Build log: Corsair 600T

[Sprucemoose]

Waiting to RMA the graphics card, I have fun fiddling around with the Aquaero control system and my CPU-only loop.

I am able to set Delta-T as a virtual sensor. I guess the real Delta-T is the difference between air-in and water-out, but I haven't connected the water-out sensor (need extension cable) and used the water-in temp instead. The real delta-T should theoretically then be lower.

I can then set my fans to try to hold a constant Delta-T value. As the Delta-T reaches let's say 5 degrees, the fans start reving up trying to keep it a the set value. Increasing the setpoint value will result in a more quiet but hotter system.

Just for fun, I set the Delta-T setpoint to 1 degree (fans will be forced to run at 100%) when running Intel Burn Test with turbo @3,8GHz enabled. Pump is for the occation at 100% (it says 61%, but that is equal to the rpms at 100%) resulting in a flow of 1,6GPM. At this picture, the temps seems to have come to some sort of equilibrium (Photobucket won't let me use the max resolution, so some of it can be a little hard to read):

Temps won't be this good when I put in the graphics card, though

 

[hokiealumnus]

Sprusemoose, I'm sorry your thread just got crapped on. I've cleaned it up for you. Hope you get your GPU RMA soon!

m0r7if3r, you could have used something a bit more tactful than 'useless', which prompted the response you got.

Gritash, next time I catch you circumventing the swear filter - especially for directly insulting someone - using dollar signs or any other method, deleted posts are going to be the least of your worries. It would behoove you to read the rules, as direct insults are against rule #1. On that subject, this will be the only warning you receive.

Conumdrum, sorry; your post just didn't have any reference once the rest were gone. You did nothing at all wrong.

-hokie

 

[Sprucemoose]

It seems like the thread got out of hand a bit, thanks for reeling it in. A tiny update on the RMA situation. i got myself a new PSU, it's waiting to be installed.

The gfx-card however, needed a rethinking. The new ETA of the 6990 card is now 1st of november, and I have no intension of waiting for this long. I'm changing my buy to the nVidia side. I've just ordered the ASUS GTX 590 (with an ETA of only 3 days). I plan to put a full cover EK copper block on it, but I'll test it on air prior to switching to water.

Reading up on the GTX590, I realize that there is not much room for overclocking, but it will probably be better regarding the vrm squeel of the 6990s.

Hopefully, everything will be completed soon. More pics and temps will be posted then.

EDIT: Changed ETA of the ASUS card, this time two weeks (and still an unconfirmed date). Going with the Zotax GTX 590 instead (which is in stock).

 

[Gritash]

Best of luck with the build Spruce. Hope it all works out for ya. (wont be posting anymore "useless" information)

 

[balizeiro]

I just bought an AquaComputer Aquaero 5 LT fan-controller, and I was wondering how to secure it to the case. Could you please post a picture of how you did it?
Thanks

EDIT: Nevermind, I hadn't realized you bought an Aquaero bay mounting kit

 

[Sprucemoose]

I just bought an AquaComputer Aquaero 5 LT fan-controller, and I was wondering how to secure it to the case. Could you please post a picture of how you did it?
Thanks

EDIT: Nevermind, I hadn't realized you bought an Aquaero bay mounting kit

Yeah, I didn't know about needing an extra mounting kit when buying the Aquaero either. At that time, the only place I could find it was at the AquaComputer webshop. I kinda wonder why they couldn't just include the two metal brackets in the first place. But it sure is a great fan controller!

 

[Sprucemoose]

Time for the weekly update for those interested: I finally got my Zotac GTX590. Surpisingly, the card lies in a regular plastic bag, not one of those anti-static ones. Looking at some of the unboxing videos online, however, this seems to be what Zotac does.

Anyway, I installed the card and fired it up. No artefacts in BIOS this time around . I installed the newest drivers, and ran 3DMark 11 with a score of P8863 (which is about what I was expecting).

Even on idle, the fan noise from this card is pretty loud. Concidering that the HD6990 is even louder, I can't imagine how videocard manufacturers can continue making these monsters without seriously concidering implementing new ways of cooling. Watercooling FTW!

I had previously thought about getting an EK block for this card, but the nickle plating incident put me off. I have the option of buying the pure copper+Acetal block in a local supplier, but I become somewhat skeptical to the whole product line when the manufacturer is caught cutting corners.

So I went ahead and ordered the Koolance VID-NX590 block, which should be a good choice. Hopefully I'll have it installed within a week. Final pictures will follow!

On a side note, the water temps and Delta-T was somewhat influenced by installing the video card. Even though there is no water going through the 590, the radiator fans had to increase their rpm to keep a set Delta-T of 5 degrees. I guess this can be explained by how the aircooling on the video card works: The fan blows half of the airflow out the back of the case, but half of it is recycled into the case itself, thereby increasing the air temperature. This would in turn impact the efficacy of the radiators. I guess this is a sign of things to come when I connect the video card to the loop.

 

[Sprucemoose]

Express delivery! The GPU waterblock arrived today, and I got to fiddle around with it tonight.

The parts:




Taking off the stock cooler was no biggie, plenty of screws to unscrew, though. Most of the screws where very tiny torque screws, and I had to buy a special screwdriver for this job.

Cleaning didn't take much time, Q-tips with industrial ethanol did the trick:




The most time consuming was cutting all the thermal pads to the right size. The pads had protective film on both sides, and I chose to keep one side on until directly prior to installing the block. Some of you observant readers may see that I miss two pads on the picture (far left corner). A good advice is to double- and triple-check everything, and I was able to spot it before installing. Getting it right the first time around would also be acceptable .

Thermal paste on the GPUs is Arctic Silver 5, the only paste I can get without ordering from abroad. I know that it's not at the very top, but hopefully it'll do. I don't plan to overclock this card with the VRM issue around, so I guess it'll do.




In the end, it all came together. I found that the easiest way to align all the holes in one try was to have the block rest with pads facing upwards, putting the card back down on it. It looks quite nice, doesn't it?




I'll post some photos on the complete rig when I get the card installed and connected to the loop.

 

[Conumdrum]

Yea, TTape is a pain. I found that not cutting individual peices for the ram and Mosfets made life easier. The way you installed it is what I have fornd best for mating the parts too.

Also, Koolance support for spare oring kits and TTape is very good.

 

[hokiealumnus]

Very nice looking card & block, congrats!

 

[Sprucemoose]

Ok, so I got it installed, and it's still working . Now, I want to start testing the cooling solution, and I need some advice.

I was thinking that I have to establish how well the cooling solution works at full heat production combined with both pump and fans at maximum. This should give me a baseline for the cooling capabilities of my rig.

Now, I was a little unsure on how to test GPU and CPU at the same time, but I ended up running Furmark v1.9.1 using Xtreme burn-in at 800x600 windowed mode while running 6 simultanous threads in Prime95 (In-place large FFTs). If Furmark only utilizes 1 thread, I guess that I should have run 7 threads in Prime95. Anyway, the test should produce more heat than any real-life situation, right?

The CPU runs overclocked at 4,5GHz (100x45) and vCore set manually to 1.35V in BIOS. Flow is 1,2GPM at full speed. All fans at 100%

CPU and GPU readings below are the hottest core registered using RealTemp and Furmark/GPU-Z.
Water, Air and Delta-T are the average registered temp last 10 minutes at steady state.

Well, the numbers (centigrades) are as follows:

IDLE:
Water-out - 29,9
Air-in - 25,7
Delta-T - 4,2
CPU - 34
GPU - 34

MAX LOAD:
Water-out - 41,2
Air-in - 25,7
Delta-T - 15,5
CPU - 73
GPU - 52

Now, my first thought is that the Delta-T is too high for a CPU-loop, but borderline for the GPU. What I don't understand, though, is that even with such a high Delta-T, both the CPU cores and the GPUs seem to have temps that are viable. The GPUs actually seem to thrive compared to some air-cooled GTX590 charts I've seen.

Now, I'm really curious about what you guys think of the total cooling solution of my rig. I think I had done decent research prior to buying parts for this build, but I've been a bit worried whether or not my radiators/fans were really up for the job, so this is the moment of truth...*crossing fingers*

PS: Pics will follow.

 

[bennoculus]

Dude! This build is sick! Keep it up

 

[m0r7if3r]

Yea, your DT's a bit high, air temp out is pretty toasty though, so it's just lack of rad I'd say...if it's working, then don't sweat it, your temps are in check, loop looks good to me.

 

[Conumdrum]

Now run Heaven benchmark and prime with 7 cores. Thats more CPU than gaming would ever see. It's a much more realistic GPU load too, that 590 on Furmark is pumping out tons of heat. The CPU temps on your tests are higher than I would like. You 'need' 4.5 on the CPU?

 

[Sprucemoose]

Now run Heaven benchmark and prime with 7 cores. Thats more CPU than gaming would ever see. It's a much more realistic GPU load too, that 590 on Furmark is pumping out tons of heat. The CPU temps on your tests are higher than I would like. You 'need' 4.5 on the CPU?

The only reason for me running it at 4,5GHz is for bragging to my friend who bought an unoverclockable iMac (comparing Geekbench scores)

Now, to the temp issue. I reduced the multiplier to x40 and redid the above test. I also tested with the Heaven benchmark as Conumdrum adviced:

 

[Sprucemoose]

I did some gaming running Battlefield Bad Company 2 single player mode (all settings on full, resolution 1920x1200). Now, there are some temp-issues I don't grasp. I thought that increasing the flow would give me the better temp results, but the opposite happens.

If I set the pump on full speed the difference between water-in and water-out is appr 1 degree while gaming. If I reduce the pump speed, this difference increases and the temperature of the water coming out of the radiator decreases. Delta-T is reduced proportionally.

It seems that reducing the speed of the pump actually gives the CPU a cooler environment.

Some approximate temps:
@1,2GPM
Water-out: 38,5 deg
Delta-T: 13 deg
Delta-water: 1 deg

@0,5GPM
Water-out: 37,5 deg
Delta-T: 12 deg
Delta-water: 2 deg

@0,3GPM:
Water-out: 37 deg
Delta-T: 11 deg
Delta-water 3,5 deg

Now, what is going on here? Giving the CPU the coolest water seems to be smart, and by reducing the pump speed, I'm actually able to do so. The GPU runs at appr. 50deg, so I'm not really worried about those.

 

[m0r7if3r]

not sure what those numbers mean or where you get them, but just use realtemp and see what your core temps are.

EDIT (for cross-post): I don't think that you're grasping what you need to be looking at. Post core temps, not water temps, the resolution on your sensors is almost undoubtedly not enough to provide 1 decimal place of accuracy. Look at the actual component temps, as those are the sensors with the highest resolution (probably) in your system.

 

[Sprucemoose]

not sure what those numbers mean or where you get them, but just use realtemp and see what your core temps are.

Just to clearify where the temps come from: All water/air temps come from the Aquasuite software reading off the AquaComputer air/water sensors. I've posted the average of the last ten minutes after the temps have reached their steady state. The CPU temps are from RealTemp (I've registered the max temperature of the hottest core). The GPU temps are the hottest of the two cores registered in GPU-Z.

EDIT (for cross-post): I don't think that you're grasping what you need to be looking at. Post core temps, not water temps, the resolution on your sensors is almost undoubtedly not enough to provide 1 decimal place of accuracy. Look at the actual component temps, as those are the sensors with the highest resolution (probably) in your system.

I checked the CPU core temperature using Realtemp running the game with different pump speeds. No matter what pump speed (flow at 1,2/0,5/0,3 GPM), the hottest core max out on 60 degrees.

 

[m0r7if3r]

I checked the CPU core temperature using Realtemp running the game with different pump speeds. No matter what pump speed (flow at 1,2/0,5/0,3 GPM), the hottest core max out on 60 degrees.

That's rather odd...how about GPU?

 

[Conumdrum]

Heat pump dump. Water temps will go up with more heat from the pump. Flow rate for the parts you have is not making a big difference? I'll dig deeper into the posts later. Cooking a wonderful meal for the wife.

Orrrr, just get rid of the sensors, problem solved.