Overclock.net banner
21 - 40 of 121 Posts

·
(not)New to Overclock.net
Joined
·
853 Posts
Quote:
Originally Posted by Moonmanovich View Post

According to my testing it is about 0.8°C in favor of 90° CCW.
Thanks! I guess I won't be disassembling my loop anytime soon, after all...
biggrin.gif
 

·
Premium Member
Joined
·
2,332 Posts
delta T is core temp minus water temp.

Thanks for testing! Interesting to see Supremecy tested better than Koolance on one cpu, but near same on the other. I dont think platform differences matter as much as different IHS and different samples of waterblocks. Have seen just as much variance (including my own testing) within similar cpus as across platforms. Not surprising that Sten got Koolance better vs vice versa in this test. Two different samples of waterblocks, slightly different bows, different tims, different mounting pressure, different IHS shapes which make a huge difference as to how the bow affects things.

Nonetheless you can get see the ones that tend to perform best across wide variety of cpus.
 

·
Registered
Joined
·
229 Posts
Quote:
Originally Posted by Moonmanovich View Post

I found it heating a piece of pure copper bar by inserting one point in hot water and the piece is completely heated.
After getting one point of the copper bar in cold water, it cools completely.

In this case, I think the block orientation fits depends on how comfortable each to build your loop as being vertical or horizontal, the block warm and cool in the same way as explained in the example.











 

·
Premium Member
Joined
·
4,967 Posts
FYI, I found orientation preference typically has more to do with mechanical TIM contact and how the bases are bowed much more than fin/pin or nozzle arrangement.

The proof and result is always in looking at your TIM paste post block removal. Some blocks are bowed mostly in one direction while others are two dimensional bows. You want that thin TIM contact patch to align with core direction. Also larger cores like 2011 prefer less bowing than smaller. Example:




Raystorm copper for me did poorly on 2011 because the base is bowed too much.
IMO, mechanical TIM contact and base bowing is key, not the nozzle or fin orientation.

Of coarse test best assumes the average user understands that AND willing to compromise performance over visual. Most will likely be clueless and install with the block name lettering so it's easy to read or if there are ram conflicts whichever way works.

The ideal methodology probably would take an average, or use the most likely used orientation, not necessarily the best but it's always debatable.

As mentioned above processor IHS and block sample variances can have more affect which we just can't test for in one test. You would need 10 block samples AND 10 processor samples to rule anything out which isn't possible in hobby level testing. Only the manufacturers have that sort of luxury to test sample variances out.

IMHO any block within 1 degree is statistically insignificant and equal.
 

·
Registered
Joined
·
2 Posts
Thanks for your efforts, such are always very welcome. I'd suggest, however, that readers might do well to re-plot the provided data against an origin of zero:

ek Sup 52.58 1.00
uc-2 lt 53.03 1.01
Swif HD 54.64 1.04
Swif XL 58.01 1.10
ek LTX 59.68 1.14,

doing so puts the ratings in better perspective. For my part, I far prefer the mounting (and remounting to get things, "just right") ease of SwifTechs in my (many) machines and happily accept the small hit which is vastly offset in my, admittedly, over-radiatored builds. The CPU thermal contributions pale in comparison to that of the multiple GP-GPUs.
 

·
Registered
Joined
·
229 Posts
Quote:
Originally Posted by Moonmanovich View Post

Eze2kiel: I don't think I understand what you are trying to say
redface.gif
Excuse me.
Sorry for my bad English.

I found out that putting the square of a pure copper bar in hot water, the whole piece heats.

On the contrary, put the square of the same piece in cold water, the whole piece cools completely.

To this I meant with respect to you comment of the block orientation:
Quote:
Originally Posted by Moonmanovich View Post

 

·
Premium Member
Joined
·
4,222 Posts
Quote:
Originally Posted by Eze2kiel View Post

Excuse me.
Sorry for my bad English.

I found out that putting the square of a pure copper bar in hot water, the whole piece heats.

On the contrary, put the square of the same piece in cold water, the whole piece cools completely.

To this I meant with respect to you comment of the block orientation:
Quote:
Originally Posted by Moonmanovich View Post

Hello and thank you for your interest, questions and comments!
thumb.gif

The best orientation for water block is when the micro-channel is parallel to the core. Picture bellow better explains it:


I believe it may have something to do with the CPU die not being in the center of the CPU package. Maybe this is not the case but these were the best orientations according to my testing.
The mounting appears to be producing less than satisfactory mounting (clamping) force. Not so much on the LGA-1150 but on the LGA-2011 this is quite noticeable. Will take a photo of the TIM imprint soon.
Hello Martin!

The water blocks were tested using their 'box' mounting mechanisms. The results reflect what one receives when one purchased a certain water block.
I forgot to mention it (rarely do such reviews) - the TIM used was test standard - Arctic Silver Ceramique 2. I have loads of it
biggrin.gif

According to my testing it is about 0.8°C in favor of 90° CCW.

Edit: typos.
Well, "completely" is misleading because the fact is the temperature will not be uniform. It will be coldest/hottest at the point of the thermal source and the opposite at the point furthest away from it with a gradual change in between. More importantly, with a waterblock you don't just have one thermal source. You have the heat on one side and the cool on the other competing against each other. That's where the the micro-channels and their orientation to where most of the heat is coming from matters. The micro-channels create more surface area for the cooling side and cause turbulence causing more of the molecules of coolant to come into direct contact with the block than if the same amount of flow just went smoothly by.
 

·
Registered
Joined
·
130 Posts
Quote:
Originally Posted by Eze2kiel View Post

Excuse me.
Sorry for my bad English.

I found out that putting the square of a pure copper bar in hot water, the whole piece heats.

On the contrary, put the square of the same piece in cold water, the whole piece cools completely.

To this I meant with respect to you comment of the block orientation:
This is pretty easy to rationalize if you recognize (as Martin did) that the issue has more to do with the bowing of the base than the true direction of flow across the core. If you're making better contact, you're allowing more heat to flow into the cold plate, which can then be conducted across the cold plate and ultimately removed by the water.
Quote:
Originally Posted by Unicr0nhunter View Post

Well, "completely" is misleading because the fact is the temperature will not be uniform. It will be coldest/hottest at the point of the thermal source and the opposite at the point furthest away from it with a gradual change in between
True, IIRC the strength of the gradient should depend on the thermal conductivity of the material. A good heat conductor such as copper may only have a small gradient across it. I think that is ezekiel's point, that the position of the water flow itself will be small due to the high conductivity of the copper basically equilibrating the temps across the cold plate. I think he is probably correct (at least to some extent, i'm not sure) but the bowing means that contact is the deciding factor, and that there is in fact a reason to put some thought into the orientation. This could probably be tested on a fully lapped/flat block
 

·
Banned
Joined
·
2,388 Posts
So the micro channels should run parallel to the die using an EK EVO and a 3770K.

That's the CCW, ok slightly better thermals ok.

Wasn't 90degrees from manufacturer's recommended what stren was calling "goofy" mount?

Which gave slightly worse results?

Just triple checking since I'm building tomorrow wanna make certain.

Thank you awesome work Moon.
 

·
Premium Member
Joined
·
2,332 Posts
The only way you will know which is better rotational mount wise is test it on your own waterblock and IHS. Usually only 1-2C difference or so, unless IHS/block have unusual shapes or tester is using light pressure. But you can look at the tim footprint both ways and get an idea. But after testing numerous waterblocks myself with calibrated probes on multiple cpus, including 2 of same cpus (since wife has computer and can swap out hers), what matters is IHS shape of cpu and bow/shape of waterblock, and which orientation makes the best footprint. Martin testing showed same.

And often I get the opposite results of other testers, but that is expected since Im testing a different IHS/waterblock. So just use whatever rotation suits your build best, or try testing yourself if you have accurate enough equipment. You can use your gpu temp as water temp, it is accurate to within about 1C and try measuring that way as well.
 

·
Premium Member
Joined
·
17,355 Posts
so on the mainstream 4 core haswell chips it didnt really show a difference between the supremacy and the supremacy evo, but on the enthusiast platform the difference was noticable. Thats enough for me to buy a upgrade kit for my supremacy.

5960x is ready to go, just waiting on some last pieces, this being one of them.
 
21 - 40 of 121 Posts
Top