Back to Media Center

RECOGNI DELIVERS 1000 TOPS (MICROPROCESSOR REPORT)

RECOGNI DELIVERS 1000 TOPS (MICROPROCESSOR REPORT)

Driving autonomy inherently requires visual inputs which in turns necessitates a massive amount of computation.  Seeing both far and near and in every direction is imperative at high speeds and with minimal processing delay.  While high compute requirements typically goes hand-in-hand with high power consumption, Recogni has effectively resolved these diametrically opposing requirements.

Recogni’s Vision Cognition Module (VCM®) utilizing Recogni’s inference accelerator with 1000 TOPS of compute capacity and novel compute approach is capable of processing multiple 4K high dynamic range streams & classifying objects and their depth.  High resolution is a key enabler in resolving objects at far distance and minimizes the need for use of disparate sensory components.  Considering the low power consumption of 25W (TDP), Recogni’s solution has unmatched performance efficiency amongst automotive inference processing providers.

With Recogni’s unparallel performance (1000 TOPS) at only 25W of power, VCM® delivers industry-leading perception processing and accuracy in real-time without compromising the driving range of electric vehicles.  Low power consumption is critical to both lowering environmental impact and accommodating the user to preserve power for longer range driving rather than cognition processing.

Recogni’s low power consumption is realized by several contributing factors:

  • Use of novel logarithmic computation that significantly reduces the compute power consumption
  • Highly optimized acceleration engine particularly for 3×3 convolutions
  • Novel compression scheme to reduce memory data transfers
  • Minimized DRAM accesses

As a standard benchmark comparison, a 224 x 224 ResNet 50 runs at 50,000 fps and MLPerf  SSD-Large object detection using 1200×1200 pixels is realized at 300 fps.  Such performance is not achieved by any other similar class product in the market.

Recogni’s architecture is achieving 1000 TOPS with high efficiency whereas other general-purpose offerings in the market are at sub 50% efficiency. This translates into power wasted and is clearly seen with other chipsets and systems’ TOPS/Watt ratio.

The industry is taking notice and is excited about what Recogni brings to realm of vehicle autonomy. Recogni’s processing pipeline, acceleration and memory architecture, as well as competitive benchmarks are published in Microprocessor Report, June 6, 2022, by Byron Moyer.  For further information on the report, please visit:

Recogni Delivers One Petaop/s.


You May Also Like

3rd Party Media
3rd Party Media
Articles
Converting a network for the fastest inference chip on the market

Recogni Core-Tech Team *Conversion: Making a full-precision network compatible with our chip Introduction Deep Neural Networks (DNNs), in particular convolutional ones, are the go-to approach for visual perception tasks that…

Subscribe to Our Updates

Join our newsletter to stay up to date with Recogni!

Enabling ADAS to Full Autonomy