Kioxia and WEKA: The Epic Quest to Calculate Pi to 300 Trillion Decimal Places
When it comes to groundbreaking technology and record-breaking calculations, few endeavors are as exhilarating as the quest to compute the digits of Pi. Linus Tech Tips, a Canadian YouTube channel led by the dynamic duo of Linus Sebastian and Jake Tivy, has teamed up with industry giants like AMD, Micron, Gigabyte, Kioxia, and WEKA to achieve an astonishing milestone—calculating Pi to a jaw-dropping 300 trillion decimal places.
The Collaboration Behind the Record
In a recent video, Linus and Jake reveal the intricate details of their record-setting journey, which has officially secured a spot in the Guinness World Records. This calculation showcases not only their passion for technology but also the relentless pursuit of challenges that push the boundaries of computational capability.
From left, Linus Sebastian and Jake Tivy
The Mind-Boggling Setup
The ambitious project was executed using a sophisticated setup featuring nine Gigabyte storage servers alongside a single compute node, all seamlessly running on Ubuntu 22.04.5 LTS. The core of the operation relied on Y-cruncher, a memory-intensive application designed to harness the power of external storage as RAM swap space.
Exquisite Technical Specifications
- Storage Servers: The powerhouse behind this feat consisted of 1 RU Gigabyte R183-Z95-AAD1 systems, each armed with dual AMD EPYC 9374F CPUs and ~1 TB DDR5 ECC memory.
- Network Capability: With dual ConnectX-6 200 GbE network cards, the storage servers boasted an impressive bandwidth of 400 GbE.
- Storage Solutions: Utilizing Kioxia NVMe CM6 and CD6 series SSDs, they achieved a staggering 2.2 PB of storage, spread across 32 x 30.72 TB CM6 SSDs and 80 x 15.36 TB CD6 SSDs.
The Compute Node
The compute node, a Gigabyte R283-Z96-AAE1, was equipped with dual 96-core EPYC 9684X 3D V-Cache CPUs and a mind-blowing 3 TB of DRAM. This powerhouse setup was crucial for optimizing calculations and minimizing latency to deliver an unparalleled computational experience.
WEKA: The Secret Sauce
To further enhance their performance, the team employed WEKA’s parallel access file system software, enabling a single file system across all servers for optimal interaction with Y-cruncher. This sophisticated arrangement allowed them to efficiently manage the massive data flow and keep latency to a minimum.
Technical Nuances
- Configured Containers: Tivy configured two 18 GB WEKA client containers, each assigned to 12 cores, maximizing cache utilization and ensuring that Y-cruncher’s buffers remained within the efficient limits of the CPU.
- Throughput Achievements: WEKA achieved impressive write and read speeds, with write throughput exceeding 70.14 GBps and read speeds reaching up to 122.63 GBps.
Avoiding cross CPU memory transfers
A Thrilling Journey
The quest for Pi began on August 1, 2024, but faced its share of challenges, including a multi-day power outage. The team persevered through power cuts and air-conditioning failures, ultimately completing the calculation at 191 days. And the big reveal? The 300 trillionth digit of Pi is 5!
A Glimmer of Future Possibilities
In a world where technology constantly evolves, one can only wonder what the future holds for the calculation of Pi. Perhaps a 1 quadrillion decimal place Pi calculation is within reach, should there be a shift to GPU-powered servers.
A Look Back at Pi Calculation History
The journey to compute Pi has a fascinating chronology:
- 250 BCE: Archimedes approximates Pi (~3.142857).
- 5th Century CE: Zu Chongzhi calculates Pi to 7 digits (3.1415926), holding the record for nearly 1,000 years.
- 1949: The ENIAC computer achieves a groundbreaking 2,037 digits calculation.
- 2022: Google Cloud sets a record by computing Pi to 100 trillion digits.
For a more comprehensive timeline of Pi calculations, check the complete chronology here.
Conclusion: The Final Word
With the remarkable collaboration between Kioxia, WEKA, and Linus Tech Tips, the record for Pi calculation has reached dazzling new heights. The realm of computational prowess continues to inspire innovation and ambition, making us wonder what records will fall next. Will GPU servers lead the charge for even higher calculations? Only time will tell!
As tech enthusiasts, we stand on the cusp of a thrilling frontier—glory awaits!
