Overview / Usage

My dream PC would help to have a successful college experience as a machine learning student. I currently have a MacBook Air with M1. The computer performs well for everyday needs. However, as a developer I have struggled setting up my projects, even using Rosetta the user experience was frustrating. For my recommender systems project, I needed to request a cloud hosted machine to complete my tasks as a research assistant. My dream PC would help any Computer Science student with a dream to run their data in minutes, instead of having to run their projects overnight.

Methodology / Approach

For this deployment I have laid out an ideal list of components and given my rationale for each of them.

CPU: Intel Core i9-12900K

The i9-12900k is the top performing CPU across a broad spectrum of workloads, the new BIGlittle architecture also opens new possibilities when paired with the new scheduler in Windows 11 to keep inconsequential background process out of the way of work and while providing the highest possible core clocks on any CPU on the market today. Additionally it's support for the latest PCIe Gen 5 devices and DDR5 memory give it incredible I/O and data throughput capabilities.

OS: Windows 11 Pro

Windows 11 is the best OS to use for this deployment as the implementation of Thread Director as well as other kernel optimizations give it the best chance to take full advantage of the BIGlittle architecture on the 12th Gen Alder Lake processor I have selected for this system. Additionally Windows has the deepest back catalog of software, and being the most used OS on earth it is always a priority for developers, be it bug fixes or new features Windows will get them first from most if not all major software companies. Additionally with Windows Subsystem for Linux I can gain access to the powerful  CLI and tool set available from linux without having to be subject to the performance implications of their kernel's interaction with the new CPU architecture.

CPU Cooler: Noctua NH-D15 Black Edition

Noctua makes some of, if not the best and most reliable fans and coolers on the market, the NH-D15 has proven itself to be an exceptional cooler rivaling even 360mm CLC liquid coolers while having far fewer points of potential failure and maintaining very low noise levels due to it sheer size and very high quality fans.

RAM: Crucial DDR5-4800Mhz 64GB (2x32 GB)

DDR5 has been shown to provide a notable improvement for production tasks like machine learning and code compilation, as this machine focused is around those workloads a fast kit with a high capacity is very important. Crucial is owned by Micron who has a proven record of being a manufacturer of very high quality memory, in addition to being one a very short list of companies that fabricates the DRAM packages. 

**MotherBoard: **Asus ROG Maximus Z690 Hero

Asus has a track record of making very high quality motherboards with exceptional feature sets, as well as a very configurable and user friendly BIOS/UEFI interface for tuning performance.  While the 3080 TI selected for this system only supports PCIe Gen 4, being on PCIe Gen 5 board now ensures the longevity of this system and when a device that can take advantage of PCIe Gen 5's full bandwidth arrives to market I will be able to take advantage of those performance gains without needing to upgrade the motherboard. Additionally this board has a very robust and well cooled VRM and power delivery system which will allow the 12900k to draw as much power as it needs to achieve its full potential. This motherboard also has a plethora of I/O available, including 2 Thunderbolt ports,  Wifi 6, and 2.5 Gb Ethernet. 

**Storage (Boot): **Intel Optane SSD P1600X Series 118GB

The boot drive will be an Optane SSD module, as 3DXpoint provides incredible performance for the kind of file operations that occur when using a storage device for an operating system. While the Optane drive only supports PCIe Gen 3.0 x4 the access latency and random read write performance is unparalleled and this drive's only job will be handling windows and applications, with data storage for active projects handled by devices that are optimized for sequential performance. While the Optane drive provides uncompromising random performance for the OS. 

**Storage (Scratch): **2 TB Samsung 970 Pro NVME M.2 X 2 (RAID 0)

Active projects and file storage will be on a pair of Gen 4 x4 M.2 SSD's configured in RAID 0, this gives the system extreme sequential read and write performance while having sufficient space for a large project or multiple concurrent projects. This array is meant for Active projects, closed projects or those that have become inactive will move to the Bulk storage array.

Storge (Bulk): 2 X 8TB WD Black (RAID 1) 

The Bulk storage array is a pair of WD black 8TB drives in RAID 1 that will provide storage for closed and inactive projects as well as supplementary bulk storage for any exceptionally large files. A single drive of redundancy provides some protection from a drive failure as unlikely as it given WD's track record, however to provide a proper backup I will be using duplicati, and windows fork of the tool duplicity from linux to handle nightly backups of both arrays, so that in the event of hardware failure there will be a point from within the last 24 hours that I can restore to.

External Backup: WD Elements 14 TB (Backup for Bulk RAID Array and the Scratch Array):

For this deployment I am going to use an external hard drive to hold backups of the Scratch and Bulk storage arrays, I have chosen a 14TB external drive from Westen Digital for this to allow for a proper differential backup of both arrays up to their full size with room to spare, as by the time I have accumulated that much data another solution such as a NAS will have become necessary. Additionally with that many projects it may also be time to look at geo-replication to protect the work in the event of a natural disaster.

Graphics Card: EVGA GeForce RTX 3080 TI  - FTW3 ULTRA iCX3

For my work having dedicated AI and ML hardware is very important and the current generation Nvidia graphics cards are the undisputed champions at the consumer level. The 3080 TI has nearly the same performance as the 3090 but does not have the same memory thermal issues that the 3090 presented as most card designs do not have sufficient cooling for the memory modules on the back of the card. Additionally from my research 12 GB of GDDR6X should be entirely sufficient for my workloads. The EVGA FT3 series has been one of the best implementations of coolers for graphics cards for several generations and provides exceptional cooling while maintaining very low noise levels, additionally the iSX sensors on the card provide and very detailed picture of the thermals on the card and ensure that all components are running at the optimal temperature.

Case: Corsair 5000D Airflow (Black)

The Corsair 5000D Airflow is an excellent enclosure for this build, as it can accommodate the massive Noctua NH-D15 CPU cooler and supports the additional drives I need. The case ships with a basic complement of fans, however due to the thermal demands of the components inside I am going to replace the stock fans with 4 of Noctua’s pressure optimized NF-A12x25 (Chomax Black Swap) fans. These will allow the system to maintain optimal temperature while keeping noise levels reasonable. With a 3 fan intake and 1 fan exhaust the case will have a positive airflow pattern, and thanks to the 5000D Airflow’s dust filters the case system can be protected from dust while still having exceptional airflow.

**Power Supply: **SeaSonic PRIME Platinum 1300W

Seasonic makes stealler power supplies backed by an exceptional warranty, 1300 watts is appropriate for this build as the 3000 Series Nvidia cards have become somewhat notorious for their immense power draw, and a full load on the CPU & CPU would be a load that a lesser unit may struggle with, which was the motivation behind a unit with such a high capacity. Additionally the exceptional efficiency of the unit will keep the power draw of the system during minor tasks and idle time lower.

Displays: Dual LG UltraFine 32UN880-B (aka LG Ergo Display 32”)

For the displays, I have chosen 2 4k 32" in displays from LG as they provide enough screen space for my work while being at a size where 4k makes sense and allows me to have as much screen space as I could need without one of the displays going to waste by only showing static content. Additionally the exceptional adjustment range of the ergo stand allows for quick changes to the monitor’s orientation and position.

**Mouse: **Logitech MX Ergo

I have a history of wrist and shoulder pain and to mitigate this I am going to use Logitech's MX Ergo Mouse as the trackball style of the mouse puts far less strain on the wrists and shoulders than a traditional mouse or trackpad.

**Keyboard: **Keychron K8

The compact size of the Keychron K8 is well worth the lack of a number pad for me. As it would allow me to keep my hands in a more natural position when switching from mouse to keyboard and back. Also the smooth low effort actuation of the Gateron Red switches would allow me to type comfortably for extended periods of time. The compact layout is also very familiar to me as I do most of my work on a laptop currently.

​

This is an ideal configuration for what I need to do, truly a Dream PC for any Computer Science student.

Technologies Used

Intel Core Processors: Core i9-12900K

Intel Optane SSDs: P1600X Series SSD

​

​