← Back to Overview
Ancillary Tile · Intel 14A-E · 2DKR · Architecture Deep Dive

2D KANVAS
RENDERER

Purpose-built 2D silicon. Zero 3D ALUs. Zero shaders. Zero Z-buffers. Every transistor dedicated to one mission: rendering the desktop at sub-millisecond latency while consuming less than a watt.

Intel 14A-E Modernised Blitter Engine Zero 3D ALUs Sub-ms UI Latency <1W Active
<1W
Active Power Draw
Entire tile under load
0
3D ALUs
Pure 2D silicon
<0.3ms
Frame Latency
Desktop composition
14A-E
Process Node
Intel efficiency node

WHY DEDICATED
2D SILICON

Every modern GPU — from integrated to discrete — wastes significant silicon area and power budget on desktop composition. Window borders, taskbar animations, scrolling text, and notification overlays all compete for the same 3D pipeline used by games and compute workloads. This is architectural waste.

The 2D Kanvas Renderer eliminates it entirely. By moving all 2D composition to a dedicated tile built on Intel 14A-E, the Eventide APU achieves something no competitor can: the desktop is completely invisible to the 3D GPU.

During gaming, the Arc Elementalist Xe5 dedicates 100% of its shader array, memory bandwidth, and thermal budget to the 3D workload — while the 2DKR silently handles the Steam overlay, the Discord notification, the taskbar clock, and the FPS counter. No context switches. No shader preemption. No stolen cycles.

Process Intel 14A-E
Architecture Scanline-based 2D pipeline
3D ALUs None (0)
Shaders None (0)
Z-buffer None
Blitter engine Modernised, hardware-accelerated
Alpha blend units 8 parallel
Overlay layers 16 hardware layers
Max resolution 8K60 per output
Active power <1W (full desktop load)
Idle power <0.1W (static desktop)
Die area ~4 mm² (Intel 14A-E)

THE SCANLINE
MACHINE

The 2DKR operates on a fundamentally different rendering model than any GPU. Where GPUs tile-rasterize triangles in parallel, the 2DKR processes rectangles scanline-by-scanline — the natural unit of 2D window composition. This isn't a compromise; it's an optimization.

Desktop windows are axis-aligned rectangles. Scrolling is a vertical shift. Text rendering is a horizontal blit. The scanline model processes these operations in a single clock cycle per line, with zero triangle setup overhead. The Modernised Blitter Engine at the core of the 2DKR is a direct descendant of the blitter hardware that powered every PC from the IBM PC/AT onward — re-engineered for 8K HDR composition with hardware alpha blending, gamma-correct sub-pixel anti-aliasing, and a 16-layer overlay compositor.

STAGE 1 CMD FETCH Command ring buffer read STAGE 2 RECT DISPATCH Rectangle decomposition STAGE 3 SCANLINE RAST Line-by-line rasterization STAGE 4 ALPHA BLEND 8-unit parallel alpha composite STAGE 5 OVERLAY COMP 16-layer overlay stack STAGE 6 DISPLAY OUT Direct scanout to panel

The overlay compositor supports 16 hardware layers — each independently scrollable, alpha-blendable, and addressable. The taskbar, notification center, Start menu, and in-game overlays each occupy their own layer. Layer switching is instantaneous (no framebuffer rebuild), and the display controller reads directly from the overlay compositor's output buffer — eliminating the copy-to-framebuffer step that plagues GPU-based composition.

INVISIBLE
POWER

The 2DKR consumes less than one watt under full desktop composition load — rendering windows, compositing alpha-blended overlays, running smooth 120Hz scrolling, and driving animated taskbar elements. Under static desktop (no movement), power drops below 100 milliwatts.

Compare this to GPU-based desktop composition on competing platforms: a discrete GPU in "desktop idle" still draws 5–12 watts maintaining its 3D pipeline, shader array, and memory controller in a ready state. The 2DKR has no 3D pipeline to power, no shader array to idle, no high-bandwidth memory to keep warm. It draws only what 2D composition demands — and 2D composition demands almost nothing.

The result: during gaming, the desktop overlay adds zero watts to the GPU power budget. The FPS counter, the Discord overlay, the stream chat widget — all rendered by the 2DKR, all invisible to the Arc Elementalist's thermal envelope.

vs GPU-BASED DESKTOP COMPOSITION
Desktop idle W
−85%
UI frame latency
−92%
GPU freed for 3D
100%
Overlay GPU cost
0W
Full composition load <1W
Static desktop <0.1W
Overlay during gaming 0W GPU overhead
120Hz scroll smoothing <0.5W
Peak (all 16 layers, 8K, HDR) ~0.8W

TWO PIPELINES,
ZERO CONTENTION

The Eventide driver stack implements hardware-level API routing that separates 2D and 3D rendering at the command level. When a game runs, the driver identifies 2D overlay surfaces — HUD elements, Steam overlay, Discord chat, system notifications, video capture widgets — and routes them to the 2DKR's command queue. 3D draw calls, compute dispatches, and ray tracing commands route to the Arc GPU pipeline.

This split happens at the API layer, before command buffer submission, with zero latency overhead. The game sees a single GPU; Windows sees a single display adapter. But internally, two independent silicon pipelines execute simultaneously, each optimized for its workload class. No mode switches. No shader preemption. No pipeline stalls. No resource contention.

DirectComposition / DWM → 2DKR (all desktop compositing)
DXGI overlay surfaces → 2DKR (game overlays, stream widgets)
Windows.UI.Composition → 2DKR (UWP/WinUI animations)
Direct3D 11/12 → Arc Xe5 / Xe4E (3D rendering)
Vulkan → Arc Xe5 / Xe4E (3D rendering)
DirectCompute / OpenCL → Arc Xe5 / Xe4E (GPU compute)
Video decode (H.266/AV1) → Media Engine (shared)
Display scanout → 2DKR (final output compositing)