Every performance claim traces back to RTL.
Not a spreadsheet model.

IPC within 5% of RTL.

Correlated cycle-by-cycle, pipeline-stage by pipeline-stage using Dhrystone (CPU) and Matrix Multiplication (VPU). Branch prediction and cache hit rates match RTL at 100%.

5%

IPC correlation tolerance vs. RTL

100%

Branch prediction match vs. RTL

100%

Cache hit rate match vs. RTL

Configurable from the model.

Configurable parameters

I-cache and D-cache: 16KB – 128KB
L2 cache: 512KB – 4MB
Memory latency: 10 – 128 cycles
Vector register width (VLEN): 128 – 2048 bits

What's in the package

GCC RISC-V compiler · Whisper ISS · Cycle-accurate model
Benchmark suite: Dhrystone · MatMul · Qwen · YOLO
Instruction manual — run it yourself

Cycle-accurate. Stage by stage.

IPC within 5% of RTL — correlated cycle-by-cycle, pipeline-stage by pipeline-stage using Dhrystone (CPU) and Matrix Multiplication (VPU).

Qwen 2.5 7B — verified on the model.

Performance model result
28.8 tok/s

Qwen 2.5 7B · FP16 · batch 1 · 100-cycle DRAM penalty

Simultaneously achieves

17 SpecInt2k6/GHz (8-wide) or 14 SpecInt2k6/GHz (4-wide) running concurrently with Qwen — with reduced resource conflict due to TBS scheduling.

CircuitSutra partnership.

Performance modeling is conducted in partnership with CircuitSutra for independent TBS performance correlation. UCI compiler development is ongoing alongside RTL verification.

Granted and filed patents.

Microprocessor Patents

Patent #TitleGrant Date
US-11829187-B2Microprocessor with Time Counter for Statically Dispatching Instructions11/28/2023
US-11829767-B2Register Scoreboard for a Microprocessor with a Time Counter11/28/2023
US-11829762-B2Time-Resource Matrix for a Microprocessor with Time Counter11/28/2023
US-12190116-B2A Microprocessor with Apparatus and Method for Replaying Instructions1/7/2025
US-12169716-B2Microprocessor with a Time Counter for Statically Dispatching Extended Instructions12/17/2024

Vector Processor Patents

Patent #TitleGrant Date
US-12106114-B2Microprocessor with Shared Read and Write Buses and Instruction Issuance to Multiple Register Sets10/1/2024
US-12112172-B2Vector Microprocessor with a Time Counter for Statically Dispatching Instructions10/8/2024
US-12124849-B2Vector Microprocessor with Extended Vector Register Sets10/22/2024
US-12288065-B2Microprocessor with Odd and Even Registers4/29/2025
US-12282772-B2Vector Microprocessor with Vector Data Buffer4/22/2025

Additional applications filed and pending. Full list available on request.

From humanoid robotics to sovereign AI infrastructure.

See Applications