Products

“GigaMACS” stands for “Giga Multiply and Accumulate,” plural; it is a commercial-ready AI Accelerator.

GigaMACS takes your TensorFlow or other CNN (convolutional Neural Network) model, as-is, and uses our patent-pending technology to compile a hyper-optimized bitstream to use in your FPGA or for your custom ASIC.

GigaMACS will automatically accelerate your model to have near-zero latency, require no buffering, and enable your model to handle full camera HD, 4K or 8K at input speed, even in real-time. GigaMACS works with all Convolutional Neural Network models and delivers a FPGA or ASIC ready-to-use solution.

If you have a camera at 80 fps or even faster, GigaMACS will make sure your model works at the same rate. GigaMACS can easily move models to 240 fps in full 4K and in real-time. The only speed limit for GigaMACS is your camera.

Other technology benchmarks a model with the 224x224x3 thumbnail images. While attempting to keep up with processing, the other technology solutions cannot process full-frame high-resolution images without dropping a large percentage of the frames. While nVidia will perform up to 75 fps with these small images, GigaMACS will exceed 3,000 fps.

A test with the Tesla nVidia V100 on AWS with SqueezeNet can reach 25 fps with a full 1920x1080x3 frame. GigaMACS automatically optimized SqueezeNet on an FPGA hardware test and reached the full 80 fps (the input rate). The nVidia hardware clock runs at 10x the FPGA’s speed, and GigaMACS still outperforms it by a magnitude. Also, nVidia’s latency was nearly 60 milliseconds, compared to GigaMACS’ which was less than a millisecond. GigaMACS can perform as fast as the input.

Why not go to full-frame HD, 4K or 8K and be real-time with GigaMACS?

Adding more memory and faster clocks to GPUs or TPUs is not the answer to accelerating your Neural Network model; it ultimately leads to a dead-end with minuscule gains at best.

Filtering low-weights from the calculations potentially decreases the accuracy of the model, which requires re-training and re-certification. Changing the accuracy of the model impacts the Neural Network model as it evolves and becomes more complex.

GigaMACS will remove redundancies and looping while automatically identifying and implementing optimizations; also, it will make your model perform in real-time with near-zero latency. In the process of using an FPGA, or making a custom ASIC for the ultimate performance, your model will reduce power consumption by 90%.

GigaMACS does not change your AI/ML model in the process but it does make it much faster!