Parallel GPU Blocks and temps simple test

[Hondacity]

Well a lot don't like parallel loops, I can't blame them, I myself questioned it. But to my amazement and curiosity. I've done a minnie testing on my new GPU blocks.



IDLE no load


One loaded and the Other idled


One Loaded and the other idled


Both Gpu's running Folding@home


Notes

The gpu's have almost the same load temps, pcb temps were different' one gpu had more airflow at the pcb.

I've seen series temps and the heat from one gpu transfers to another...i don't think thats the best setup.

on a side note:

these GTX285's are clocking really high........



Heatkiller's GPU WATERBLOCKS are HIGHLY RECOMMENDED

Enjoy

HONDACITY

[ecat]

Interesting results

Can you run a comparison with a series set-up ?

[air_ii]

I've seen series temps and the heat from one gpu transfers to another...i don't think thats the best setup.

With sufficient flow, i don't think that heat transfer is the issue. As water has high heat capacity, it doesn't heat up quickly enough for it to make a significant difference.

[Hondacity]

Interesting results

Can you run a comparison with a series set-up ?

wish i could, try sli is coming

With sufficient flow, i don't think that heat transfer is the issue. As water has high heat capacity, it doesn't heat up quickly enough for it to make a significant difference.

what is best then? series? sure one gpu is ok, lets say for example you have 1loop... cpu is a core i7, and you have tri sli, cpu first gets the cold water..dumps its heat to the next gpu, the next the next.

but yeah you have a point with the and series and that heat dump...how much? i don't know...is it better than parallel flow? NO

[generics_user]

parallel has the advantage of much lower restriction = more flow for the cpu block, additionally gpus are not was dependant on low temps as the CPU and the restriction is almost 100% identical for both blocks = same temps on both gpu cores

in series yu have 1 core heating the water for the second one = better temps on first core, worse temps on second core

with sli setups parallel flow is a much better option than series as all GPUS get the same cooling peformance = easier to get the same clocks on all cards

[Hondacity]

parallel has the advantage of much lower restriction = more flow for the cpu block, additionally gpus are not was dependant on low temps as the CPU and the restriction is almost 100% identical for both blocks = same temps on both gpu cores

parallel loops are 100% ok as long as the restriction and heat load of both loops are equal

dam post is way too slow, i'm still waiting for my X-res and HK 3.0

i think you're right with the restriction...the heatkiller285 block has a small inlet and outlet...but doubling the block sort of decreases the restriction..not necessarily doubling.

i disagree on the low temp dependency of gpus.....thats false...you can't get 810MHz on aircooling... let alone have it stable with aircooling...

[DRAGONKZ]

So whats the suggestion for Tri SLI? Try running in Parallel?

[generics_user]

i think you're right with the restriction...the heatkiller285 block has a small inlet and outlet...but doubling the block sort of decreases the restriction..not necessarily doubling.

i disagree on the low temp dependency of gpus.....thats false...you can't get 810MHz on aircooling... let alone have it stable with aircooling...

i meant that you won't notice any difference between different waterblocks (except for innovatek, i bet their blocks are worse than a decent air setup )

anyways great setup, once you get a third GTX 285 you might want to consider this watercool link part

[Sparky]

What don't you like about the MCR320? Seems to be doing a fine job

[veedubfreak]

Im running TRI-SLI with 3 koolance blocks on their own loop. At load i get around 45c, but i probably need more radiator. A single swiftech 320 is probably at its limits cooling 3 285s.
Running series just for ease of installation.

[Hondacity]

i meant that you won't notice any difference between different waterblocks (except for innovatek, i bet their blocks are worse than a decent air setup )

anyways great setup, once you get a third GTX 285 you might want to consider this watercool link part

i already did, its coming friday

What don't you like about the MCR320? Seems to be doing a fine job

its not an mcr420, even koolance and xspc are first to make a 480 rad, and the damn thing needs high speed fans to compensate, i already have 2 480gtx for the replacement...and quiet fans for it. MASSIVE IMPROVEMENt

Im running TRI-SLI with 3 koolance blocks on their own loop. At load i get around 45c, but i probably need more radiator. A single swiftech 320 is probably at its limits cooling 3 285s.
Running series just for ease of installation.

45c...thats ok...i'll post my temps with tri-sli when its done...i'll take note of that 45c and see how it compares to my hk tri-sli setup

[SNiiPE_DoGG]

well thats not what everybody expected.... good stuff though now everyone can run parallel for SLI and get less restriction!

[NaeKuh]

wish i could, try sli is coming

what is best then? series? sure one gpu is ok, lets say for example you have 1loop... cpu is a core i7, and you have tri sli, cpu first gets the cold water..dumps its heat to the next gpu, the next the next.

but yeah you have a point with the and series and that heat dump...how much? i don't know...is it better than parallel flow? NO

Honda it takes roughly 300W of heat @ water moving 1.5gpm for it to go up 1C.

So on average systems where the flow is about .9-1.2gpm, your looking at roughly 250-275W of heat required for h2o to go up 1C.

You sure u dont want to retract your statement?

Paralell doesnt work for high pressure/high flow systems. Its been tested over and over and over again.

[BlueAqua]

Great test. I'm about to install a parallel gpu setup myself for ease of plumbing. I figure as long as you use the same gpu block then they'll have equal restrictiveness and should have equal flow.

[SNiiPE_DoGG]

Paralell doesnt work for high pressure/high flow systems. Its been tested over and over and over again.

he's got a single loop with presumably a DDC running and 2 gpu blocks.... that seems High flow enough for me to say that parallel is perfectly fine. I said parallel was bad before too but I admit now I think I was wrong.

[Hondacity]

Honda it takes roughly 300W of heat @ water moving 1.5gpm for it to go up 1C.

So on average systems where the flow is about .9-1.2gpm, your looking at roughly 250-275W of heat required for h2o to go up 1C.

You sure u dont want to retract your statement?

Paralell doesnt work for high pressure/high flow systems. Its been tested over and over and over again.

first you talk about 300w and 1.5gp and 1c, radiators? and fans?....for all i know thats fud not fact.

high pressure? high pressure and low flow?
high flow? high flow and low pressure?

which one?

[Hondacity]

Great test. I'm about to install a parallel gpu setup myself for ease of plumbing. I figure as long as you use the same gpu block then they'll have equal restrictiveness and should have equal flow.

yep if they're the same, i hope yer build is good like the rest your buids...

[NaeKuh]

first you talk about 300w and 1.5gp and 1c, radiators? and fans?....for all i know thats fud not fact.

high pressure? high pressure and low flow?
high flow? high flow and low pressure?

which one?

Uhhhh.... those numbers is straight from physics. If you think im wrong do the math out.

And wow you wana try to get smart with me?

The specific heat of water is 4.184joules/gram-C

Q = mc(DeltaT)

Q = 1,000g(4.184j/g-C)(1C)

Q = 4,184 joules per 1 degree C rise in temperature

A joule is a watt-second so:

4,184 watts will heat a liter up by 1C every second.

2,092 watts will heat a liter up by 1C every 2 seconds.

1,046 watts will heat a liter up by 1C every 4 seconds.

This is with 100% efficiency. In reality, there will a small amount of heat loss to ambient. Minimize your exposed surface area to minimize losses.

Roughly translates when using a different forumla that my numbers were correct.

BTW martin and about 2/3rds of my original TEC thread posters came up with that number as well. Math is way more important when messuring Qmax value when your TECing.

And im gonna ignore your last comment, because you should know what i mean about high pressure and high flow.

If you cant take the criticism dont test, why do you think i quit?
If you think im wrong, then prove me wrong with a complete testing using flow meters and a nanometer to show your head pressure and flow variance and how they scale.

I admit im human and i can be wrong, but id like to be shown when wrong.

[DarthBeavis]

math is for n3rds.

[nikhsub1]

I've seen series temps and the heat from one gpu transfers to another...i don't think thats the best setup.

How do you know? With only data from parallel not comparing to series = test and thread FAIL.

/end thread.

[Hondacity]

math is for n3rds.

lol

How do you know? With only data from parallel not comparing to series = test and thread FAIL.

/end thread.

thanks NOT

[SNiiPE_DoGG]

I thought the point of this thread is to show that parallel works fine and temps on each card are the same. just that is enough for me to say yeah P beats S because of the restriction reduction and the ease of plumbing

[Hondacity]

Uhhhh.... those numbers is straight from physics. If you think im wrong do the math out.

And wow you wana try to get smart with me?

The specific heat of water is 4.184joules/gram-C

Q = mc(DeltaT)

Q = 1,000g(4.184j/g-C)(1C)

Q = 4,184 joules per 1 degree C rise in temperature

A joule is a watt-second so:

4,184 watts will heat a liter up by 1C every second.

2,092 watts will heat a liter up by 1C every 2 seconds.

1,046 watts will heat a liter up by 1C every 4 seconds.

This is with 100% efficiency. In reality, there will a small amount of heat loss to ambient. Minimize your exposed surface area to minimize losses.

Roughly translates when using a different forumla that my numbers were correct.

BTW martin and about 2/3rds of my original TEC thread posters came up with that number as well. Math is way more important when messuring Qmax value when your TECing.

And im gonna ignore your last comment, because you should know what i mean about high pressure and high flow.

If you cant take the criticism dont test, why do you think i quit?
If you think im wrong, then prove me wrong with a complete testing using flow meters and a nanometer to show your head pressure and flow variance and how they scale.

I admit im human and i can be wrong, but id like to be shown when wrong.

i'm tired, hey it works. simple test.

about quitting? it never crossed my mind

[Hondacity]

I thought the point of this thread is to show that parallel works fine and temps on each card are the same. just that is enough for me to say yeah P beats S because of the restriction reduction and the ease of plumbing

[NaeKuh]

i'm tired, hey it works. simple test.

about quitting? it never crossed my mind

Honda your one of the few people i actually respect when it comes to testing.

But to a tester he needs to listen to criticism. And the way a tester pwns the critique is with raw observed numbers so there is no room for arguement.

When have you seen me complain about a non valid testing point to martin/skinnee or gabe/stephen?

Im just telling you what ive seen in the 9yrs ive been watering.

I thought the point of this thread is to show that parallel works fine and temps on each card are the same. just that is enough for me to say yeah P beats S because of the restriction reduction and the ease of plumbing

no to validate the test we need to see how they scale, and what the difference is between paralell and serial on his own loop records as scott mentions.

Dude im not going to let another half completed test slide on this forum. Im honestly getting tired of it.

I honestly expect about this much from honda, as i said he is one of the few i see a lot of potential in.