Tesla details how it finds punishing defective cores on its million-core Dojo supercomputers — a single error can ruin a weeks-long AI training run

Tesla's Stress tool detects and disables faulty cores in Dojo wafer-scale processors, which power Dojo clusters with millions of cores, without interrupting AI training.

https://www.tomshardware.com/pc-components/cpus/tesla-details-how-it-finds-punishing-defective-cores-on-its-million-core-dojo-supercomputers-a-single-error-can-ruin-a-weeks-long-ai-training-run

Utworzony 3d | 7 cze 2025, 14:10:05


Zaloguj się, aby dodać komentarz

Inne posty w tej grupie

 Nvidia's 20-core N1X leaks with 3000+ single-core Geekbench score — Arm chip could rival Intel and AMD's laptop offerings

A preliminary Geekbench listing for Nvidia's future N1X SoCs for mobile devices suggests performance near Intel's Arrow Lake-HX and AMD's Strix Halo chips.

10 cze 2025, 13:50:08 | tomshardware.com
 Starlink shut down on United regional jets — Reports of radio interference on Embraer E175 aircraft prompt switch off

United has turned off Starlink on its regional Express service following reports of radio interference.

10 cze 2025, 13:50:07 | tomshardware.com
 Getting To Know the Linux Filesystem — Demystify the directories

It may look intimidating, but the Linux filesystem isn’t a maze of dead ends. Here’s the map that you need to learn more about an OS that runs on computers big and small.

10 cze 2025, 13:50:07 | tomshardware.com
 This giant 42-inch OLED gaming monitor from Asus hits an all-time low price at Amazon

Save 29% on this 42-inch Asus ROG Swift PG42UQ OLED gaming monitor

10 cze 2025, 13:50:06 | tomshardware.com
 'Our single chips still lag behind the US by a generation' — Huawei CEO says Washington has exaggerated its achievements, saying the company isn't that powerful

Huawei CEO Ren Zhengfei said in an interview on the People's Daily that its Ascend chips are a generation behind U.S. tech.

10 cze 2025, 13:50:06 | tomshardware.com
 FSP MEGA TI 1650W Titanium power supply review

The FSP Mega TI 1650W Titanium PSU offers extreme-efficiency power delivery for the most demanding systems, featuring premium performance and an exceptional thermal desig

10 cze 2025, 13:50:05 | tomshardware.com
 1.5TB of James Webb Space Telescope data dumped on the internet — new searchable database is the largest window into our universe to date

NASA-backed astronomy project has just released the 'largest look ever into the deep universe.'

10 cze 2025, 13:50:05 | tomshardware.com