Intel has confirmed the existence of their Xe HPG Graphic cards during Intel Architecture Day 2020. One of the benchmarks showed a Monster 42 TFLOPs 4 Tile ‘Arctic Sound’ Xe GPU with 16,384 Cores, running at a whopping 1.3 GHz clock speed. Intel benchmarked the card running at 1.3 GHz with a resounding 42 TFLOPs of FP32 compute and near-perfect x3.993 scaling on a 4-tile MCM GPU. The main target of Xe GPU
Intel’s Senior Vice President Raja Koduri confirmed that the company is working on 1 Tile (512 EU, 4096 cores), 2 Tile (8192 cores), and 4 Tile (16384 cores) versions of the Arctic Sound GPU, which will be based on the HP architecture. He said that the drivers and revisions of the chips are currently at their starting stages, but they still managed to run some benchmarks to show off their product.
The benchmarks showed that a single tile of Xe HP chip can transcode 10 separate streams of 4K 60 HEVC content – the 4-Tile version should be able to transcode 40 different streams at the same time. The Xe HP will use HBM2e memory, has 64-bit floating-point support and Tensor cores. The latter two are common features used in data center compute, deep learning, and AI environments. HBM2e will provide a bandwidth of 307 GB/s, which is huge compared to 72 GB/s of GDDR6.
For the gaming side, Intel Xe HPG will have the FP64 support removed or limited and will use GDDR6 memory in them. Intel also confirmed that Xe HPG will support hardware ray tracing. This will make them compete with the upcoming NVIDIA’s Ampere series and AMD’s Big Navi cards, both of which will support Ray tracing.
All these benchmarks bode well for Intel. If they manage to get ahead of NVIDIA and AMD or even come level to them, then it will be the best for consumers as more competition will mean better products coming out. Intel has announced that these cards will release sometime in 2021, so we still have some time to see what Intel will be competing against, with the release of Big Navi and Ampere cards just around the corner.
