2 x D5s - increase loop temp.. interesting...

[Phatboy69]

Just been doing some testing with the new Aquaero controller and noticed that with one pump on each loop at safe minimum run voltage of 7.2V it decreased water temp by about 1c in each loop but increased GPU/CPU temp between .5c and 1c. So the blocks/pumps are either dumping less heat or rads are removing more heat or maybe both.

Thats with .6 Gal/min CPU and .4 Gal/min GPU compared to 1.5 Gal/min CPU and 1.1 Gal/min GPU at max speed.

This is with idle temps BTW

Will be testing load temps to see what happens tomorrow and checking if CPU/GPU actually run cooler or hotter.

Very interesting. There might be a sweetspot somewhere in the flow vs heatdump tradeoff to gain a couple degrees better cooling.

[Waterlogged]

Sounds like your really close to the thermal limit of the loop if the pumps are affecting temps like that. IIRC, each D5 dump ~22-24W of heat into the loop at full speed, that's 44-48W total. Are you changing the fan speeds as well when you turn the pumps up? Does it drop the temps down again when you do? If not, you might want t think about better fans (Gentle Typhoons) or even more rad.

[Phatboy69]

Sounds like your really close to the thermal limit of the loop if the pumps are affecting temps like that. IIRC, each D5 dump ~22-24W of heat into the loop at full speed, that's 44-48W total. Are you changing the fan speeds as well when you turn the pumps up? Does it drop the temps down again when you do? If not, you might want t think about better fans (Gentle Typhoons) or even more rad.

This is with fans set to max at 1750RPM.
I dont think its a thermal limit problem its just a fact that 2 D5s dump a lot of heat into a loop and the science of thermodynamics. Im sure the increase would be more noticeable with fans at idle speed.

I can see from the graphs the controller creates that the return water temp for GPU goes up by 2c-3c while the cold side actually goes down by about 1c. So the loop is actually taking more heat from the GPU at the lower flowrate with less pump heat going into it too. But the chip temp actually goes up by 1c because of the low flowrate. Its just interesting to learn how this all works.

At the end of the day it looks like performance of the loop is determined by component temp rather than minimum water temp. More flow = higher ambient water temp but lower component temp within the sweetspot. Go outside the sweetspot and your probably either taking too much heat from the component per gal/min or too little heat and not enough time for the radiators to have time to cool it. Guess thats why car engines have thermostats eh!

[NaeKuh]

no ive been telling you guys this!

D5's dump a TON of heat back into the loop.

How come no one believes me?

Trust me i was really shocked at the results too when i threw on the dual D5's vs dual DDC's.

At the end of the day it looks like performance of the loop is determined by component temp rather than minimum water temp. More flow = higher ambient water temp but lower component temp within the sweetspot. Go outside the sweetspot and your probably either taking too much heat from the component per gal/min or too little heat and not enough time for the radiators to have time to cool it. Guess thats why car engines have thermostats eh!

no.

Minimum water Temp = Floor = the Lowest possible u can go b4 u break science.
More flow = higher holding capacity of water which means you can carry more heat per pass b4 your water goes up 1C.
More flow = always good.. your get diminished returns as you reach the end tho.. but u still get gains.. never a loss.

uhh.. no.. your not thinking in the right concepts of thermodynamics... there is no such thing as "sitting time in radiator" in a closed system.

[Phatboy69]

no ive been telling you guys this!

D5's dump a TON of heat back into the loop.

How come no one believes me?

Trust me i was really shocked at the results too when i threw on the dual D5's vs dual DDC's.



no.

Minimum water Temp = Floor = the Lowest possible u can go b4 u break science.
More flow = higher holding capacity of water which means you can carry more heat per pass b4 your water goes up 1C.
More flow = always good.. your get diminished returns as you reach the end tho.. but u still get gains.. never a loss.

uhh.. no.. your not thinking in the right concepts of thermodynamics... there is no such thing as "sitting time in radiator" in a closed system.

Im not sure I agree completely with you. mostly though..

My testing revealed that both the rads and block were removing more heat at a lower flow rate because the delta Tin/Tout was 5c with a lower Tin compared to a higher flowrate with a delta Tin/Tout of only 1c at a higher Tin water temp. But the lower Tin didnt translate to the lowest GPU temp because flow was too slow.

When I post some graphs tonight you'll see what I mean.

[NaeKuh]

Im not sure I agree completely with you. mostly though..

My testing revealed that both the rads and block were removing more heat at a lower flow rate because the delta Tin/Tout was 5c with a lower Tin compared to a higher flowrate with a delta Tin/Tout of only 1c at a higher Tin water temp. But the lower Tin didnt translate to the lowest GPU temp because flow was too slow.

When I post some graphs tonight you'll see what I mean.

then your data has errors. (meaning hardware is not calibrated properly or within testing range scope)

Because thermodynamically speaking...

More flow = more passes at radiator.
Closed loop works on a principle of equalibirum, there is no hot / cold side in a closed loop, unless u have extreme heat loads on one end.

And usually when you have more flow, you can carry more heat, and once again, more passes at the radiator = better results.

What your speaking is not thermodynamically possible

[Martinm210]

Heat is a result of two things. For one the DDC series is much more electrically efficient:


Followed by the metal vs plastic pump housing. Metal transfers heat very well (D5 series), and plastic does not(DDC series):
DDC is plastic (Insulator)


D5 is metal (Conductor)


Good news is you can pretty much insulate a D5 without any worry about heat buildup. Also the D5 is much happier at higher flows than the DDC series for heat.

But as far a heat dump, D5's do dump quite a bit more heat into the loop.

[NaeKuh]

and in that infomercial no D5 or DDC's warrenty was voided.

lol....

[shazza]

no D5 or DDC's warrenty was voided

I think we better send you bak to kollige for sum more of that educimication you missed



Back on topic ... I knew the D5s threw out more heat, but didn't realize it was this much.

[NaeKuh]

Back on topic ... I knew the D5s threw out more heat, but didn't realize it was this much.

no i was seriously shocked...

This emote would not be enough... followed by WTF i should of just went all DDC's!!!!

You can ask Vapor i was telling him WTF! while he was laughing at me going Duh.

note i had 2 DDC's i could of toss'd at any given time in my loop but i could only afford 2 bay res... and no i dont want to spend any more money on my blackhole of a wallet..errr unfinished build.

[Phatboy69]

Here's an interesting conversation from Hardwarecanuks that supports my theory of flowrate vs time in Rad cooling... higher the flow the less time for cooling. As I said thats why cars have thermostats to regulate flow rate in a radiator or cars would overheat.. replace engine with CPU in our setup and you have the exact same formula. Eventually heatload runs away from you. Thats exactly what my testing demonstrated.

"Well said Scruff. Dazmodes chart is scaled to look impressive but check the data on the chart, it says a 400% increase in flow rate equals a 6% drop in temps.

What happens over time? The high flow rate coolant does not have time to get cooled in the rad and the temps creep up.

When we removed the thermostat in our race car we had to put a restrictor plate in or the car overheated in no time."

http://www.hardwarecanucks.com/forum...errated-3.html

Some other interesting facts I found about the place..
Maximum flow velocity over gpu and CPU block and minimal flow velocity over heat exchanger (Rad) generates the best results.
So you need high velocity restrictive high pressure at the CPU/GPU and unrestricted low pressure low speed through heat exchanger. (RAD)

Parallel flow rads would appear to be the winner here because that would slow the velocity in half and create low pressure! Might have to look into how I convert my series rads into parallel.

[Metal Wolf]

I though this line of thought was killed dead back in the old little river whitewater / danner mag7 days.

If there is water in the radiator, then the radiator is radiating heat at the rate it will radiate heat at. It is not related to the flowrate of the water through it.

Small exception in that if you slow down the flowrate enough that you get a wonky delta then you start soaking out huge amounts of heat into the air. However your parts are general well overheated by that point so its not good to do on computers.

If you dont believe me thats fine but this is something thats been long dead.

also, parallel radiators do benefit from a reduced flow rate. That part is correct mostly, the pressure drop thing is well no idea where thats coming from.

[Phatboy69]

I though this line of thought was killed dead back in the old little river whitewater / danner mag7 days.

If there is water in the radiator, then the radiator is radiating heat at the rate it will radiate heat at. It is not related to the flowrate of the water through it.

Small exception in that if you slow down the flowrate enough that you get a wonky delta then you start soaking out huge amounts of heat into the air. However your parts are general well overheated by that point so its not good to do on computers.

If you dont believe me thats fine but this is something thats been long dead.

also, parallel radiators do benefit from a reduced flow rate. That part is correct mostly, the pressure drop thing is well no idea where thats coming from.

The higher flow rate through the rad means it is working at a lower efficiency.
The rad has much high cooling capacity than I can utilise due to flow rate being too high and the T deltas are also lower than what you see in the most efficient heat exchanger setups.

I believe there will be an optimal flow rate for heat extraction and heat rejection and that will vary from system to system.

What I intend to start testing tonight is the maximum T Delta water in/out and GPU T Delta to water I can achieve at different flow rates and with 1 or 2 D5s running.

[Church]

Phatboy69: no, it's one of false myths of liquid cooling, that rad will cool better if water flows through it slower. As skinneelab rad test curves at different flow speed prove - rads benefit from higher flow too, albeit a bit less then waterblocks and imho minimum efficient enough flow level is less.
In this case it's simply problem of extra pump's additional heatdump into loop offsetting benefits of higher flow. If there were more rads in loop, then that extra heat dump would be relatively less to rad capability, yet all those rads would benefit from higher flow, thus one should see temp gains, though small ones (diminishing returns).
With slower or faster flow water still spends same time in component in closed loop (double flow, and while water will pass in half of the time, it will do that twice as many times in same timeframe), but with slower flow there also is less turbulence and thus less effective heat transfer to/from water, be it in WBs or in rads. So there is no optimal maximum flow speed per rad or per waterblock. Just diminishing returns from flow above 1gpm that is more then offset with pump heat dissipation in water. With sufficient rad area and not overexagerated pumping power not that much of an problem.

[Waterlogged]

Maybe we should reference the "2 race cars going different speeds on the same track" analogy for the 2,000 time?

If you have 1 car traveling at 50MPH and another traveling at 100MPH, the car that's going faster will make 2 laps for every one the slower car makes. Same thing with the water, the faster flowing loop will pick up less heat per cycle but it will make 2 cycles and therefore pick up the same amount of heat in the same timespan.

[TJ TRICHEESE]

^^Yea exactly what i was going to say, the water spends the same amount of time in each component regardless of the flow (as long as there is some flow and that flow is constant)

[Martinm210]

Just look at the facts..more flow is better and moreso for stronger fans...


You do have to watch where youb temperature sensors are. The average water temperature is different than rad water out only. Lower the flow the higher the in out delta, so it easy to confuse.

[Phatboy69]

Well I did some more testing and looks like even though i get lower Water Tin with a lower flow rate I get a very slightly higher temp on the GPU and CPU. Doing the same test under load conditions and the Tin/out delta is even greater and the GPU runs 2c hotter while the CPU is only about 1c more at lower flow.

The graphs show 5 mins normal idle temp, 5mins with 1 D5 off in each loop and 5 mins with all 4 D5s at max.

So the moral of the story is, as you guys have been saying, more flow is king. The D5s obviously dump about 1c into each loop but they make up for it by improving heat extraction from the block.

FYI my D5s are before the Rads on the hot side.

I will still try running the rads in parallel another time to see if I can squeeze some more out of it.

[NKrader]

no ive been telling you guys this!

D5's dump a TON of heat back into the loop.

How come no one believes me?

Trust me i was really shocked at the results too when i threw on the dual D5's vs dual DDC's.



no.

Minimum water Temp = Floor = the Lowest possible u can go b4 u break science.
More flow = higher holding capacity of water which means you can carry more heat per pass b4 your water goes up 1C.
More flow = always good.. your get diminished returns as you reach the end tho.. but u still get gains.. never a loss.

uhh.. no.. your not thinking in the right concepts of thermodynamics... there is no such thing as "sitting time in radiator" in a closed system.

so your saying. the more pumps you put in a loop the cooler it will be?

[Church]

NKrader: keyword is "diminishing returns". As you increase more and more flow temps improve less and less, yet heat dump from each extra pump is always added to coolant without a fail.
So in very resistive loop adding 2nd pump might improve things a bit, but adding even more so - at some point pros will be offset by cons.
That's why you often see mentioned 1gpm as nice flow to have. To increase it past that makes no sense financially and that pros/cons balance wise. It's more not "to have as much flow as possible", but rather "to have enough flow" (in my eyes ~ 0.8-1.1 gpm). Below 0.5gpm temps are hurt much more because of lack of flow, probably because at such numbers there is way less turbulence in flow - thus worse heat exchange in waterblocks and rads with heat transfer to/from water.

[RacingTurtle]

So you talk about the D5 dumping heat into the water, mostly due to the plastic of the housing being a poor conductor - do you think aftermarket housings would help with heat dissipation?
Such as the Bitspower Mod kit which is metal?

[NaeKuh]

no because the water itself is cooling the metal plate..

There is nothing u can do on a D5 to help disapate the heat unless u want to change the metal to plastic.
Then u dont have a D5, u have a crappy D5 that thinks he's a DDC.

[Alexandr0s]

no because the water itself is cooling the metal plate..

There is nothing u can do on a D5 to help disapate the heat unless u want to change the metal to plastic.
Then u dont have a D5, u have a crappy D5 that thinks he's a DDC.

A D5 with an identity crisis... Well I have to say, that's new !

But yeah, it has been said a billion times before, but there comes a point where the amount of heat dumped by the extra pumps is greater than the amount of heat dumped because of the higher flow. At that point, every pump you add is only there for the e-peen (not necessarily a bad thing, especially if you got big rads ).