TDX vs MCW6000, OCers vs Procooling
 

I was looking at the performance difference between the TDX and the MCW6000 on OCers and noticed the TDX wins .136 C/W to .141 C/W. pH's results are the exact opposite showing the MCW6000 to win by quite a lot at 1gpm. What gives?

Different die sizes on the testbeds: ~130mm² for OC.com vs 84mm² for Procooling.

Too many variables to explain just why/how a certain design will work better with a larger die size without potentially confusing people or having them run away with a too-minimal understanding of the issues, suffice to say that the MCW6000 will favor smaller die sizes over the TDX for bare-die CPU's at the least.

Die size, secondary cooling, maybe the mount (Joe does one mount, I believe, and regulates the mounting, while pHaestus uses the stock mounting system and mounts a whole bunch of times)

While we're on this topic, does Joe insulate the top of his die sim (not the die itself, the surrounding area)? If not, things with small bases (TDX for example) will outperform things with large bases (MCW6000, for example)

Joe tested with the #4 nozzle on the TDX; my review was originally of the TDX with stock #1 nozzle. Even still MCW600x does better at 1GPM in my testing; maybe the 15lbforce JoeC uses for testing favors thinner bps?

it was my understanding that joec doesn't test each waterblock at 1GPM, rather, set up the bench loop at 1 GPM with no waterblock, and then each waterblock is tested, and flow drops in each according to each block's pressure drop characteristics.....thus, a more restrictive block would suffer from have a lower tested flowrate than a less restrictive one....

er, that doesn't seem right....

so, for clarification, JoeC has each of the waterblocks tested at 1GPM?

edit: surely the TDX has a thinner BP than the MCW-600x?

Think OC.com is 140mm²:-
To quote Bill - "JoeC told me yesterday the die area was 140mm²
sorry Les, should have posted" - from here

What you described was his very early procedure. After much protest when he changed pumps, he switched to testing at various data points, which is apparantly 0.5gpm, 1.0gpm, and 1.5gpm (or 0.3gpm if 1.5gpm is not achievable).

Cheers. It was always a little ambiguous.

oh, right, switched pumps...during the RBX testing, no?

been a while since I read OCers....

pH have you concidered moving to a barton core, or maybe even moving to an A64? Your setup seems a bit dated making real world decisions harder. I wonder about a 2.6Ghz mobile barton personally. :)

Edit: Ok now I know that you are working on a test bench that will include a die simulator. Will it be able to simulate multiple contact areas?

if i switch cores then i have to start over (~7 months work down tube). I am testing at a resonable power level (~70W). Same power joec currently uses and bill used to use. Barton temps would likely be a small bit cooler than TbredB.

Your info is still comparing apples to apples so I think you are right on. And although I often times don't know WTH you guys are talking about in here, your reviews and comparissons are great. Your one of the few opinions I trust.

It wouldn't be 7 months down the tube, you could retest the current few top performers and move on to new block tests as they come out. It's just that your current cpu has a small contact area and low wattage compared to modern P4's and A64's. Hammer 193mm, Newcastle 144mm, Northwood 146mm, Prescott 112mm. If the primary cause for difference between Joe's setup and your setup is the die size then I would say that it's important to move to something larger. Ideally three sizes like 80mm, 110mm, and 140mm should be used but that's a lot of work. 70W is also kind of low, most higher end cpus are >=90W especially the Prescotts.

By this time next year it seems likely everything will be using a heatspreader...

Razor6:
Why not ask me to test with AMD64, Prescott, Barton, AND TBred?
It's not going to happen. I could increase the power to 100W easily enough but it will change nothing of any importance other than I'd have to retest everything again.

I will retest currently offered commercial wbs on the die simulator whenever I get that working and move onwards and upwards at that point

All I'm hoping for is that the die simulator simulates more modern CPUs. I wasn't suggesting you go buy and test on several new mobos and CPUs because now I know the die simulator is on its way.

I could test at 50-150W and results (C/W) should be the same. Probably the die simulator will not simulate more modern CPUs any better than the TBredB does. Very different situation between localized hot spots in silicon (and with heatspreaders now) and an even copper die. I'll probably go with 100mm^2 or so