Apogee test done at System Cooling

[seamadan000]

Fair enough, sounds like a good idea to me.

Just a question, what else matters, other than performance and max OC...

Thanks for the advice.

 

[IYIENACE]

As much as I'm liking the Storm block, I have to admit I'm very pleased with how my old RBX is still running up there with the big dogs.

 

[Les56]

..
By using die sims of known size, and known even thermal output spread and applied to the base of the waterblock, and by calculating the temperature rise of that die sim we can arrive at a fairly confident approximative value for h. ....

For your Storm G4 you infer a value of ~77000 W/m²K for "effective thermal convection transfer efficiency" at "6W of hydraulic pumping power" (Link) .
This would clash with my 170,000-240,000W(link for h(eff); h(eff) defined here.
So I guess "effective thermal convection transfer efficiency" is h(convection).
To arrive at a value of h(convection) you have to model the waterblock.
Are you now in a position to reveal your model?
It can then can then be used, in conjunction with my model (link) to aid analysis of Incoherent's future data(link to analysis stagnation)

 

[Cathar]

Les, I was talking about h(eff). I work off my own test data. You're working off Robotech's. Two different testbeds, therefore two different results for h(eff). It's all good so long as the testbed is consistent.

 

[Les56]

Les, I was talking about h(eff). I work off my own test data. You're working off Robotech's. Two different testbeds, therefore two different results for h(eff). It's all good so long as the testbed is consistent.

No
h(eff) is testbed independent
Something is sorely amiss ..
My inferred h(eff)=170,000-240,000W/m²K and modelled h(eff) should apply to both Robotech's data and your "own test data".
Likewise your h(eff)=77000 W/m²K should apply to Robotech's data

 

[Cathar]

I don't know what to say Les. I can't, for the life of me figure out how you derived an h(eff) of 170-240K. Lee's die size is 14x14mm, or 196mm^2. Lee's calculated gross C/W is 0.132 (for Storm @ 4.5W). If we discount the TIM layer, (est. 0.032C/W for 196mm^2), we arrive at a C/W of 0.100 for the 196mm^2 die size. h(eff) in that scenario would then be around the 37000 mark.

Heck, even if we were assuming a straight up size for size convectional area for the heat-source (i.e. a 14x14mm waterblock), and with a 0.1mm copper base-plate, h(eff) at best would be ~50000.

Unsure how much of Lee's values are modified to account for the TC offset (depth) from surface of the die sim. That would definitely throw out any predicted h(eff) derived from that test data, which is what I meant by testbed dependent. Or are we interpreting that paper's definition of h(eff) differently, which may be possible.

At this point, I'm pretty much confused by your values.

 

[Les56]

You are missing die sensor offset.
"Memory guess" at value I use ~0.06c/w for the 196mm^2 die.

Edit: For clarity;inserted "You are".

 

[Cathar]

You are missing die sensor offset.
"Memory guess" at value I use ~0.06c/w for the 196mm^2 die.

Meaning that Lee's thermal probe is located approximately 3/16" from the die surface? Seems quite a distance... Has Lee disclosed that publically anywhere?

[Edit: Looking at his die-sim here, it doesn't even look like there's 3/16" high of the 14x14mm raised section into which one could assume a linear temperature gradient between a thermal probe located 3/16" beneath the surface, and the die-sim surface. Any discussion on how the 0.06C/W die-offset value came to be?]

 

[Les56]

Meaning that Lee's thermal probe is located approximately 3/16" from the die surface? Seems quite a distance... Has Lee disclosed that publically anywhere?

Yes, implies ~ 4.5mm with 392 W/mK Copper.with uniform flux.
Have not seen any values published and have no private information.
Care should be taken in taking the "physical distance". As the analysis(eg link) for Incoherent's flux-bloc indicate "virtual distances" should be considered.

 

[Les56]

[Edit: Looking at his die-sim here, it doesn't even look like there's 3/16" high of the 14x14mm raised section into which one could assume a linear temperature gradient between a thermal probe located 3/16" beneath the surface, and the die-sim surface. Any discussion on how the 0.06C/W die-offset value came to be?]

"Any discussion on how the 0.06C/W die-offset value came to be?"

Best fit of MCW6000(196) model to Lee's 196 data..
The MCW6000 model has some validation with Incoherents Virgin MCW600 Resistance at both 100mm^2 and 140mm^2 (eg link.
Although model is based on hearsay dimensions(have been repeatably asking for real) and not guaranteed to work at 196mm^2 it is the most validated I have.

 

[Cathar]

Agree, "virtual distance" is going to be the key point here, especially since even at a 4.5mm-4.7mm vritual depth (depending on one's assumption of copper's conductivity) this would still physically estimate placement of the probe well beneath the "plateau" of the main die body, let alone the 14x14mm raised section.

Since we don't know the actual thermal die offset, nor its exact C/W corrective offset, then perhaps we have found the reason for our "difference"? Also need to factor in differences between 6W hyd. pwr vs Lee's 4.5W.

If we said at 6W hyd. power that the Storm was at 0.13C/W for Lee's testbed, minus TIM C/W, giving 0.098C/W. In this scenario, the thermal die offset C/W would need to be around 0.035-0.040 to be within 5% either side of my h(eff) estimates. Compared with your 0.06 offset C/W estimate, this is a small enough difference, but still well within reason, to possibly explain any perceived discrepencies.

 

[Les56]

......to possibly explain any perceived discrepencies.

The differences in predicted h(eff)s are very real and radically effect the cooling characteristics of a wb; as illustrated here and further pursued here and here

 

[Cathar]

The differences in predicted h(eff)s are very real and radically effect the cooling characteristics of a wb; as illustrated here and further pursued here and here

Let me just state that when I was referring to "impossible", I was referring to myself having never seen any technical paper ever adequately prove to have sustained an h(eff) of anything much more than 200,000, and even those ~200K values were obtained using around 250-300W hydraulic pumping powers. As it stands, I am still unable to accept that even the Storm could have anything like a 250K h(eff).

I don't know what the issue is. Something is awry. I still think that a 0.06C/W thermal die offset for Lee's testbed is too high, and that is the cause of any disagreement here.

Anyway, I'm off to bed.

 

[Les56]

The differences in predicted h(eff)s are very real ....

To ellucidate:
Your predicted h(eff)=77000 W/m²K is based on your own test data("I was talking about h(eff). I work off my own test data") and not Lee's.

 

[Les56]

.... As it stands, I am still unable to accept that even the Storm could have anything like a 250K h(eff).......

The 170-240K was obtained, for 0.1-1mm bp, by inference from Lee's data
My Modelling suggests the lower range is more appropriate for the Storm 4.
However would predict values of 250K for the MP-05-SP

 

[Cathar]

Storm's bp is substantially thicker than 1mm.

 

[Les56]

Storm's bp is substantially thicker than 1mm.

Any corrections to dimensions are welcome.and will remodel accordingly
Will be noted that was not happy with dimensions used: "Dimensions are sometimes assumed and/or hearsay:applies particularly to MCW6000 and Storm G4 "

A repost of original including attachments(resized)

"Snide's Model of Storm G4
Taken as: 35 Cups(3.1242mm Diam x 4mm Deep) in 4.7625mm Base with 0.7366mm(ID) Jets
Modelled as: 37Cups(3.1242mm Diam x 4 mm Deep) in 4.7625mm Base on a 20.7x20.7mm Area, with 0.7366mm(ID).
Cup walls modelled as hexagonal Arrays of (0.4or0.8) x3.1x4mm Pins on 0.7625mm bp Effective Base Area=21.5452x21.5452 mm = 464.195643 mm^2
Effective Cup Base Area=330.8391314 mm^2
Effective Cup Wall Area=1694.330101 mm^2

Notes on Modelling
All very crude
Applies to uniform flux sources and are average Temps and Watts
All need fudge factors for h(conv)
All need better correlation between h(conv) and h(eff) : particularly Storm G4
Dimensions are sometimes assumed and/or hearsay:applies particularly to MCW6000 and Storm G4

Edit: Corrected boob in StormG4 h(eff) profile . Stupidly, was showing the "Cupped Area" to be double its actual size"