Overclock.net banner
1 - 20 of 116 Posts

·
New to Overclock.net
Joined
·
2,505 Posts
Discussion Starter · #1 ·
A few peeps have asked me to provide this info in an easily accessible / findable spot so the don't have to remember what thread it is in or take snippets from multiple threads. If ya find it useful, great .... if ya like something better, by all means use what ya comfy with. First off, I did not create any of the test data ... it's just not available in any one place so that's the only real help I can offer..... besides letting the spreadsheet "do the math" for all of us.

Step 1 - Heat Load Calculation.

a) CPU Load - For this I usually use this PSU calculator

http://extreme.outervision.com/PSUEngine

Plug in ya CPU, anticipated OC and voltage.... hit "overclock my CPU", thatz ya number....for me it was 135 watts. We don't wanna use the total PSU wattage, it accounts for capacitor aging and adds headroom.....just use the "overclock my CPU" subsection where it gives ya the wattage for CPU only

b) GFX Card Load - For this I usually use Guru3D.com's Power Consumption Data

Example: http://www.guru3d.com/articles_pages/geforce_gtx_780_review,9.html
Quote:
Subjective obtained GPU power consumption = ~ 231 Watts
Now I typically shoot for a 25% overclock .... Now at the very least, if ya wanna go 25% faster, ya gonna use at least 25% more power. But I remembered reading at some point that the way some sites recommended estimating OC'd wattage was to take the % speed increase and square it .... so if I wanna go 1.25 times faster, 1.25^2 was 1.56. But that seemed way to much especially w/ today's voltage limits so I usually go with 1.28 - 1.35 for a 25% OC.

231 x 1.28 = 296 watts or 592 watts for two cards.

c) RAM - I don't see any reason to cool RAM unless ya doing LN2 but if that's ya thing, I figure 3 -5 watts per stick based upon the above PSU calculator and and things I have read in various stickies over the years.

d) MoBo VRMs and Chipsets - Martins and most sites I have visited recommend 10 - 20 for each or 20 - 40 total.

e) Pumps - Most peeps forget about this item .... but ya can find the wattage for ya pumps on martins site

Example: http://martinsliquidlab.org/2012/01/29/swiftech-mcp35x2-pump/6/

Here you will see a max wattage of 51.52 watts for the twin pumps and 46 watts of heat.

Summary: For me it was 135 (CPU) + 592 (GFX) + 20 (MoBo VRMs) + 46 (Pump) = 792 watts

Step 1 - Heat Load Allocation.

Th next step I take is to "guestimate" what portion of the heat load is to be handled by the rads / fans. To my understanding, Martin specifically designed his tests to measure ONLY the cooling provided by the rad and fans.... even thermal radiation from the rad shroud thru the sides is excluded. And don't forget , heat is being radiated by all the tubing, fittings, exposed portions of water blocks and backplates, etc. ..... and not everything hits max load at the same time.

Now I'd like to claim some scientific method or keen insight was used but I basically came up with the numbers based upon reading all of peoples build results here on OCN..... looking at what the numbers said they should have and what they actually had ....and then looking at my own system (six temperature meters installed to measure temps at various points in the loop + ambient and case temps). At the end of the day .... it seemed for the most part, if ya had enough rad / fans to provide about 60% of the calculated heat load, you were doing OK. Now this is for the typical office application / gaming system and a higher number should be used for 24/7 folding, bitcoin mining or other similar uses. Suggestions welcome.

So in my case.... 792 x 60% = 475.2 so I went a lookin for 465 - 500 watts of rad cooling as a minimum

Would like some input here as Im sure 60% is low for a CPU and 60% is high for a GPU but the 60% average seems to work..... anyone who has rad in / out temps..... measured Delta Ts nor other relevant data, please post.

Step 3 - Rad Estimation.

All data based upon test results from martins liquid lab

http://martinsliquidlab.org/2012/04/12/alphacool-nexxxos-ut60-360-radiator/4/
http://www.xtremesystems.org/forums/showthread.php?287200-Flower-Labs-News-Comparison-of-Radiator-Cooling-Efficiency
http://martinsliquidlab.petrastech.com/Radiator-Fan-Orientation-And-Shroud-Testing-Review.html

Martin has a great set of data on his site for numerous 360 radiators including the Alphacool ST30, XT45 and UT60 each of which "tops the charts" from 1000 rpm on up. Xtreme systems had comparable data on the Monsta. As I found it difficult to try and convert the 360 to 240 and 480 equivalents each time I sized a system, I made a spreadsheet to "do the math" for me". Now I am by no means suggesting that all the data is, using that technical term from "My Cousin Vinnie" ..... "Dead-on balls accurate". It is however several orders of magnitude better than the "1 per heat producing component" rule of thumb.

The only data here which is "Dead-on balls accurate" is the data for the 360 rads (indicated by gray shading) as that is a direct quote of what is in Martin's charts. And while we all know that a 120 rad will not exactly be 1/3 of a 360, it's certainly the best approximation I can come up with....if we can accept that, then it's not hard to jump to the 240 being 2 x the 120 and a 480 being 4 x the 120. For the 140mm width rads, again, not "dead on balls accurate" but the numbers are a simple surface area allocation....essentially 140 x 140 / 120 x 120..... until we actually have a set of test data to work with on 140mm, it's the best idea I was able to come up with.

The Monsta data came from the xtremesystems site and was allocated in the same manner as above. Finally the fan data came from the results in Martins fan orientation test linked above. And yes, all data is based upon a "High End Cooling System" with a target Delta T of 10C. If your case and or budget can not sustain a system capable of doing 10C, then try the following:

-Divide the recommended amount of rad space for a high end system at 10C by 1.5 for a Delta T of 15C for a Mid to High End Cooling System

-Divide the recommended amount of rad space for a high end system at 10C by 2.0 for a Delta T of 20C and Mid Range Cooling System

-Divide the recommended amount of rad space for a high end system at 10C by 2.5 for a Delta T of 25C for a Mid to Entry Level Cooling System

-Divide the recommended amount of rad space for a high end system at 10C by 3.0 for a Delta T of 30C for a Entry Level Cooling System


Again, I sat down to design my loop looking for 60+% of the heat load in push / pull with 1250 rpm fans .... taking that 500 watt target and dividing it by 85 watts (140mm UT60 @ 1250 rpm), it would seem that I'd need about 5 x 140mm worth of Rad which was perfect for the Entrhoo Primo case I had and the numbers said I'd get 64% of my 475 watt target.

But I wanted to see how well I'd do with just one set of fans (53% of load) and I'm doing just fine with a 8.4C delta T at 1200 rpm and case filters removed....and 12-14C with fans capped at 850 rpm and filters back in. At this point, the only reason I think will make me add fans is curiosity. GPUs peak at 39C under Furmark @ 1200 / no filters and 44c with filters and 850 rpm cap. RoG Bench brings the CPU to 74C at 4.6C tho a new BIOS upgrade has currently left that unstable at my original voltages and haven't decided whether to re-tweak or roll back. Adding a 3rd 240mm rad would in push would take me to about 80%

So again, I hope peeps find it useful, and this will give me something to refer to instead of retyping snippets when needed to answer peeps questions.

The resultant spreadsheet has the aforementioned data including:

a) Watts Allocation for 120, 240, 360, 480, 140, 280 and 420 radiators for fan speeds of 1000, 1250, 1400, 1800 and 2200 rpm
b) Heat Load Generator
c) Heal Allocation Table
d) Case Air Flow table
e) Miscellaneous Data

AquacoolEstimator.XLS 38k .XLS file


Please advise of any errors or confusion and I will try and get fixed ASAP.

XSPCEstimator.XLS 33k .XLS file


EKEstimator.XLS 29k .XLS file


SwiftechEstimator.XLS 29k .XLS file


HardwareLabsEstimator.XLS 32k .XLS file
 

Attachments

·
Banned
Joined
·
12,731 Posts
That is AWESOME JackNaylorPE. +Rep

While I'm not using Alphacool's anymore it does give me a jump off point/estimate for some other radiator types (namely a 30mm 480 radiator).

So it looks like my build will be able to roughly dissipate 291 (30mm 480 P/P) + 219 (60mm 360 P/P) + 125 (60mm 240 P) or 635 watts of heat (using 1200 RPM fans).

That should be good enough for a 3770K and 2 GTX 780's, all overclocked.
biggrin.gif
 

·
Premium Member
Joined
·
4,898 Posts
Great guide, with great being the emphasis! Saving this thread for sure. I'd even go so far as to say this should be Sticky'ed
 

·
Banned
Joined
·
12,731 Posts
Quote:
Originally Posted by morencyam View Post

Great guide, with great being the emphasis! Saving this thread for sure. I'd even go so far as to say this should be Sticky'ed
If it got expanded to include other radiators, like say the XSPC, Phobya, Koolance and Swiftech's (to name a few), that would be a truly awesome addition.
 

·
New to Overclock.net
Joined
·
2,505 Posts
Discussion Starter · #5 ·
Well......

1. Martin and ExtremeSystems did all the hard work..... I just did the math.

2. The main reason I did it was just about every sticky has the "rule of thumb" and while adequate and useful as a "rule of thumb", it just doesn't cut it to do any real planning. I just couldn't get past using the same rules for a 133 watt and a 333 watt GPU .....especially when that differences gets compounded 2,3 and 4 times with SLI/CF. Didn't wanna fions myself picking case based upon the rule of thumb and then finding out my rads and case were undersized.

3. OK, you nudged me .... the reason that I did the Alphacool line was that a) ....martin had results for 3 of the 4 models and b) they were at the "top of the charts"......I'll take a look at what's available for what and see what I can do.

4. I create a lot of spreadsheet based 3rd party tools for a game called "The Saga of Ryzom" ..... it's a 10 year old MMO that for whatever reason has an average player age of 35+. I never was into multiplayer anything before of since and started it only because my then 11 year son was down in the dumps cause he didn't get into WoW. My company did a lot of beta testing for memory managers and technical programs and I used the resume to get "me" into the SoR alpha / beta figuring once "in", I'd turn it over to him .... well he played for about 6 months and 10 years later, I'm still popping in now and then. But I ran outta tools to make for them and so I need another project
smile.gif
. So if ya have specific suggestions or know of other data sources than Martins, please advise.

** BTW, if ya looking for a social game with a great community w/o a lot of tweenage testoserone driven behaviors, Id suggest giving SoR a try.....free till level 125 ..... send a private message to FyrosFreddy , Misatey or Azad when ya arrive ..... they are neutral faction payers who can give ya clue in to the factions, what they are about etc. and what guilds might fit ya play style.**

5. If anyone has any insight on why "everything seems to work" w/ just asking the rads to handle 50-60% of the heat load, I'm all ears. I expected significantly less than 100% but not that low.

Quote:
Originally Posted by 47 Knucklehead View Post

So it looks like my build will be able to roughly dissipate 291 (30mm 480 P/P) + 219 (60mm 360 P/P) + 125 (60mm 240 P) or 635 watts of heat (using 1200 RPM fans).

That should be good enough for a 3770K and 2 GTX 780's, all overclocked.
biggrin.gif
Well.... I have a 4770K (4.6 Ghz) and two 780s (25+% on core / 20+% on mem) and I'm at 53% (508 watts) .... so you're way ahead of me and my delta T under Furmark is 8.4C with 420 + 280 and 5 phanteks rad fans at 1200 rpm......I'd say ya golden with a bout 80%.

EDIT:

1. XSPC Rads Added .... no data for AX
2. EK Rads Added
3. Swiftech Rads Added
4. Hardware Labs Added
 

·
Premium Member
Joined
·
4,967 Posts
Nice work!

Although, I've learned in time just how hot the 3930K runs. I now have my machine in a small office and I can not leave the door closed when gaming or it turns into a sauna.

This old chart I found does't seem all that far off, at 4.5Ghz and on upward heat takes off sharply. Between 4 and 4.8 Ghz, you're talking over double the heat.
devil.gif


wattvfre.jpg


Glad to see the general emphasis in estimating the heat load, overclock plays a huge role as well.
thumb.gif
 

·
New to Overclock.net
Joined
·
2,505 Posts
Discussion Starter · #9 ·
You did all the work .... I just did some math
smile.gif


Since you been at it so long, and as we have discussed before, would much appreciate your insight as to just what % of the heat load should be expected to be carried by the rads.... for example, I can see tubing accounting for some % but that % won't be static as when ya go from 1 GFX card to 2, 3, and 4 as the total length of tubing won't change much adding 1-3 sets of crystal links.
 

·
Premium Member
Joined
·
4,018 Posts
Quote:
Originally Posted by JackNaylorPE View Post

You did all the work .... I just did some math
smile.gif


Since you been at it so long, and as we have discussed before, would much appreciate your insight as to just what % of the heat load should be expected to be carried by the rads.... for example, I can see tubing accounting for some % but that % won't be static as when ya go from 1 GFX card to 2, 3, and 4 as the total length of tubing won't change much adding 1-3 sets of crystal links.
The amount of tubing wont, but your ΔTemps sure will. XD
 

·
New to Overclock.net
Joined
·
2,505 Posts
Discussion Starter · #11 ·
Well that's what I am getting at ..... if we say 60% of the heat load handled by the rads, then what accounts for the other 40%.

Could be argued that say 10% of GPU heat is radiated out thru the surface of the block..... well with SLI and a space between the rads it *could* be argued that 10% still holds with 2 cards in SLI with that space between them ..... but if we had 4 cards in SLI, then if the original 10% is good for 1 card and possible for 2 cards with adequate air flow between, , I don't think 10% *holds* for 4 cards.

Similarly, if we said 5% of the heat will be radiated by the tubing .... well if 5% is *good* for 1 GFX card, as the length of tube doesn't increase proportionally for 2, 3 and 4 cards, the 5% will be more and more *wrong" as the number of cards increases.

So what I am pondering is .... this model could probably be further refined to change from say everything totaled up x 60%

CPU Load x 60%
MoB VRM x 60%
GPU1 x 50%
GPU2 x 70%
GPU3 x 90%
GPU4 x 100%

Not suggesting the numbers are "real" just that the conditions they serve under warrant them being different.
 

·
New to Overclock.net
Joined
·
2,505 Posts
Discussion Starter · #12 ·
Nor sure wghat happened here but my edit from last night on the OP was repeated in this message spot..... prolly my fault (way past my bed time)

So as not to waste the space, will answer Jamaican Reaper post from immediately below here
Quote:
Originally Posted by Jamaican Reaper View Post

Page Bookmarked....
thumb.gif


Thx alot for doing this....
Well, it's not like i was being altruistic and all...... I did it initially for my own selfish reasons
smile.gif
cause I got tired of recalculating everything. (Should also note that Martin created waaaaay better spreadsheets ..... but were based upon specific rads and ones I was using weren't there.) Then as more and more peeps asked.....it kinda grew and was expanded into a bigger and bigger project.

I just updated the Alphacool Estimator and will re-upload in Post No. 1. The new version has some additional sheets which give a sharper estimated based upon fan RPMs in increments of 50 ..... ya input the number numbers in the yellow boxes (i.e. No. of of 120/140.. fan mounts) ya have (i.e. 3 for a 360) and it tells ya watts for single fan, watts in P/P as well as cost per watt for both. Good for comparing for example

XT45-420 w/ 3 1250 rpm fans ($130) = 250 watts in push @ a cost of $0.52 per watt
XT45-480 w/ 4 1250 rpm fans ($165) = 245 watts in push @ a cost of $0.67 per watt

This additions were "a bear" as I made a lot of copy / paste errors .... Went back and think I caught most of them ..... hopefully someone can go thru and check the math and pick up any mistakes I made....especially the cost tables.

Still looking for input to refine the % allocations for each component. We all know that electronic components convert 100% of their energy to heat. But our rads don't handle all the heat.

Depending on how ya sized ya PSU and how good it is .... your PSU turns about 10-15% of the power drawn from the wall into heat..... on my system, I draw about 685 watts under Furmark but the CPU is not full stressed, drawing about 30 - 40 watts according to AiSuite 3 and HWInfo64.... so add 100 and ya at 785. I calculated 792 so looks like it's right on the money.

Using Stock Settings BIOS Profile = 589 watts calculated
CPU (Under Furmark) = 40 watts
MoB VRM = say 15 watts
GPU1 x 110% TDP = 231 x 1.10** x 0.9* = 229
GPU2 x 109% TDP = 231 x 1.09 x 0.9* = 227
35x2 @ 40% PWM (From Martins test) = 12 watts

* 231 is at wall so adjusting for PSU efficiency
** Cards running @ 100 / 109% TDP

So the above has would have the rads (420 watts) handling about 79% of the heat load. Now getting back to allocations, I'll put some "shot in the dark" numbers in ( ) ..... 90% for CPU stuff (small block surface area), 70% for GFX stuff (large block surface and backplate area) and 100% for pump as that was a direct measurement.

CPU (Under Furmark) = 40 watts x (90%) = 36
MoB VRM = say 15 watts x (90%) = 13.5
GPU1 x 110% TDP = 231 x 1.10** x 0.9* = 229 x (70%) = 160.3
GPU2 x 109% TDP = 231 x 1.09 x 0.9* = 227 x (70%) = 158.9
35x2 @ 40% PWM (From Martins test) = 12 watts x (100%) = 12

Total is 380.7 and the Rad Estimator says my rads do 420 watts @ Delta T of 10C.... Looking at the temp meters, I'm showing 8.9C, so my load must be lower than 420 ..... 420 x 8.9/10.0 = 373.8

Now that does that mean we nailed it..... no, those numbers were guesses so we'll need a lot more data .... I started at stock because my BIOS upgrade borked my OCs and haven't been doing much benchmarking till get answers from Asus .... but its Furmark stable so will do 4.5Ghz and 25% OC on the GFX and see if numbers still hit.
 

·
Registered
Joined
·
3,283 Posts
Page Bookmarked....
thumb.gif


Thx alot for doing this....
 

·
Registered
Joined
·
20 Posts
I have a system that's a little different. Could you help me figure out how much power it can handle?

I'm using a generic 3 x 120mm fan cross-flow radiator @ 30mm. It has 18 FPI and 12 flat tubes. Total dimensions are 390mm x 118mm x 30mm.

I have a pump that pushes 3 gpm through the circulation.

I'm using 3 120mm Sanya Denki San Ace 9GV1212P1J091 fans. They're 38mm and here's what I've been able to determine about the fans (each):

3A 12V 36W
224 CFM
6400 RPM
7 Blades
1.45 in H2O = 36.8 mm H2O = 2.7 mmHg static pressure
loud ~ 64 dB-A

Assuming 25C ambient temperature, how much heat can I expect this radiator / fan assembly to extract?

The total power dissipated by the electronics is 600W. The pump will add another 30W. Assume that the electronics cooling blocks to fluid have almost no temperature drop (1-2C).

In order to get the electronics to operate where I need them, I need to keep the temperature under 35C. Is this possible with this setup?
 

·
Registered
Joined
·
20 Posts
I don't care about noise level at the fans (clearly)
biggrin.gif


If this isn't sufficient, I'm open to additional cooling with peltier devices between the electronics and the cooling blocks to create a heat pump and inject more heat into the cooling liquid to keep the electronics cooler.
 

·
Registered
Joined
·
2,024 Posts
At room temp about 300 Watts for a 10°C delta,
although this is relative, since the larger the delta water-air, the more wattage a rad can dissipate

so, overly simplified
liquid temp 35°C - Ambient Air temp 25°C ~ 300 Watts
liquid temp 35°C - Ambient Air temp 10°C ~ 600 Watts

all this +/- 10%-15% variation, cause there are other variables, like it might actually perform better if the pump is set a little under 2GPM iso +3GPM .... for starters the pump will add less wattage
smile.gif

Might be easier to just add another 360 radiator

Now, with TECs
overly simplified
liquid temp 45°C - Ambient Air temp 25°C ~ 500 Watts
liquid temp 70°C - Ambient Air temp 25°C ~ fail of pump or blocks or fittings (o-rings) or hoses

You do know the problem with tec... for each watt, you have to handle double or triple
For a 600 w load on the cold side, you'ld be looking at at least 1000-1200w to dissipate from the hot side... a simple 3x120 rad cant do this
You'ld have to go to at least 3x 360 rads to handle the hot side

Not to mention you'ld have to use multiple 400w or 331w TECs parallel at half voltage/current (eg 12V iso 24V umax otherwise they are just heat generators iso heat pumps)

And not the cheap ebay ones either.. the good stuff like http://www.customthermoelectric.com/tecs/pdf/19911-5M31-28CZ_spec_sht.pdf

The again, the benefit of TECs would be, that with the right PWM controller, you could "fix" the cooling by the TECs at a set temp.
But you'ld still need more Radspace to dissipate it all.

All in all, i think you'll have to consider using double or triple the radiator space....

Now, if you say "money is not of the issue" , then there are some more options
smile.gif
 

·
Registered
Joined
·
20 Posts
thanks.

your description lines up with a standard rule of thumb for a standard system, but I got the highest FPI radiator I found and these Sanyo Denki fans screech at 6400 rpm with 224 CFM and 37mmH2O?

I can go to a lower flow pump, but the water circuit is long (24ft total) so I didn't want it to move too slowly.

Can you walk me through the math or the options?

I wasn't going to use 600W TECs. I was only looking to use them as a boost - 216W distributed over 6 small TES1-12704. I was expecting that wouldn't increase the water temperature by more than 20C (and cool the ICs by that much more than a non TEC version).
 

·
Banned
Joined
·
3,726 Posts
So my i7 is 130w at stock speeds, I am guessing that at 4.0 ghz its around 190W
OC Wattage = TDP x ( OC MHz / Stock MHz) x ( OC Vcore / Stock Vcore )^2
OC Wattage = 130 x (4000/3200) x (1.30/1.20)^2
OC Wattage = 130 x (1.25) x (1.083)^2
OC Wattage = 130 x (1.25) x (1.173)
OC Wattage = ~190.62W
580 GPUs each are 244W each at stock. so I'll assume....hmmm...310watts each at a 850Mhz, OC?
190 + 310 + 310 = 810watts total.
810 x 0.85 = ~688.5 watts to be dissipated.

the MCR320's radiators are rated for ~350W

Depending on fans used, that drops me in the approx range of ~350watts per 360mm radiator to be dissipated with 1200-1500rpm fans.

so I needed 2 radiators

http://www.overclock.net/t/1573189/serial-vs-parallel-9-6lpm

by the time I got done I had a 4.4C delta T for this system
ambient temp of the room is at 28.6C
water temp at load is 33C
 

·
Banned
Joined
·
3,726 Posts
1 - 20 of 116 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top