If you buy this stuff, then you have an excuse for that 200 dollar pump of yours. :p:
Printable View
If you buy this stuff, then you have an excuse for that 200 dollar pump of yours. :p:
A bit later, please, I was really busy with some other things.
It took some time to make a decision how to set up the RAD. Temperatures will be very low anyway for now, as I plan to use it with huge reserve - it should be able to handle QX9650@~5GHz and 2x9800GTX in the future, each 9800GTX will probably consume 300-350W - may be a bit more (comparing to 8800GTX and new 3870 X2), and CPU will have to dissipate around 150W @5GHz, AFAIK, in total it will be up to 850W to dissipate. It is not much, but now I have only 8800GTX and QX9650 to handle, therefore it is around ~350-400W in total (CPU@~5GHz and GPU@650MHz).
I am going to try and run it in a passive mode and get the temps reading and after that I am going to use 5xNoctua P12 in ULNA mode ... at least for now.
QX9650 is a great CPU, it can run at 3.2 GHz with 37 Idle/49 Load (2 hours running Prime 95 stress test) using ZeroTherm Nirvana 120 air cooler (I use it temporarily) at low fan speed (~1300-1500RPM) and 41 Idle/54 Load at lowest fan speed (~1000RPM), therefore I expect very good temps with a WC system.
I have made a test run for WC system, it looks that the flow is quite good. My setup for now is D-Tek fuzion (a bit bowed with a small o-ring)->MCP60->1080->MicroRes->Laing D5-38/810
where are you getting these wattage figures from they seem to be off mate...
Mostly from reviews, here are some of them:
http://www.hwupgrade.com/articles/vi...-in-sli_5.html -that's a consumption measurement for a whole PC
Take 8800GTX SLI data and remove single 8800GTX data, that will give you the results for one card itself, for us it is 464-289=175W (full load), 317-171=146W (idle), I took 150W as an approximate, we can use 200W, it doesn't really matter in this brief guessing.
Yes, it will consume a bit more in OC mode (not significantly IMO without raising the voltage), I will measure and provide the data.
For CPU http://www.digit-life.com/articles2/...650-page4.html
Having 76W result at 3GHz my assumption is that it will not pass 150W @5GHz, but again, we will see. Having a HUGE reserve in dissipation power with 1080 (I can always use more fans, or run them at higher RPM, or even replace them with more powerful ones) ... I don't need to worry about such exact calculations ... IMO ;)Quote:
Despite the claimed 130 W TDP (identical to the QX6850's), the real power consumption of QX9650 equals 76 W at 100% load. It's even lower than that of the 65nm lower-clock QX6700. And even if we consider this value as absolute, 76 W is really low for a top CPU in the series. It's been a long time since we saw a high-end processor not exceeding the symbolic 100-watt line.
Also I would like to note that consumption power is not equal to required dissipation power for CPU+GPU (the last one will be somehow lower), as we need to keep in mind some power, dissipated automatically inside the case, through motherboard elements, etc.
5mhz or 5ghz?
I will be interested to see temps and how you mount this...
Will post some pictures ASAP. Temps will be available after the weekend, I hope.
Thanks for your input, andyc, 5GHz is not a goal to run with WC stable 24/7 (IMO it is impossible with WC), I just want to try to reach that limit, but it will be definitely limited by the power supply voltage, I will not go over 1.7V for that result - whatever it will be possible to reach is going to be fine.
And it will be just a test result, I am going to run the CPU at 4.2-4.5 GHz 24/7 ... depends on the voltage results.
I have found quite interesting info on the power consumption of QX9650, my assumptions were wrong, it consumes <150W only before 4 GHz, after that it consumes significantly more.
http://www.anandtech.com/cpuchipsets...spx?i=3184&p=2
Thanks, that what I have seen for others ... I will definitely try to run it at 670.
well, all dice sessions, you see everything above 5 ghz at about 1.5 volts. On water and above ambient temperatures on cpu, even pushing 1.7 volts and can't break 4.3 ghz
There is very big difference between "seeing" some high number and actual stability at that speed. Those who have high OC with 45nm CPUs mostly do not test well. Almost all high OC'es are unstable, whether this semi-stable state is acceptable mostly depends on intended usage.
Thanks again, andyc.
OK guys, as promised, here is my decision how to mount it.
http://www.audio-measurements.com/images/mount-1080.jpg
Yes, it is outside of the PC case, and there are several thought why:
1. It is better to direct heat flow/air flow specifically, so it will create a good environment for me. The RAD itself it big enough to mount it inside a PC case, especially for CM Stacker 830 SE, only possible logical solution IMO is to use something like this, if someone wants to use 1080 internally.
If such things such as Thermaltake Symphony or Rocket, Zalman Reserator 1,2 or XT exist using external cooling idea and only 3/8" tubing, why similar project can't exist using 1/2" tubing?
2. I don't want to worry about future/possible leaks of distilled water+additive, therefore will keep only CPU and GPU water blocks inside the PC case to lower the risk.
3. For me it is good to design a refilling output for used coolant, therefore I will be able to replace it or clean the system easily. It is also easier to create using external design.
The project is not moving forward as fast as I want due to some absolutely unexpected events. For example:
http://www.audio-measurements.com/images/No-space.jpg
As we can see, the recommended Swiftech design (MCP60+MC8800) doesn't allow me to fit tubes, it is not enough space. I had to correct it, spending a lot of time on something, which was already "designed" by some engineers.
The other discovered thing was even worse:
http://www.audio-measurements.com/images/latex-MC21.jpg
Provided in MC8800 set rads didn't want to stick to the chips. I have found some information in forums that the surfaces of chips had to be cleaned, but I already cleaned them carefully using standard for such purpose isopropyl alcohol. It didn't work, the rad surfaces were not sticky at all. I have discovered that if I clean surfaces with pulti-purpose remover (actually it's a latex solvent), I can stick the rads, they fit well - it looks well. But as you can see on pictures, the contact surface was extremely small. The reason why is - those sticking surfaces of used thermal tape were covered with latex paint, the paint dried out due to a long storage time for rads, and to refresh their sticking ability someone has to use a solvent, then those pained surfaces again work fine (if you want to wait while everything will be dried out).
I still want to blame "engineers", who have used latex as a thermal pasta :ROTF: while such things as 3M™ 8815 Thermal Tape exist, and even used in competitive for Swiftech products, for example:
how much temps you get on this Core??
I have the test run results, its 31 idle 41 load at 3.2Ghz and 37 idle 44 load at 3.9 Ghz with Prime95, with Everest full load stress test @3.9Ghz it is 37 idle 50 (50 is a peak spike) load. Currently I use only 3 fans - 2 of those DFS122512L-PWM (52.71CFM@1200RPM@25dBA) and one CM TLF-S12-EB (42.73CFM@1200RPM@22dBA) while I wait for Noctua P12. CPU uses auto mode @1.2V core ATM, the ambient temp was 72F.
Those results include CPU and GPU (EVGA 8800GTX SC@621 - will be running at 670 a bit later)
Temps with Noctua P12 fans, as well as for higher freq will be posted later.
Playing with those stress tests I can see that you all guys are right, no way I can get anywhere close to 5GHz with WC and 4 cores fully loaded during hours, at least I can try to break 4.3 a bit later ... one thing is good though, the system is almost whisper silent and absolutely stable running at 4GHz with CPU temps around 50C.
I got some interesting results, the FSB is max stable @1750 MHz (with auto voltage settings), after that I got a "blue screen" randomly. Got CPU to 4.156GHz@1.4V@x9.5@1750FSB stable with full load @53C(during the whole test several temp spikes @57C and one @60C, average is 53C) while GPU is at 675/1566/1080@40-41C, will continue experiments with core voltage.
I still use those 3 simple mediocre fans ... Magicool 1080 is a good RAD, IMO. It allows me to achieve the electronic limits of chips and boards still having good temps.
To keep the flow rate I would not use any washers/nozzles on D-Tek Fuzion, just a small o-ring to avoid the discussed leakage and still keep it's restriction low.
http://www.audio-measurements.com/images/RAD.jpg
Here is the final version of mounting with finally arrived Noctua P12, I use 5 of them all in ULNA mode (900RPM), as planned before.
Temperatures are even a bit lower than with 3 previous mediocre fans (running at 12V with full speed), peak spike on the processor temp was 59C max and 53C was an average temp during several hours of full 4-cores 100% loading.
The noisiest part of my system are hard drives for now, though a CM 830 rubber separated rack is used. The carpet separator helps to create a correct airflow (moving heat up and out of the user), I intend to add two more similar separators on the sides.
Please note, that RAD is fixed using small angle also for creation of a correct airflow - out of the user.
Very nice mate!
Whats the proc clocked at?
gmcg-you have one awesome project
I'm thinking about doing the same with an antec titan case and have the rad mounted on the side. the antec case, btw can be found here and comes with a true power trio 650 psu for only 80 bucks.