A unique way to remove unnecessary data

A typical microchip memory contains up to 70% unnecessary information. ZeroPoint technology removes all that information by combining ultra-fast data compression with real-time data compaction and transparent memory management. Integrating these three areas into one solution for maximizing performance per watt makes ZeroPoint unique.

How ZeroPoint technology removes unnecessary data in a three-step process

Since Intel launched its first microprocessor in 1971, the computational performance has doubled almost biannually. As compute engines are separated from memory, performance scaling has been challenging for decades and is now becoming a showstopper. ZeroPoint is committed to getting rid of this roadblock once and for all.

Compression efficiency and latency in the memory hierarchy

Is it possible to deliver both high compression ratio and low latency? In fact it is. For those interested in ultra-low latency in the nanosecond domain paired with high compression ratios, ZeroPoint technology is the only solution available on the market today.

To get the full picture of ZeroPoint technology, we suggest you read our whitepaper.

To get the full picture of ZeroPoint technology, we suggest you read our whitepaper.


Our technology explained in a few words

video thumbnail
play video icon
video thumbnail
play video icon

Long version

Most frequent questions

and answers

What do you deliver?

We deliver a hardware IP block (data compression and compaction) that customers integrate on a CPU or SoC and customized software (memory management).

How can I evaluate your technology?

We believe a good approach is to follow a three-step evaluation funnel: (1) Compression ratio analysis based on your data, (2) Memory management emulation based on your system architecture and (3) Architectural simulation based on your application. Evaluating the compression and memory expansion potential only takes a few weeks.

What is the maximum throughput?

The maximum throughput of our technology depends on a combination of design, process node, cell libraries and bus architecture. Contact us to take part of the test results.