Texas Instruments Interview Presentation

34
THE CLOVER POT CUSTOM LIQUID NITROGEN GPU COOLING CONTAINER Chance Coats

Transcript of Texas Instruments Interview Presentation

Page 1: Texas Instruments Interview Presentation

THE CLOVER POTCUSTOM LIQUID NITROGEN GPU COOLING CONTAINER

Chance Coats

Page 2: Texas Instruments Interview Presentation

Personal Introduction

Page 3: Texas Instruments Interview Presentation

Building Computers I assembled my first computer in 2009. Incremental upgrades annually since

then. Saving money lead me to push my

components as far as possible. Overclockers.com became my main

forum in early 2009 and I began overclocking.

After joining the Benchmarking team in 2011, I have been competing online ever since.

Page 4: Texas Instruments Interview Presentation

Benchmarking and Extreme Cooling Benchmarking results are tracked on

“HWBot.org” Extreme cooling is not required to compete. I felt a strong desire to push my hardware even

further past typical limits. Ambient cooling no longer allowed me to reach my goals.

I purchased my first CPU container and used dry ice.

After gaining enough experience with dry ice, I switched to liquid Nitrogen for cooling.

Page 5: Texas Instruments Interview Presentation

Why use Extreme Cooling ? As I pushed my devices further, power

consumption and temperatures quickly got out of control.

Sub-ambient cooling methods allow for greater power dissipation, lower operating voltages, and most importantly higher clock speeds.

Page 6: Texas Instruments Interview Presentation

Building my own Container CPU containers are extremely useful, but

GPU containers offer increased performance results for graphics cards when using extreme cooling.

Commercially available GPU containers are expensive and difficult to find in good condition.

My advanced drawing class offered me access to heavy duty machining equipment.

I enjoy getting my hands dirty and working through projects.

Page 7: Texas Instruments Interview Presentation

Clover Pot – Design Process

Page 8: Texas Instruments Interview Presentation

Hand Drafting My initial designs were all done by hand

using pencil and paper. The first draft was very far from the

finished product.

Page 9: Texas Instruments Interview Presentation

CAD Design Process My hand sketches went through multiple

revisions as I learned about the machining process and its limitations.

After finding a general design, I moved onto CAD drawing using AutoDesk AutoCAD.

Once switching to electronic drawing, I was allowed more freedom to explore base designs and could visualize the final product in 3D.

Page 10: Texas Instruments Interview Presentation

Creative Design Process Trial, error and testing lead to my final

design. Early revisions would change as often as

everyday.

Page 11: Texas Instruments Interview Presentation

The Clover Pot was born My final design revision gave me a

strong sense of achievement. I had created a unique design with strong performance analysis results.

Page 12: Texas Instruments Interview Presentation

Intermediate Design Step Because school gave me access to a 3D

printer, I was able to take my design process one step further after finalizing. The design was created at half scale using extruded ABS plastic.

Page 13: Texas Instruments Interview Presentation

Design Choices A two-piece design was chosen for cost

reasons. Copper was used as the base material to

offer superior thermal conductivity near the GPU core. The increased specific heat of Copper also offered better load handling properties.

Aluminum was used for the top of the container. Lower cost allowed the container to have a 2-3 fold increase in volume with a slight increase in cost.

Page 14: Texas Instruments Interview Presentation

Machining Process

Page 15: Texas Instruments Interview Presentation

Materials and Tools After finalizing my design, I purchased

the raw materials and specialized tools necessary to machine my design.

Page 16: Texas Instruments Interview Presentation

Design Specifications and Precision I set lofty goals for myself when

machining this project. Having no prior experience using a 3-axis mill, I knew the learning curve would be steep.

I set a goal of ±0.01’’ tolerances for my general machining, and ±0.005’’ tolerances for the mating surfaces between the halves of the container.

I met my goal for most of the project, but unfortunately did not meet these specifications for all cuts. In the end I was very happy with the results.

Page 17: Texas Instruments Interview Presentation

The Container Top For reasons including material cost,

material hardness and design complexity, I chose to start with the top half of the container.

The general process of machining includes facing and squaring materials before any cuts are made.

Raw materials are often close to square, but not close enough for precise machining.

All materials were faced to remove major surface imperfections and squared for precise machining.

Page 18: Texas Instruments Interview Presentation

Container Top Machining The top half of the container is

essentially a hollow metal tube made to mate with the base of the container.

Page 19: Texas Instruments Interview Presentation

Base Machining After successfully completing the top

half of the container and meeting nearly 100 percent of my specifications, I moved on to machine the copper base.

This half of the container was considerably more complex and required finer measurements as well as nearly three times as long to machine.

Page 20: Texas Instruments Interview Presentation

GPU Core Contact Patch A key aspect of the base was the section

designed to make contact with the GPU core when in use. This section required a very smooth finish to provide optimal heat exchange, but was left rough for the better part of the machining process.

Page 21: Texas Instruments Interview Presentation

Internal Base Design Challenges

During the design and revision process, I decided on a relatively complex base design. This offered superior surface area while still maintaining a large amount of mass near the core contact patch.

Due to time and machine constraints, the design was simplified during the process.

Testing after completing the project has shown the container still offers strong performance and load handling characteristics.

Page 22: Texas Instruments Interview Presentation

Internal Machining Once the outside of the container was completed,

I began machining the internal surfaces and the surfaces used for mating with the top half of the container.

Page 23: Texas Instruments Interview Presentation

Final Internal Machining Seen below is the base just after internal

machining was completed. Visible in the holes is the cutting oil used

throughout the process to ensure smooth cutting and proper cutting head cooling.

Page 24: Texas Instruments Interview Presentation

Completed Internal Surfaces After many hours revising the original

design and machining the new features, I had nearly completed the base of the container.

The final steps of the base included drilling and tapping holes used to attach mounting brackets.

These brackets were designed for smaller aluminum plates and cut using a CNC machine.

Page 25: Texas Instruments Interview Presentation

Attached Mounting Hardware Once the brackets were cut and all holes

were tapped, the only remaining step was to mate the upper and lower sections of the container.

Page 26: Texas Instruments Interview Presentation

Clover Pot Results

Page 27: Texas Instruments Interview Presentation

Benchmark Testing After spending approximately 4 months

on this project split between design and machining, I was itching to test the finished container in the real world.

In the world of benchmarking, there are programs to test nearly all aspects of a computer. From CPU and memory to GPU and overall system performance, the list of benchmarks is incredibly long.

Page 28: Texas Instruments Interview Presentation

Aquamark3 Aquamark3 is a GPU specific benchmark

created in 2003 to test the rendering power of high powered graphics cards.

Despite its age, world records in this benchmark are still fought over.

After hours of pre-testing, I was ready to freeze one of the graphics cards I had purchased specifically for benchmarking.

Page 29: Texas Instruments Interview Presentation

Number 69 in the World

Page 30: Texas Instruments Interview Presentation

Really Cool Pictures

Page 31: Texas Instruments Interview Presentation

Hardware Setup Used I’ve always been fascinated by pushing

computer hardware to the limits. Results and pictures of my sessions provide a view into the insane world of competitive overclocking.

The hardware used for the previous results include:

Intel Core i7 processor and High Performance DDR3

Asus Motherboard and Nvidia Graphics Card

Page 32: Texas Instruments Interview Presentation

Mid Session

Page 33: Texas Instruments Interview Presentation

Aftermath

Page 34: Texas Instruments Interview Presentation

Thank you very much for this opportunity to tell you about myself.

Questions?

Thank you!