Soft Machines, Inc.

Operating R&D Center

Shareholders in Portfolio Company 
Samsung Ventures, AMD, Mubadala, Russian Venture Company, KACST, RUSNANO, TAQNIA

Industry Sector 
Optics and Electronics

Production Location 
Santa Clara (USA), Moscow (Russia), Hyderabad (India)
Jobs created: 250

Investment Started:  2013

Development of a new architecture for processor cores. The new architecture, christened  “Variable Instruction Set Computing” (VISC), opens the way for processors with much higher energy efficiency and performance compared to processors with existing core architectures. The project is also developing a system-on-chip (SoC) structure with VISC microprocessor cores.

Soft Machines is a startup company in the semiconductor field. The company is dedicated to the licensing and joint development of products using VISC architecture for the Internet of Things (IoT), mobile telecommunications and cloud computing. The company is headquartered in the city of Santa Clara (California, USA).

Soft Machines is developing VISC architecture for energy-efficient, high-performance processors and building system-on-chip (SoC) prototypes using this architecture. Some of the company’s technologies have already been made public and patented (40 patents have been obtained), but most of Soft Machines’ technologies are still being prepared for patenting. The company’s new microprocessor outperforms analogues from ARM and Intel by several times in execution of average tasks. The processor architecture is compatible with all major architectures on the market from Intel and ARM, making it possible to address any tasks in applications written for ARM and x86 architecture. VISC compatibility with existing architectures opens up tremendous potential for entry to one of the fastest growing segments of the server market and of cloud infrastructure.

Soft Machines has offices in Russia (the Russian subsidiary is SMWare) and India (the Indian subsidiary is SMSilicon). SMWare LLC employs over 50 highly skilled engineers, who are tackling problems in the field of computer architectures, operating systems, compilers, virtual machines and systems programming. The “software virtualization layer” is a key component of VISC architecture, since it is what enables the use of VISC technology in any (existing and new) software ecosystems without any changes. The software virtualization layer has been developed, from start to finish, by the SMWare team. Soft Machines has been active in Russia since 2009 under the leadership of Alexander Drozdov.

Areas of application

  • Mobile solutions, data processing


  • Manufacturers of mobile devices, “smart home” solutions and servers

Competitive advantages

  • Binary compatibility with operating systems and software written for both x86 and ARM architecture.
  • It is expected that VISC architecture will enable performance of 3-4 times more instructions per clock cycle compared to analogues, which will increase productivity by 2-4 times per watt in single-threaded and multi-threaded applications.
  • VISC architecture SMI can run applications written for ARM and x86 architectures, making the Soft Machines product competitive in all server and mobile device market segments (phones, tablets, laptops/ultrabooks).

VISC architecture is based on the concept of “virtual cores and virtual hardware threads”. This new approach enables dynamic allocation of resources between cores. Microprocessors based on CISC and RISC architectures use physical cores and program streams, and this approach has come up against a technology and economic wall due to limits on transistor use and power consumption, and also problems associated with frequency increase. Use by the VISC architecture of a “lightweight software virtualization layer” is also important, since it means that the architecture can be used in both existing and new software ecosystems.

In 2013 Soft Machines released a silicon prototype of the system on a chip with microprocessor cores based on the VISC architecture. On 23 October 2014, Soft Machines presented results from use of the prototype at the Linley Group Processor Conference 2014 and demonstrated to semiconductor industry experts how the prototype works.

Results of the prototype testing on Dhrystone, SPEC2006, etc. benchmarks and comparison of its performance with that of microprocessors based on ARM Cortex A15, A57, Apple A7, Haswell 1-C and several other cores can be found in the Soft Machines presentation at the Linley Group Processor Conference 2014, which is available on the company’s website.