1 / 29

Lecture 9: FPGAs vs. ASICs

Lecture 9: FPGAs vs. ASICs. Spectrum of Design Choices. Fast, Inflexible. Choices. Full Custom. Polygons. ASIC. Standard Cells (LTE Modem). FPGA. Logic Network (Intel/Altera, Xilinx). Specialized Processor. Program (e.g., GPUs). GP Processor. Program (e.g., Intel x86, ARM).

shana
Télécharger la présentation

Lecture 9: FPGAs vs. ASICs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Lecture 9: FPGAs vs. ASICs

  2. Spectrum of Design Choices Fast, Inflexible Choices Full Custom Polygons ASIC Standard Cells (LTE Modem) FPGA Logic Network (Intel/Altera, Xilinx) Specialized Processor Program (e.g., GPUs) GP Processor Program (e.g., Intel x86, ARM) Slow, Flexible

  3. “Wires” (Routing) implemented by programming connectivity boxes (CBs) and switch boxes (SB) “Logic gates” implemented by programming configurable logic blocks (CLBs) Modern FPGAs have more than 1 million equivalent logic gates General FPGA Layout

  4. Basic CLB comprises “Lookup Table” (LUT) and D-Flip Flop The MUX allows selection of either the LUT output or the D-FF output Configurable Logic Blocks

  5. Combinational functions created with programmed “tables” connected to cascaded multiplexers LUT inputs are MUX select lines Lookup Tables (LUTs) Programmed Levels (EEPROM or SRAM) 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 1

  6. The basic “Adaptive Logic Module (ALM) Block Diagram” Note the fast adder carry chain (does not require going out to programmable switch boxes) Intel Stratix II FPGA Architecture

  7. Each ALM can be configured to one or two logic functions ALM Flexibility

  8. ALM Flexibility

  9. ALM Flexibility

  10. Connectivity Between ALMs

  11. Intel Stratix 10 GX 5500/SX 5500 FPGAs implemented in 14 nm process Contains 1,867,680 ALMs, which can implement roughly 5,510,000 logic elements (logic gates). Contains 7,470,720 ALM registers Also contains Quad ARM Cortex-A53 CPU cores Latest Stratix 10

  12. Integrated Quad ARM Cores

  13. Modern FPGAs have many built-in interfaces. DRAM PCI Express USB SATA (disk drives) etc Makes them easy to integrate into compute environments Many Built-In Interfaces

  14. FPGA vs. ASIC

  15. ASIC (Application-Specific Integrated Circuit) designs are usually implemented using “standard cells” Standard cells are pre-designed layouts of transistors for implementation of common logic gates and registers (D-Flip Flops) Standard cells are be pre-characterized in terms of cell area, cell delay, and cell power consumption Simplifies design flow, design verification, and timing analysis ASICs

  16. N-Channel MOS (NMOS) transistors turn “ON” when the Gate voltage = “1” P-Channel MOS (PMOS) transistors turn “ON” when the Gate voltage = “0” CMOS Gate Gate “1” “0” Source Source Drain Drain

  17. CMOS Inverter Layout Figure derived from slides by S. Edwards from his CSEE4840 class

  18. CMOS NAND Gate Figure derived from slides by S. Edwards from his CSEE4840 class

  19. CMOS NAND Gate Figure derived from slides by S. Edwards from his CSEE4840 class

  20. CMOS NAND Gate Figure derived from slides by S. Edwards from his CSEE4840 class

  21. CMOS NAND Gate Figure derived from slides by S. Edwards from his CSEE4840 class

  22. Standard Cells of Logic Gates

  23. CMOS D-Flip Flop

  24. Standard Cell of D-Flip Flop Edge-Triggered D-Flip Flop with Asynchronous Reset

  25. Standard Cell Layout

  26. Standard Cell Layout

  27. Standard Cell Layout

  28. NVIDIA Tesla GV100 GPU in 12 nm process Contains 23 billion transistors Latest NVIDIA GPU

  29. FPGA and ASIC have similar design flows Design Flows FPGA Flow ASIC Flow SystemVerilog Translation SystemVerilog Translation Logic Optimization Logic Optimization Logic Clusteringto LUTs Technology Mapping to Standard Cells Floorplanning Assignment & Programming CLBs Placement “Fitter” Routing & Programming CBs/SBs Routing

More Related