It’s not a rarity that tech giants will release or leak information regarding their products online, by accident. This time around, it’s Intel who’s in the spotlight, along with new information about four new variants of its unreleased Xe GPU.
Details of at least four Xe GPUs were first discovered on the Anandtech forum, and then republished by user “mikk”, but at the time writing, it seems that the post has already been taken down. Specifically, three of the Xe graphics cards mentioned go product code seen below:
The first letter, “i” is how all Intel GPUs will be referenced. The following letters, “DG” and “HP”, is speculated to be an acronym for “discrete GPU” and “high powered”. The numbers at the end could possibly be the execution units; Intel’s equivalent to NVIDIA’s CUDA cores.
By comparison, the numbers sound low compared to NVIDIA’s own offering, it should be made clear that Intel’s GPU architecture is not the same as NVIDIA’s. As per an example given by Techspot, Intel’s Gen11 GPU provides 16 flops per execution unit per clock. In other words, 512 execution units operation at 1800MHz means that the Xe GPU would be churning 14.7 TFLOPS, which is statistically higher than what a GeForce RTX 2080 Ti produces.
There’s also mention of two other discrete GPUs, an “iDG1LPDEV” and “iATSHPDEV”. These two don’t have as much information or details surrounding them, but the former sounds like a low-powered option that Intel is still trying to get its head around. However, the DEV at the end of both product codes makes it pretty obvious that the two are development units.
It’s important to note that the numbers and decoding made here are purely speculations, and there is a possibility that the information could be wrong. Until Intel officially debuts its Xe architecture, we do advise you take this with a grain of salt.