Intel's next-generation Falcon Shores GPU is going to be a power-hungry beast as revealed to Computerbase during ISC 24.
During ISC 24, Intel and its partners happened to have teased the power consumption figures for the upcoming Falcon Shores GPUs which will be the follow-up to Gaudi 3. While the Gaudi line of accelerators has been dedicated to the AI segment, Intel seems to have taken a step back with its standard HPC & AI GPU offerings. Recently, we reported how Intel has ended the deployment of its first true HPC GPU, Ponte Vecchio.
Now, all eyes are set on Falcon Shores which will be the first true GPU IP based on the next-gen Xe graphics architecture. While we are still waiting to get the exact details on Falcon Shores, the GPU has seen vast changes from its original design.
Originally planned as an XPU which was going to be a combination of both x86 CPU and GPU cores similar to AMD's Instinct MI300A APU, the product was later downgraded to a GPU-only design. Intel has so far confirmed that Falcon Shores will take the best of Gaudi's AI prowess and combine it with the next-gen Xe graphics architecture for HPC & additional compute-heavy workloads. The GPU will also be backed by a strong software ecosystem in the form of oneAPI.
Based on the information from ISC 24, it is reported that Intel's Falcon Shores will feature a TDP of up to 1500W. This TDP is much higher than NVIDIA's flagship Blackwell AI GPU configuration. Those are rated at 700W, 1000W, and up to 1200W. Even AMD's Instinct MI300X and MI300A accelerators are rated at 750W and 760W, respectively.
With such high TDPs, the cooling requirements of the Intel Falcon Shores GPUs are also going to be quite extravagant. It is reported that Intel and its partners are
Read more on wccftech.com