Highlights:

  • ZeroPoint refers to the Ziptilion-BW’s performance as enabled by a collection of entropy-based compression methods.
  • The company plans to introduce more hardware acceleration products using the capital raised in its recently disclosed fundraising round.

ZeroPoint raised USD 5.5 million in the initial-stage funding. The startup develops hardware that facilitates processors to use DRAM efficiently.

The lead investor in the Series A financing, according to the announcement of ZeroPoint Technologies AB, was Munich-based Matterwave Ventures. Besides, Chalmers Ventures, Climentum Capital, and Industrifonden were involved.

ZeroPoint is a Gothenburg, Sweden-based spinoff of Chalmers University of Technology. The company was started with the intention of commercializing data compression research that its co-founders, Per Stenström and Angelos Arelakis, conducted at the university. According to ZeroPoint, this technology can save operating costs and increase the power efficiency of data center servers.

The performance of a server CPU is strongly impacted by how quickly it can get data from DRAM. As a result, one of the top priorities for semiconductor designers is to accelerate data retrieval times. Compressing the files that a chip processes can help speed up the retrieval workflow because a smaller file can be fetched from memory more quickly.

This kind of compression-based chip speeding was historically unfeasible. The most sophisticated compression algorithms are not very fast, so any processing gains they could offer through DRAM optimization are offset by their relative slowness. While quicker algorithms exist, their ability to reduce data size is much less.

ZeroPoint says it has found a solution to this trade-off. The Ziptilion-BW is a small semiconductor module that the company sells and claims to be quick and efficient at reducing data. The chip claims to operate 1,000 times quicker than some conventional compression techniques while reducing dataset memory footprints by up to 400%.

ZeroPoint refers to the Ziptilion-BW’s performance as enabled by a collection of entropy-based compression methods. The company claims that those algorithms evaluate a chip’s memory and identify the different kinds of data stored there by applying statistical techniques. Next, based on this information, the Ziptilion-BW determines the most effective method for data compression.

The device integrates several more performance optimization techniques with the entropy-based algorithms developed by ZeroPoint. Notably, an onboard cache is employed to hold previously used data. Calculations are sped up by servers’ ability to obtain data faster from this cache than from the DRAM, where it is typically kept.

ZeroPoint claims that the Ziptilion-BW can be used as a semiconductor module that is only 0.4 square millimeters in size. Because of this, chipmakers may incorporate the technology into their CPUs quite easily. According to ZeroPoint, the Ziptilion-BW can lower the total cost of ownership for data center servers by 25% while increasing performance per watt by 50%.

ZeroPoint Chief Executive Officer Klas Moreau said, “ZeroPoint Technologies’ customers include some of the biggest semiconductor companies in the world and our products are in demand by data center operators looking to overcome the mounting challenge of memory bottlenecks.”

The company plans to introduce more hardware acceleration products using the capital raised in its recently disclosed fundraising round. To support the growth goal, ZeroPoint also intends to increase staffing levels in Sweden and the United States and to scale its sales operations.