most likely prototype right now. Should be coming out soon.