Hondacity's HeatKillers GTX285 in Triple Sli (simple test)

[Liu Kang]

blah blah blah

Good looking block with good results... sounds like a win/win scenario to me.

Nice work, Honda.

[justin.kerr]

That depends mate... with medium speed Yate Loons and an MCR320 you are looking at around 13-15C delta between water and air temp loaded. While that is more than cool enough for GPUs (calculate air temp + 15 = water temp + 20 for OCed GPUs = GPU temp), it's not great temps. If you just want to silence them and don't care if they run at 50 or 60C, it's enough. Otherwise get a GTX 480/Feser/PA120.3 or equivalent.

I see, lol, When I seen Honda's dual quads for tri GPU's, I knew it might be a little overkill, but maybe not.. Thanks for the input !!

[shazza]

I see, lol, When I seen Honda's dual quads for tri GPU's, I knew it might be a little overkill, but maybe not.. Thanks for the input !!

The results I posted were with 3 285s on a PA120.3. Temps are fine for all practical purposes ... both at stock and even with the cards overclocked. You will not really be loading the cards at 100% during most gaming or other sessions. Even 50 degrees ... or 60 ... or 70 will be better than you can do on air, and the cards will run fine. So, I'd say give it a whirl with the MCR320.

[sidewindercomp]

and whatever agenda there is with Sidewinder Computers, so be it.

Dear IanY,
If my company or I have done something to offend you, please accept our apologies. My small staff and I work very hard to keep our customers happy and carry as many parts as customers request. If we messed up somewhere and did not fix it, please contact me here. I take full responsibility for whatever mistakes were made.

If this is not the case, I'd appreciate an explanation as to what you are implying or accusing my company of doing?
Gary Stofer
Sidewinder Computers

[septim]

basically this thread shows people that parallel setup can work. it just depends on what you're using...

[Erasmus354]

basically this thread shows people that parallel setup can work. it just depends on what you're using...

It also adds further reinforcement against the "Card 2 will get no flow!" myth that certain people have been pushing in regards to this type of tri-sli parallel setup.

Keep in mind however, that when you split a loop 3 ways each part of the loop (in this case each card) will get 1/3 of your overall flow. So you need to have a pump with good flow capacity (head pressure is less important in these setups because you are also making the resistance much much less than normal by going parallel). So there are certainly many things you still need to be careful of, just that you don't need to worry about each card getting different amounts of flow.

[skinnee]

Ummm I was actually trying to add some humor! Man everyone is so stiff tonight. What gives?!? Sorry Gary it was not meant as an insult at all.

I still don't know how to pronounce Naekuh right.

Seriously...help me sound it out.

Nay-koo
Nay-ka
Ni-koo
Ni-ka
rama-lama-ding-dong

[SNiiPE_DoGG]

I always said Nak-youuuuuu

[b@llz0r]

Keep in mind however, that when you split a loop 3 ways each part of the loop (in this case each card) will get 1/3 of your overall flow. So you need to have a pump with good flow capacity (head pressure is less important in these setups because you are also making the resistance much much less than normal by going parallel).

Thats a good point.... so the MCP655 would be preferable for 3 way over the MCP355

MCP355
head (@ 12 V) 15 ft (4.5 m)
discharge (@ 12 V) ~ 120 GPH (454 LPH)

MCP655
head 10 ft (3.1 m)
discharge ~ 317 GPH (1200 LPH)


so many tests to perform so little time

[Zehnsucht]

What I'd like to see is the following test:

And noting the flow when the resistance is equal for valves, squewed etc.

If I manage to buy a second pump used I might consider doing this myself.

I think that the parallel topic is quite interesting, and there is certainly not enough testing. That's why we are having these discussions.

[skinnee]

Zehnsucht, your diagram is similar to what I have planned for testing.

[jcool]

I still don't know how to pronounce Naekuh right.

Seriously...help me sound it out.

Nay-koo
Nay-ka
Ni-koo
Ni-ka
rama-lama-ding-dong

I pronounce it Nä - koo but that's probably totally wrong

[Sadasius]

I still don't know how to pronounce Naekuh right.

Seriously...help me sound it out.

Nay-koo
Nay-ka
Ni-koo
Ni-ka
rama-lama-ding-dong

?!? Oookay! Help me out I don't get it!

***EDIT*** Okay got it now! Thanks Skinnee!

But yeah to keep this on topic it does show that parrellel blocks are a viable thing to look into obviously. Just need comparison runs now between parallel and serial to see what is easier to setup and which is more beneficial.

[Zehnsucht]

Zehnsucht, your diagram is similar to what I have planned for testing.

Awesome

[IanY]

Dear IanY,
If my company or I have done something to offend you, please accept our apologies. My small staff and I work very hard to keep our customers happy and carry as many parts as customers request. If we messed up somewhere and did not fix it, please contact me here. I take full responsibility for whatever mistakes were made.

If this is not the case, I'd appreciate an explanation as to what you are implying or accusing my company of doing?
Gary Stofer
Sidewinder Computers

Dear Mr. Stofer,

I prefer to respond to you publicly in open forum if you don't mind.

I don't have any greivances against your company, nor do I harbor ill will against yourself personally.

You are an excellent retailer whom I would recommend.


With regards,
Ian Yeoh

[Vapor]

Thread cleaned...let's keep the crap, the attitudes, and the off-topic BS out of XS guys.

[weston]

The problem I see with a high flow parallel setup is that the flow is still relatively low in the cards, but very high in the radiator giving less contact time with the fins to cool down. Sure it works fine with 2xquad rads, but wouldn't anything? The test I would most like to see is a direct comparison between parallel and serial: Same pump, rads, fans, water blocks, etc... except for the way the tubing is connected.

[Erasmus354]

The problem I see with a high flow parallel setup is that the flow is still relatively low in the cards, but very high in the radiator giving less contact time with the fins to cool down. Sure it works fine with 2xquad rads, but wouldn't anything? The test I would most like to see is a direct comparison between parallel and serial: Same pump, rads, fans, water blocks, etc... except for the way the tubing is connected.

It doesn't work like that. In fact, the radiator performs better with more flow. Just go look at any of the testing charts. The Watts the radiator can dissipate at any given deltaT goes up with flow rate (not by much but it does). This is because it helps lessen the boundary layer between the tubing and the water.


Sure if you slow down the flow rate you will lower the temperature of the water more from rad in to rad out. However you are also increasing the temperature RISE in the waterblocks by slowing the flow down. The Watts dissipated is still going to be the same either way (it always has to equal the Watts put off by the components). Therefore you want to get the most efficient heat transfer, which is with more flow.

Hypothetical Example:
Flow 1: 2GPM
Air Temp : 23C
Rad In: 33C
Rad Out: 32C
Because of fast flow the system can dissipate the heat put off with a 10C deltaT between water and air. However the fast flow results in only 1C temperature drop.

Flow 2: 1GPM
Air Temp: 23C
Rad In: 35C
Rad Out: 33C
Because of the slower flow the radiator needs a higher 13C deltaT between water and air to dissipate the same amount of heat. Notice that the difference between rad in and out is higher but overall water temperatures are higher as well to compensate for the loss of efficiency in the radiator with lower flow.


Now those numbers are greatly exaggerated. Water flow is a relatively minor component in the performance of a radiator, but you get the idea.


Really you shouldn't think about things in terms of change in temperature across a given component (radiator or waterblock). Think about things in terms of Watts dissipated and delta between air and water temperature.

[vinister]

It must be almost May. Every 3 months, like clockwork, this forum sees a series vs. parallel debate.

I just had to post here because if the thread has the words parallel and Naekuh in it an alarm goes off next to my pc and I get a text message and an email reminding me to get involved.

Anyways I cool my 3 gpu's on just one double rad, load temps are 40-50C (they're old 3870's), in a similar parallel arrangement. Parallel rocks, but if there is one thing Naekuh is right about, its that the inlet and outlet should be on opposite sides of the setup, because of the effects of the velocity of the water and the inertial forces involved. Either way, its a matter of a couple degrees, which can easily be due to differences in chips, TIM, mounting pressure, etc etc.

Yay for another parallel success thread!

[Erasmus354]

It must be almost May. Every 3 months, like clockwork, this forum sees a series vs. parallel debate.

I just had to post here because if the thread has the words parallel and Naekuh in it an alarm goes off next to my pc and I get a text message and an email reminding me to get involved.

Anyways I cool my 3 gpu's on just one double rad, load temps are 40-50C (they're old 3870's), in a similar parallel arrangement. Parallel rocks, but if there is one thing Naekuh is right about, its that the inlet and outlet should be on opposite sides of the setup, because of the effects of the velocity of the water and the inertial forces involved. Either way, its a matter of a couple degrees, which can easily be due to differences in chips, TIM, mounting pressure, etc etc.

Yay for another parallel success thread!

Inertia doesn't really play much of a part in closed loops. It is more about pressure differentials.

[veedubfreak]

Nice Results!

So, after seeing this cooling setup,I was wondering if a Swiftech mcr320 with Medium Yate Loons, would be enough to cool the 3 285's?? I ask because I am putting my 3 under water next week.... Damn they get hot, with stock cooling and over clocked in tri-sli...

I'm running tri-285s on that exact setup. It's more than enough. You'll run out of voltage long before heat becomes the issue with your overclock.

[b@llz0r]

Inertia doesn't really play much of a part in closed loops. It is more about pressure differentials.

music to my ears, you obviously know a thing or two about fluid mechanics

Got any data on your parallel setup?

[b@llz0r]

heres one i did in July last year



temps on GPU 2 were higher than the others until i figured out that the sli fittings needed spacers, the threads were too long and there was only a gap of a few mm between them

[Hondacity]

let me give you 25 elbows to see if Inertia doesn't hurt flow #'s,

in a full paralle loop yes pressure matters but with mixed its not a issue

wow 25!!! 90 elbows?

i only have 10

[Sadasius]

What I am thinking is happening in the GPU blocks with parallel flow is completely different then what the blocks were designed to do. The GPU blocks were designed to have flow go through them but with parallel I don't think there is actual flow as we normally see it but is happening in a different way. It's too bad we do not have some sort of visual pinwheels on two sides near the ports on the GPU blocks to confirm this but what I think is happening is that the water is being exchanged at the ports of the blocks and are not really circulating in the block itself as much as it is just exchanging slowly off of the lip of the ports. When you stop to think about it and the results show it too is the temps are consistent through the 3 blocks. The flow is going the path of least resistance down the tubes but is scrapping off fluid along the way and is exchanging that way. So the water in the blocks is kinda sitting there exchanging a little at a time slowly but enough not to overheat the water and I think there is even some protection there as the water heats up and expands into the running water in the ports. I don't know....Just ramblings on my part and may mean nothing.