High performance and low latency hardware accelerated zram/zswap at unmatched power efficiency

Overview

The SuperRAM implements a hardware accelerator for zram compression and decompression. SuperRAM implements a ZeroPoint proprietary compression algorithm and is optimized for power efficiency, high throughput and high compression efficiency.

SuperRAM is integrated on the SoC as other hardware accelerators, as a master node on the SoC interconnect. Part of the integration includes a software driver so that the zram crypto-compress API sends a command to the SuperRAM accelerator when there is a software-triggered compression and decompression.

Key Features

  • Compression ratio: 2-4x across diverse data sets
  • Compression throughput: 8GB/s
  • Decompression throughput: 10.5GB/s
  • Frequency: DDR4/DDR5 DRAM speed
  • IP area: Starting at 0.1mm2 (@5nm TSMC)
  • Memory technologies supported: (LP)DDR4, (LP)DDR5, HBM
  • SuperRAM IP is compatible with all DRAM technologies and supports standard interfaces such as AXI and CHI. Other proprietary interfaces can be supported upon request.

Benefits

  • High performance and low latency hardware accelerated zram/zswap at unmatched power efficiency
  • Off-loading CPU – More cycles to released to user work loads
  • Power efficiency – Less energy
  • Speed – Fast compression and low latency access
  • Several in-flight compression and decompression operation operating in parallel
  • Operating at main memory speed and throughput
  • Compatible to AXI4/CHI, both 128-b and 256-b bus interface
  • Intelligent real-time analysis and tuning of the IP Block

Applications

  • Smart device SoCs and Server CPUs.
  • Smart device SoCs: The product benefit is a better user experience at unmatched power efficiency when the host processor is offloaded and the page swapping is hardware accelerated.
  • Server CPUs: The product benefit is more system performance at less power when SW-based compression is off-loaded from the the host. Operating system and hypervisor offload SW-based compression of swapped pages with a super-fast page compression technology and return more performance to the guest at unmatched power efficiency.

Deliverables

  • Synthesizable Verilog RTL (encrypted)
  • Implementation constraints
  • UVM testbench (self-checking)
  • Vectors for testbench and expected results
  • User Documentation

Technical Specifications

Foundry, Node
TSMC 7nm, 5nm, 3nm
Maturity
Tape-out
Availability
Immediate
×
Semiconductor IP