Intel’s upcoming flagship Xe-HPG DG2 Gaming GPU has made an appearance within the Geekbench database along with its specifications. The Xe-HPG lineup from Intel will be dedicated to gaming graphics cards and will be available by the 2nd half of 2021.
The GPU features 512 Execution units that round up to 4096 cores. The clock speed was maintained at 1800 MHz (max) but this might not be the final clock speed as the graphics card tested on Geekbench is still in an engineering state.
In the listings, we find that the card has a 12.6 GB memory capacity, but it seems like the benchmark is unable to properly identify the VRAM capacity and could just be 12 GB in total. The GPU also packs 4 MB of L2 cache.
To conduct the test, an Intel Alder Lake-S Desktop CPU platform which comprised of a 12th Gen processor featuring 16 cores and 24 threads was used, which is surprising. The Alder Lake-S Desktop CPU is clocked at 2.20 GHz base and 4.60 GHz boost clocks and is comprised of 32 GB DDR5 memory.
Coming to the OpenCL test result, the Intel Xe-HPG DG2 GPU is really bad and performs worse than even the GeForce 500 series graphics. But the reason for that is the tested unit was one of the first produced units, and also currently there are no optimized drivers available. Once every supporting module has been produced the performance of the flagship GPU would end up being on par with current modern-day graphics cards.
If the listing is to be believed then we can see the 512 EU GPUs in both 12 GB (192-bit) and 16 GB (256-bit) flavors. As for the variants, we only know so far that a full die with 4096 cores, 256-bit bus interface, and up to 16 GB GDDR6 memory has been listed to be created by Intel. For now, it seems that the blue team will produce more variants once it gets adequate demands from its buyers.