Chimera™ GPNPU delivers full Vision Transformer support—including attention mechanisms, LayerNorm, and GELU—without the operator limitations of legacy NPUs.
ViT-Base on Chimera GPNPU
ViT-Base on Chimera GPNPU
Vision Transformers have revolutionized computer vision, outperforming CNNs on classification, detection, and segmentation, but there's a big problem: their architecture is fundamentally different from what legacy NPUs were designed to accelerate. A programmable, general-purpose NPU is the answer: Chimera.
Multi‑head self‑attention requires matrix multiplications that don't map efficiently to legacy NPU systolic arrays.
LayerNorm, GELU, and Softmax often fall back to the CPU. The model technically runs but it's too slow and power hungry to be useful.
Variable sequence lengths in transformers break fixed-tensor NPU compilers designed for static workloads.
Legacy NPU memory hierarchies were built for sequential access and are overwhelmed by Attention's O(n²) access pattern.
Chimera's general-purpose NPU architecture was designed from the ground up to handle any ML workload, including the unique demands of current and future transformer models.
Every ViT operator runs natively on Chimera with no CPU fallback or operator gaps. Multi‑head attention, LayerNorm, GELU, Softmax, and patch embedding all execute on the GPNPU with full hardware acceleration.
Our Chimera Graph Compiler (CGC) automatically recognizes and optimizes transformer attention patterns, mapping them efficiently to Chimera's compute array and memory hierarchy.
Chimera's software-managed memory handles the variable access patterns of attention without the rigidity of hardware-managed caches, just configure the L2 Memory size to match your model's requirements.
GPNPURun industry-standard Vision Transformers on Chimera today, with more models coming soon. All available models are ready for immediate evaluation in DevStudio.
Hierarchical vision transformer with shifted windows
Bird's-eye-view transformer for 3D perception
Bring your own Vision Transformer architecture
New models regularly added: Chimera Graph Compiler (CGC) supports standard Transformer architectures, enabling porting from ONNX without RTL changes.
Chimera hardware supports and accelerates every capability Vision Transformers require.
Native hardware support for scaled dot-product attention with configurable head counts and embedding dimensions.
Hardware-accelerated Layer Normalization without CPU fallback. Handles pre-norm and post-norm architectures.
Full support for GELU, Softmax, and all standard activation functions used in transformer architectures.
Efficient image-to-patch conversion with configurable patch sizes and embedding dimensions.
INT8 symmetric and asymmetric quantization with Quantization-Aware Training (QAT) flow for accuracy preservation.
Scale ViT inference across multiple Chimera cores for higher throughput with data parallelism.
View cycle-accurate performance metrics, explore the compiled graph, and evaluate Chimera for your application.
Our engineering team can help you evaluate custom architectures, optimize performance, and plan your deployment on Chimera GPNPU.
Contact Our Team