• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Help Improving Water Loop Performance

FireDemon

Limp Gawd
Joined
Feb 10, 2006
Messages
256
Hey guys. I was looking for some input on possible steps I could take to improve the performance of my water loop.

What I have is a Crosshair V with an FX8320. the system is installed into a 4U rackmount case so due to space constraints I was limited as far as components however I chose a very good solution with a small footprint.

I used an M184 2x 92mm radiator and mounted it midway in the case with two 92x38mm 80cfm akust fans pushing through the radiator (Very well might I add with nearly 10mm static pressure). I used a single 3.5" reservoir to monitor the fluid level and make filling easy. the system is all connected with tygon silver 1/2" tubing. Coolant is straight distilled water with a drop of biocide and a single kill coil at the waterblock inlet. The pump is an MCP350 I had leftover from an old setup with an XSPC top and the waterblock is an enzotech sapphire. The order of the loop is reservoir > pump > radiator > block > back to res.

Initial startup of the system yielded very good core temps around 32C full load at stock speeds.. Once I started overclocking however the temps started climbing. Right now I have my CPU at 4.5ghz with a 1.416 vcore. northbridge is at 2.5 with 1.275V and I'm currently testing a 3ghz hypertransport with stock voltage. Load temps stay around 50C on the core. I had the chip stable at 4.64ghz however this required increasing the core voltage to 1.464 and load temps shot up to around 58c. I think the CPU has more left in it however my cooler can't keep up above 4.6 and I really don't like the high 50s either so I backed it to 4.5.

Temperature of the air blowing out of the radiator under load is reading 33C (However as you will read on it becomes more apparent that this is more ambient temp at the rear of the case than it is actually radiator air temp) and idling the fans down vs having them at 100% only yields a 1-2C and also the radiator feels barely warm to the touch which leads me to believe the radiator is not the main bottleneck.

Taking the lid off the case yields a 3 - 4C drop in core temps but nothing drastic. The radiator is receiving good airflow. I installed 4 60mm fans into the lid of the case each rated at 25cfm and with the lid on they are feeding fresh air directly in front of the rad fans.

I was thinking of possibly upgrading the block to an EK supremacy and possibly going with an MCP355 pump, or even a low powered TEC to avoid condensation issues but wanted to know what everyone thought. I honestly don't think the 355 is going to yield any gains over what I have now for a CPU only loop and the idea of a TEC kind of scares me.

I left the system running Prime95 all day (10 hours) running OCCT GPU mode parallel scheduled to turn off after 1.5 hours to ensure maximum stability with my current settings. When I came home, this is what I found:

- Ambient room temperature was 22C
- Ambient case temp directly in front of the radiator read 28C
- Air exiting the radiator was reading 33C
- CPU core temperature was at a steady 51C



Took the lid off the case to see what the temps would do and recorded the following:

- Ambient room temperature was the same
- Ambient case temp directly in front of radiator dropped to 27.5C
- Air exiting the radiator dropped to just a hair under 30C
- CPU core temp couldn't make up its mind between 48C and 49C

100% stable at the following settings (Still fine tuning and have to clock the RAM next. Currently at 1600 to ensure no false CPU errors.)



Now some photos of the case to illustrate the whole setup. I have had this case for 5 years and originally planned to build a dual opteron server out of it. When my vapochill finally started condensing and took out my old setup however, I decided it was to start anew and threw that case away. I salaved some of the hardware such as the vid cards, drives, power supply and a few other goodies. I always liked the clean low profile look of the rackmount case however knew making a gaming machine that didn't cook like an oven was going to be a challenge. So out came the dremel tool.

I cut 4 60mm fan holes in the lid to provide fresh intake air. Each of the fans at full speed is rated at 25CFM. Also occupying the 3 left 5.25 bays is an 80CFM 120MM intake fan blowing over the 4 disk RAID array. I know this isn't ideal because it is introducing the heat from the drives into the case but I wanted them all internal so it was a necessary evil. This fan however has very poor static pressure and I plan on replacing it with a better fan rated for 12mm static pressure at 150CFM (Full speed of course).




Exhaust was obviously going to be a concern. Aside from the power supply and two 460s (Which introduce probably 50% of their load back into the case), I only had a single 60mm fan to work with. I employed a 60x38mm sunon fan rated to 56CFM at full speed. At idle I run it at 30%. When the system loads up it increases to 50% (Still moves alot of air at this speed) and if the socket temp goes over 75C (Never happens unless something goes wrong) speedfan kicks it to 100%.



I wanted a radiator that was cleanly fixed to the case or inside for good looks, so I went with an M184 (2x92) radiator and mounted it on the crossbar towards the front of the case. I had to custom fabricate the bracket which is basically just a frame to hold it in place. The rear of the radiator is fully open.

The reservoir is mounted at the front of the case. It was an internal 3.5" bay but I cut the front open to make it external.








For S&Gs I measured the temp of the air exiting through the sunon fan at the rear of the case and it read 33C

Any recommendations?
 
Last edited:
Nice rig, but you really should rethink your radiator setup. Your radiator fans should be exhausting the heat they extract directly out of your case rather then blowing it around inside. Also, 60mm fans aren't all that efficient and they tend to be quite loud. You might also want to look into externally mounting your radiators or turning them 90 degrees and venting them out the case lid... perhaps even going with a push/pull config to further increase your cooling power... Shrouding the radiator can also help, but you'd lose the clean look you now have. Lots of options to consider... another would be a larger case with better fans (again, the small fans are really limiting you... Lots more cooling potential and airflow when you go with some quality 120mm fans like Noctua or Scythe Gentle Typhoon 1850's.). Cheers!
 
Nice rig, but you really should rethink your radiator setup. Your radiator fans should be exhausting the heat they extract directly out of your case rather then blowing it around inside. Also, 60mm fans aren't all that efficient and they tend to be quite loud. You might also want to look into externally mounting your radiators or turning them 90 degrees and venting them out the case lid... perhaps even going with a push/pull config to further increase your cooling power... Shrouding the radiator can also help, but you'd lose the clean look you now have. Lots of options to consider... another would be a larger case with better fans (again, the small fans are really limiting you... Lots more cooling potential and airflow when you go with some quality 120mm fans like Noctua or Scythe Gentle Typhoon 1850's.). Cheers!


Thanks for the compliment. And believe me I am not new to any of this. As much as I wanted to do an external radiator/mount the radiator directly at an exhaust port of the case, due the space constraints of a 4U case I am extremely limited in terms of options. Sure I could have reused my 3x 120mm radiator I had laying around and just sat it somewhere outside but that wouldn't have been very appealing to me. I 100% agree that my current setup is not IDEAL in terms of the hot air exhausted from the radiator being blown overtop of the motherboard, however the #1 dictating factor here is SPACE. There is absolutely nothing I can do IE mounting the M184 to the case lid as an exhaust port or doing an external radiator without running into severe/unworkable fitment issues/getting ghetto.

Also, don't sleep on the M184 radiator. It may be small but in terms of a CPU only loop I think it is a great performer. Take a look at what this guy did using the same radiator on not only his CPU, but his chipset and GPU as well: http://www.xtremesystems.org/forums/showthread.php?225268-Testing-the-HW-Labs-Micro-184-GTX-Rad

Now as far as the 60mm fans go. Again I would have loved to do something more extreme. I would have loved to go with say 3x 80mm fans at the top of the case for intake but in there wouldn't have been enough room to cram everything in there once the lid was on and in terms of being visually appealing I didn't want to get ghetto and mount them on the outside of the lid so I would have had 3 fans protruding from the middle of the case top. What you see in the photos is my best effort at yielding good performance while keeping it visually attractive. My previous rig was cooled via phase change so watercooling is like child's play to me in terms of temps however considering everything I have to work with I believe this setup has another 5 - 10C left in it the question being though what is holding it back. Is it poor case airflow or is it that waterblock?

Just to stress the importance of my current temperatures:

With the case top off, radiator intake temp difference of HALF a degree celsius. radiator exhaust temp difference of 3C. CPU core temp difference of approx 2.5C. This case does not have the BEST airflow but considering those numbers and what I have to work with so far I don't think I've done too bad at all. I am just looking for possible ways to improve of what I have done so far.

Anything else you can think of?
 
Upgrading the waterblock would probably net you better temps. Cheap upgrades include the XSPC Raystorm and Swiftech Apogee HD (don't get EK).

The FX series processors are fine up to ~70 C (not sure of the exact numbers). I wouldn't worry about your load temps until you get around 65 C or so. Processors today are able to run much hotter than older processors.
 
Temp problems in watercooling are almost always either one or some of the following:

1) sub ideal flow
2) Bad block mount
3) sub ideal heat exchange stage

I raise my eyebrows at using that tiny a radiator for that hot a CPU....and would change the rad before I bothered with the block.
 
Temp problems in watercooling are almost always either one or some of the following:

1) sub ideal flow
2) Bad block mount
3) sub ideal heat exchange stage

I raise my eyebrows at using that tiny a radiator for that hot a CPU....and would change the rad before I bothered with the block.

Not to sound like an asshat, but I would recommend doing more research before you go assuming that my "tiny ass" radiator is dragging down the rest of the loop just because I am not on the get-the-biggest-radiator-you-can-afford bandwagon.

Did you you check out the xtremesystems link I provided on this radiator? The M184 is more than up to the task of a CPU only loop.

Now as far as the mounting, I thoroughly cleaned both the CPU IHS and the waterblock with 91% rubbing alcohol. I then took some IC Diamond compound, held the syringe under hot running water to soften it up and applied it to the center of the cpu. I placed the block on top, pressed down firmly and *slightly* twisted the block back and forth (Maybe about a 1- 2 degrees). Then I tightened all 4 screws while still pressing down firmly on the block.
 
Upgrading the waterblock would probably net you better temps. Cheap upgrades include the XSPC Raystorm and Swiftech Apogee HD (don't get EK).

The FX series processors are fine up to ~70 C (not sure of the exact numbers). I wouldn't worry about your load temps until you get around 65 C or so. Processors today are able to run much hotter than older processors.

What is the best block you would recommend? Price is not an issue here. In fact the only reason I am using the sapphire is I got it free as part of a trade and figured I might as well try it out first before spending more money since I had it laying around.

Now as far as max temp, I am seeing some places say 70C but alot of people seem to have the belief the safe oimit is the same 62C as the previous AMx CPUs?
 
Last edited:
Not to sound like an asshat, but I would recommend doing more research before you go assuming that my "tiny ass" radiator is dragging down the rest of the loop just because I am not on the get-the-biggest-radiator-you-can-afford bandwagon.

Did you you check out the xtremesystems link I provided on this radiator? The M184 is more than up to the task of a CPU only loop.

Now as far as the mounting, I thoroughly cleaned both the CPU IHS and the waterblock with 91% rubbing alcohol. I then took some IC Diamond compound, held the syringe under hot running water to soften it up and applied it to the center of the cpu. I placed the block on top, pressed down firmly and *slightly* twisted the block back and forth (Maybe about a 1- 2 degrees). Then I tightened all 4 screws while still pressing down firmly on the block.

Fair enough. You're also pushing a chip known for running hot to the limit.

Presuming one of those 3 I have mentioned isn't off. But look changing your block will only net you a handful of centigrade if that in a CPU-only loop. Presuming you're right, and your heat exchanger isn't holding you back. And presuming your block mount is good...Your only option to get meaningfully better temps would be to lower your coolant temperature directly via a water chiller.

That is the shape of things, from what I've read the max on those chips is 70C so at least you're not presently cooking the chip at that OC and cooling state.
 
I believe your temps are reasonable considering the conditions. The radiator may be small, but I believe it's transferring as much heat as possible in the current configuration, and a larger rad wouldn't change the result (this can be checked by measuring fluid temps before/after radiator).

Your best bet is trying to get ambient temp. air to the rad intake, thus yielding maximum possible heat transfer per the conditions. The ~5c (~10F) difference between ambient and intake is huge, especially considering there shouldn't be any significant heat generation in the front of the case. Cutting a vent hole in the "top" panel or adding an intake fan in some spare 5.25 bays would alleviate this.

Also, the difference between the best/worst CPU block is <2C and subject to mount quality, so don't waste your money.
 
Guys, I just wanted to give an update here as I have done something very simple that yielded a small increase in lid-on performance.

If you take a close look at the photo of the hoses running to the 3.5 reservoir, you will notice the back of my power supply has a vent in it. The power supply was recycling hot air to the front of the case where the radiator pulls its air from.

I was very skeptical of doing this, but looking up photos of alot of modular supplies alot don't seem to have vents there so I think it was an extra measure Enermax took. I put a strip of electrical tape over the vents blocking hot air from the PSU entering the front of the case.

Previous core temperature difference between lid off and lid on was 3C. Now with that vent in the PSU blocked the difference between lid off and lid on is roughly 1.5C.
 
Last edited:
If cost is no concern, the Koolance CPU-380 is currently one of the best, if not the best block on the market.

Based on Martin's testing, the D-tek Fuzion V2 performed about 1-2 C better than the Enzotech Sapphire. Bundymania's testing shows the XSPC Rasa performing ~4 C better than the D-tek Fuzion V2. The CPU-370 does about 2 C better than the Rasa based on Martin's testing. The 380 is ~1 C better than the 370. Add that up, you're looking at a difference of up to 8 C.
 
Back
Top