Intel Details Gen11 GPU Architecture via Public Whitepaper
Ron Perillo / 2 years ago
GPU Improvements Under the Hood
While Intel showed off renders off their upcoming video card, details are scant as to what will the hardware configuration will look like. In fact, it might take some time before those details even surface.
For now, Intel is releasing details regarding their Gen11 GPU instead. Which would be putting pressure on AMD and NVIDIA‘s low-end/low-power GPUs. You can read this detailed whitepaper (PDF file) for yourself by following this link.
This upcoming graphics architecture was initially announced during Intel Architecture Day last year, promising significant performance improvements over existing Gen9 IGP. The paper also notes that these will be on 10nm CPUs with 3rd Gen FinFET technology.
What are the Differences Between Intel Gen9 and Gen11 Graphics?
Intel still uses their ring architecture to connect the Gen11 graphics. With the CPU core, System Agent (PCIe, memory and display controllers) and LLC cache connected via a ring bus.
The Gen11 GPU itself uses a slice architecture, with 3D Fixed Function Geometry, eight Sub-slices with Execution Units, and a Slice common for fixed function blocks. Each sub-slice has a local thread dispatcher unit, shared local memory, 3d texture sampler unit, media sampler unit and dataport unit.
In turn, each EU has a pair of ALUs with four execution pipelines each. These are capable of executing four 32-bit FP or eight 16-bit FP.
Intel also outlines the attribute differences in the table below. Showing a notably sizable level 3 cache and other attribute improvements overall. Interestingly, Gen11 also jumps into LPDDR4 support, which was lacking in Gen9.
Other improvements include a new HEVC Quick Sync Video engine, efficient 4K/8K video stream support, VP9 decode bit depth improvement, Adaptive Sync and more.