Creating a Custom Waterblock-Updated version
Creating a Custom Waterblock-Updated version
My worry about touching the heat sink surface was limited to the outer flat part that connects to the CPU. Inside the passages (where liquid moves) the main issue with a rough surface is a slight increase in pressure drop. But rough surfaces can help maintain turbulent flow at lower Reynolds numbers. Therefore, it might be advantageous to leave any internal passages with tool marks from milling.
I was somewhat inspired by Cather and this discussion: Because it's summer and I have limited activities, along with the Intel Science Fair and the Simens competition, I thought why not? I’d be able to gain some knowledge (useful later) and maybe even earn a little money (winning regional contests, etc.).
I think Wick intended both the internal and external aspects. Indeed, any finishing above 1000 grit is unnecessary since the TIM (especially components like AS5) can't reach the gaps. For external areas such as IHS-waterblock, use Ceramique or T-C Grease 0099 from TIM Consultants, or even a standard white silicone paste. These products spread more evenly when crushed, ensuring most contact remains metal-to-metal rather than filled with TIM. With my application, both IHS and waterblock were lapped to 1200 grit; using Ceramique with just half the grain size removed almost eliminated excess paste, meaning less was needed overall. The ideal finish needs no TIM at all, which seems to be reserved for high-end laboratory equipment.
Thicker pastes like AS5 (or the extremely thick Diamond7) are best suited for unlapped surfaces intended for regular users and are difficult to apply thinly without applying intense pressure.
For your square pin matrix—whether milled or forged?—the surrounding design and cooling enclosure also affect turbulence. A good example is comparing Dtek FuZion v1 (my model) with v2. Dtek reduced the pin matrix height by nearly 2mm and made minor adjustments to the plastic chamber above it, resulting in increased turbulence (about 2C improvement on average with an OC'd quad) and slightly higher flow resistance.
The eventual product will be milled using Copper C110, not forged, so I need to consider the constraints of CNC milling in the design. Thanks for the details about the TIM wuzy. The 6x6 square pin matrix was only simulated to establish a baseline for comparison. The simulation lasted about 95 seconds (~45 minutes in real time). I plan to conduct a longer test (10–30 minutes, which would be several hours) over the weekend. I have access to multiple PCs, allowing me to set up the fluid sim on those without installing Inventor (student license, no restrictions) plus Thermal Desktop (trial) on each one.
I intend to build a few prototype blocks like this:
Primarily to gain a clearer understanding of jets and similar aspects.
As the top will be a rapid prototype, from a design perspective, the possibilities are quite broad—compared to traditional methods like CNC milling or forging. I could potentially create complex designs (such as multiple flow channels within the top) that route water from the inlet to the Copper base and then out the outlet.
This inquiry checks if your inlets and outlets are included in the simulation. You described a layout similar to a CPU—center and corner placement, like GPU blocks. The prototype you mentioned resembles a diagonal crossflow design, with corners connecting. Regarding your suggestion, consider maintaining the central impingement pattern across the 2/4 core area, while enlarging those near the edges or lowering their height to improve flow after passing through the center to the outer sections of the block.