Originally designed as an XPU that combined CPU and GPU cores, Falcon Shores was later redesigned as a pure GPU. This decision aims to combine the AI capabilities of the Gaudi range with the next generation Xe graphics architecture to handle both HPC and AI workloads. Falcon Shores is supported by oneAPI, a comprehensive software ecosystem that makes it easy for developers to harness GPU power. However, the chip’s high performance comes at the cost of high energy consumption. The TDP of 1500 watts requires liquid cooling, as air cooling is not sufficient.
Compared to the flagship Blackwell AI GPUs from NVIDIA, which consume between 700 and 1200 watts, Falcon Shores is significantly higher. AMD’s Instinct MI300X and MI300A accelerators with 750 and 760 watts respectively are also significantly more energy efficient. The question remains whether the performance of Falcon Shores can justify the high energy consumption. NVIDIA is currently well ahead of Intel with its AI GPUs. Gaudi 2 and Gaudi 3 have not been able to reduce this gap.
Although Intel offers a good price-performance ratio, total sales of Gaudi are expected to be only 500 million dollars in 2024, compared to several billion dollars for NVIDIA. Even AMD is expected to generate 7 times more revenue from its AI accelerators. Falcon Shores shows Intel’s determination to gain a foothold in the AI accelerator market. However, whether the chip will be as successful as Intel hopes depends on its ability to outperform the competition in terms of performance and energy efficiency. The next few years will show whether Falcon Shores can catapult Intel into the “big leagues” of AI.
Source: Computerbase
25 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Urgestein
Mitglied
Mitglied
Urgestein
Veteran
Mitglied
Urgestein
Urgestein
Veteran
Mitglied
Urgestein
Urgestein
Veteran
Mitglied
Mitglied
Urgestein
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →