MIPS64 + SIMD Multi-CPU
Multi-core, multi-threaded, multi-cluster MIPS™ CPUs
TritonAI™ 64 Customizable IP Platform
for AI
Driving Inferencing to the Edge
TritonAI™ 64 Platform for AI-enabled Edge SoCs
Wave Computing’s customizable, AI-enabled platform merges a triad of powerful technologies to efficiently address use case requirements for inferencing at the edge.
A Scalable Triad of Technologies
-
-
WaveFlow Technology
Patented, scalable dataflow platform designed to efficiently execute existing and future algorithms
-
WaveTensor Technology
Highly-efficient, configurable, multi-dimensional TensorCore processing engines
Combination of Technologies Helps Future Proof Your Investments in a Rapidly Evolving Industry
MIPS64r6 ISA:
- 128-bit SIMD/FPU
- 8/16/32/int, 32/64 FP datatype support
- Virtualization extensions
- Superscalar 9-stage pipeline w/SMT
- Caches (32KB-64KB), DSPRAM (0-64KB)
- Advanced branch predict and MMU
Multi-Processor Cluster:
- 1-6 cores
- Integrated L2 cache (0-8MB, opt ECC)
- Power management (F/V gating, per CPU)
- Interrupt control with virtualization
- 256b native AXI4 or ACE interface
WaveFlow Reconfigurable Architecture
- 2-1K scalable tiles
- Flexible 2-D tiling layout
- Each tile contains 16 CPUs and 8 MACs (8int)
- 8/16/32 int datatype support
- Compatible datatypes with WaveTensor
- Dynamically reconfigurable
- Overlapping compute and I/O
- Concurrent network execution
- Future proof for all AI algorithms
- Add signal and vision algorithms
- Independent operation from CPU
WaveTensor Reconfigurable Architecture
- Highly efficient for CNN execution
- >10 TOPS/sq mm, 8 TOPS/watt @ 7nm
- Configurable 4×4 or 8×8 MAC tiles
- 8 int datatype support
- Compatible datatypes with WaveFlow
- Scalable # of tiles per slice
- Scalable # of slices per core
- Overlapping compute and I/O
- Independent operation from CPU
TritonAI Application Programming Kit (APK)
- GCC Binutil tools and IDE
- Debian Linux OS support and updates
- WaveRT API abstracts framework calls
- WaveRT Optimized AI libraries for
- CPU/SIMD/WaveFlow/WaveTensor
- TensorFlow-lite build support and updates
- TensorFlow build for edge training roadmap
The combination of Wave’s AI technologies and MIPS embedded, RISC, multi-threaded CPU IP enables inferencing at the edge today and training at the edge tomorrow.
AI-Native Platform Offers Significant Benefits
The TritonAI 64 platform provides the inferencing power needed for today’s AI edge applications, with the flexibility and scalability to address the rapidly evolving needs of the innovative AI industry.
Wave Innovation:
A Scalable, Unified Platform to Accelerate AI
By combining Wave’s unique, patented WaveFlow processor technology with its proven, efficient, MIPS core designs, Wave is powering the next generation of AI, with a single platform that scales from the datacenter to the edge.
