Arm Immortalis / Cortex

Arm DLSS for Android Frame Generation & Resolution Scaler Simulator

REAL-TIME NEURAL SUPER RESOLUTION MATRIX
Native Input (Low-Res Buffer)
Arm NPU Reconstructed Edges
Interpolated Optical Vectors
Temporal Artifact & Edge Fidelity Score
High Fidelity (Clean Edge Reconstruction)
Sub-pixel jitter suppression active via neural temporal feedback
Target Hardware Tier (SoC Profile) Arm Immortalis-G720 + Cortex-X4 NPU
DLSS Upscale Mode Quality (1.5x)
Neural Frame Generation
Interpolates alternating frames via Arm Optical Flow Engine
Thermal Ceiling / Power Cap 6.5 W
Effective Frame Rate
84.5 FPS
+112% vs Native 1080p
End-to-End Latency
28.2 ms
Reflex buffer engaged
Total SoC Power Draw
5.4 W
Within thermal budget
Thermal Throttling Status
Optimal (No Throttling)
Sustained clock stability
Arm Super Resolution & Neural Frame Interpolation Specification Matrix

Based on The Verge's reporting on Arm's upcoming deep learning super sampling solution for Android, mobile games can bypass GPU fill-rate limits by rendering at lower internal viewports (540pā€“720p) and leveraging NPU tensor pipelines to reconstruct fine sub-pixel geometry before display scan-out. Optical flow accelerators estimate motion between successive frames to inject synthesized intermediate frames with minimal thermal overhead.

Architecture Stage Internal Native Buffer Upscaled Presentation NPU Compute Cost Memory Bandwidth Saved
G-Buffer Rasterization 720p (1280x720) 1080p to 4K AI Upscale 1.8 ms / frame ~42% vs Native 4K
Neural Optical Flow Frame Gen Motion Vector Grid Synthesized 120Hz Phase 2.4 ms / frame +8% Buffer Read Stride