Method and structure for implementing a cache memory using a DRAM array

• US 5,829,026 A
• Filed: 03/05/1997
• Issued: 10/27/1998
• Est. Priority Date: 11/22/1994
• Status: Expired due to Term

• Current Assignee: Adeia Semiconductor Technologies LLC (f/k/a Invensas Corporation) (Adeia Inc.)
• Original Assignee: Monolithic System Technology
• Inventors: Leung, Wingyu, Hsu, Fu-Chieh
• Primary Examiner(s): Chan, Eddie P.
• Assistant Examiner(s): Ellis, Kevin L.
• Application Number:
  US08/812,000
• Time in Patent Office: 601 Days
• Field of Search: 395/445, 395/432, 711/105, 711/118, 711/119, 711/122
• US Class Current: 711/122
• CPC Class Codes: G06F 12/0893 Caches characterised by their organisation or structure
• CPC Class Codes: G06F 12/0897 with two or more cache hierarchy levels with multilevel cache hierarchies G06F12/0811

Abstract

A method and structure for implementing a DRAM memory array as a second level cache memory in a computer system. The computer system includes a central processing unit (CPU), a first level SRAM cache memory, a CPU bus coupled to the CPU, and a second level cache memory which includes a DRAM array coupled to the CPU bus. When accessing the DRAM array, row access and column decoding operations are performed in a self-timed asynchronous manner. Predetermined sequences of column select operations are then performed in a synchronous manner with respect to a clock signal. The DRAM array is operated at a higher frequency than the frequency of the CPU bus clock signal, thereby reducing the access latency of the DRAM array. By operating the DRAM array at a higher frequency than the CPU bus, additional time is provided for precharging the DRAM array. As a result, the precharging of the DRAM array is transparent to the CPU bus.

Claims

1. A computer system comprising:

a central processing unit (CPU);

a first level cache memory integrated with the CPU;

a bus coupled to the CPU wherein data is transferred on the bus in response to a first clock signal having a first clock frequency, wherein data is transferred on the bus at a first data rate; and

a next level cache memory coupled to the bus, wherein the next level cache memory comprises at least one dynamic random access memory (DRAM) array and a control circuit for controlling data transfer between the bus and the DRAM array, wherein the control circuit derives a second clock signal from the first clock signal, and causes data to be transferred to and from the DRAM array in response to a second clock signal at a second data rate which is faster than the first data rate, wherein the control circuit comprises a read buffer having a data input port coupled to the DRAM array and a data output port coupled to the bus, wherein the data input port is clocked by the second clock signal and the data output port is clocked by the first clock signal, and data is transferred from the DRAM array to the read buffer via the data input port faster than data is transferred from the read buffer to the bus via the data output port.

2. The computer system of claim 1, further comprising an SRAM cache tag memory coupled between the bus and the next level cache memory.

3. The computer system of claim 1, wherein the next level cache memory is integrated with the CPU.

4. The computer system of claim 1, further comprising a system logic controller coupled to the bus, wherein the system logic controller and the next level cache memory are fabricated on the same integrated circuit chip.

5. The computer system of claim 1, wherein the control circuit comprises a circuit path for routing the second clock signal, wherein the circuit path is located such that the second clock signal is synchronized with data values written to or read from the DRAM array.

6. The computer system of claim 1, wherein the frequency of the second clock signal is higher than the frequency of the first clock signal.

7. The computer system of claim 1, further comprising a DRAM main memory coupled to the bus.

8. The computer system of claim 7, further comprising a system controller, wherein the DRAM main memory is coupled to the bus through a direct data port of the system controller.

9. The computer system of claim 1, wherein the control circuit further comprises:

a phase locked loop (PLL) circuit coupled to the bus, wherein the PLL circuit generates the second clock signal in response to the first clock signal.

10. The computer system of claim 9, wherein the second clock signal has a frequency higher than the frequency of the first clock signal.

11. The computer system of claim 1, wherein the data input port is clocked twice during each cycle of the second clock signal and the data output port is clocked once during each cycle of the first clock signal.

12. The computer system of claim 11, wherein the frequency of the first clock signal is equal to the frequency of the second clock signal.

13. The computer system of claim 1, wherein the next level cache memory further comprises a refresh counter which stores the address of a DRAM array which is currently being refreshed.

14. The computer system of claim 13, wherein the next level cache memory further comprises a refresh collision comparator which compares the address of the DRAM array currently being refreshed with an address of a DRAM array requested by the CPU.

15. The computer system of claim 14, wherein the next level cache memory further comprises a CPU access delay circuit coupled to the refresh collision comparator, wherein the CPU access delay circuit transmits a ready handshake signal to the bus when the current refresh operation is completed.

16. The computer system of claim 1, wherein the control circuit comprises a burst sequencer which accesses the DRAM array in an asynchronous, self-timed manner with respect to the first clock signal.

17. The computer system of claim 16, wherein the burst sequencer and the DRAM array are fabricated on the same integrated circuit chip.

18. The computer system of claim 16, wherein the burst sequencer includes circuitry to access the DRAM array at both edges of the second clock signal.

19. The computer system of claim 16, wherein the burst sequencer further comprises a data amplifier circuit coupled to the DRAM array, wherein the data amplifier circuit has a plurality of synchronous regenerative data amplifier latches.

20. The computer system of claim 16, wherein the burst sequencer further comprises:

a row address register coupled to the bus, wherein the row address register receives a row address through the bus;

a column address register coupled to the bus, wherein the column address register receives a column address through the bus; and

a controller coupled to the bus and to the row and column address registers, wherein the controller provides a signal to the row and column address registers to cause the row and address registers to simultaneously latch the row and column addresses.

21. The computer system of claim 20, wherein the control circuit further comprises an address buffer coupled between the bus and the row and column address registers, wherein the address buffer stores row and column addresses required for a subsequent access while a current access is still in progress.

22. The computer system of claim 16, wherein the burst sequencer further comprises:

a sense amplifier circuit coupled to the DRAM array;

a data amplifier circuit having a plurality of data amplifiers; and

a column selector coupled between the sense amplifier circuit and the data amplifier circuit, wherein the column selector can be selected to isolate the sense amplifier circuit from the data amplifier circuit.

23. The computer system of claim 22, wherein the number of data amplifiers in the data amplifier circuit is selected to allow the data amplifier circuit to simultaneously store all of the data values required for a burst access of the DRAM array.

24. The computer system of claim 22, wherein the next level cache memory comprises a plurality of DRAM arrays operating in parallel.

25. The computer system of claim 22, wherein the burst sequencer further comprises:

a first delay circuit coupled to the sense amplifier circuit, wherein the first delay circuit enables the sense amplifier circuit in an asynchronous manner when data values are available from the DRAM array.

26. A computer system comprising:

a central processing unit (CPU);

a first level cache memory integrated with the CPU;

a bus coupled to the CPU wherein data is transferred on the bus in response to a first clock signal having a first clock frequency, wherein data is transferred on the bus at a first data rate; and

a next level cache memory coupled to the bus, wherein the next level cache memory comprises at least one dynamic random access memory (DRAM) array and a control circuit for controlling data transfer between the bus and the DRAM array, wherein the control circuit derives a second clock signal from the first clock signal, and causes data to be transferred to and from the DRAM array in response to a second clock signal at a second data rate which is faster than the first data rate, wherein the control circuit comprises a write buffer having a data output port coupled to the DRAM array and a data input port coupled to the bus, wherein the output port of the write buffer is clocked by the second clock signal and the input port of the write buffer is clocked by the first clock signal;

wherein a first set of data values is transferred from the bus to the write buffer via the data input port and stored in the write buffer, until a second set of data values is transferred from the bus to the write buffer via the data input port, after which the first set of data values is transferred to the DRAM array via the data output port.

27. The computer system of claim 26, wherein the write buffer is multi-depth buffer capable of storing at least one set of write burst data.

28. The computer system of claim 26, wherein the frequency of the second clock signal is higher than the frequency of the first clock signal.

29. The computer system of claim 26,

wherein the data input port is clocked once during each cycle of the first clock signal and the data output port is clocked twice during each cycle of the second clock signal.

30. The computer system of claim 29, wherein the frequency of the first clock signal is the same as the frequency of the second clock signal.

31. A method of using a DRAM array as a next level cache memory from a bus, the method comprising:

transferring data on the bus in response to a first clock signal having a first clock frequency, wherein data is transferred on the bus at a first data rate;

generating a second clock signal in response to the first clock signal;

operating the DRAM array in response to the second clock signal, wherein data is transferred to and from the DRAM array at a second data rate, the second data rate being greater than the first data rate;

reading a plurality of data values from the DRAM array to a plurality of sense amplifiers;

reading the data values from the sense amplifiers to a read buffer memory using the second clock signal to clock the data values into the buffer memory;

reading the data values from the read buffer memory to the bus using the first clock signal to clock the data values out of the read buffer memory so that the data values are read to the read buffer memory faster than the data values are read from the read buffer memory; and

precharging bit lines of the DRAM array during the time that the data values are read from the read buffer memory to the bus.

32. The method of claim 31, wherein the second clock signal has a higher frequency than the first clock signal.

33. The method of claim 31, further comprising the steps of:

transmitting data signals on the bus at either the rising edges or the falling edges of the first clock signal; and

performing read operations within the DRAM array at both the rising and falling edges of the second clock signal.

34. The method of claim 31, wherein the frequency of the second clock signal is the same as the frequency of the first clock signal.

35. The method of claim 31, further comprising the steps of:

performing a row access operation in the DRAM array in a self-timed asynchronous manner with respect to the second clock signal;

performing a column decode operation in the DRAM array in a self-timed asynchronous manner with respect to the second clock signal; and

performing a predetermined sequence of column select operations in the DRAM array, wherein the column select operations are synchronous with respect to the second clock signal.

36. The method of claim 31, wherein the frequency of the second clock signal is twice the frequency of the first clock signal.

37. The method of claim 31, wherein the step of precharging the DRAM array is completed before the step of reading the data values from the read buffer memory is completed.

38. The method of claim 31, further comprising the steps of:

transmitting data signals on the bus at either the rising edges or the falling edges of the first clock signal; and

performing write operations within the DRAM array at both the rising and falling edges of the second clock signal.

39. The method of claim 38, wherein the frequency of the second clock signal is the same as the frequency of the first clock signal.

40. The method of claim 31, further comprising the steps of:

reading a plurality of data values from the DRAM array to a plurality of sense amplifiers;

closing a plurality of column selector switches, thereby connecting the sense amplifiers to a plurality of data amplifiers, thereby causing the data values to be written to the data amplifiers;

reading the data values out of the data amplifiers;

opening the column selector switches, thereby disconnecting the sense amplifiers from the data amplifiers; and

precharging a plurality of DRAM cells in the DRAM array when the column selector switches are open.

41. The method of claim 40, wherein the step of reading the data values out of the data amplifiers further comprises the steps of:

writing the data values from the data amplifiers to the read buffer memory using the second clock signal to clock the data values into the read buffer; and

reading the data values form the read buffer memory to the bus using the first clock signal to clock the data values onto the bus.

42. The method of claim 41, wherein the step of precharging the DRAM array is completed before the step of reading the data values from the read buffer to the bus is completed.

43. A method of using a DRAM array as a next level cache memory from a bus, the method comprising:

transferring data on the bus in response to a first clock signal having a first clock frequency, wherein data is transferred on the bus at a first data rate;

generating a second clock signal in response to the first clock signal;

operating the DRAM array in response to the second clock signal, wherein data is transferred to and from the DRAM array at a second data rate, the second data rate being greater than the first data rate;

writing a first set of data values from the bus into a write buffer using the first clock signal to clock the first set of data values into the write buffer;

writing a second set of data values from the bus into the write buffer using the first clock signal to clock the second set of data values into the write buffer;

writing the first set of data values to the DRAM array after the time the second set of data values are written into the write buffer; and

precharging bit lines of the DRAM array during the time the second set of data values are written into the write buffer.

44. The method of claim 43, wherein the step of precharging the DRAM array is completed before the step of writing the second set of data values into the write buffer is completed.

45. The method of claim 43, wherein the step of synchronously writing the first set of data values to the DRAM array further comprises the steps of:

providing a data amplifier circuit and a sense amplifier circuit between the write buffer and the DRAM array;

disconnecting the data amplifier circuit from the sense amplifier circuit;

writing the first set of data values from the write buffer to the data amplifier circuit at the same time as performing a row access operation in the DRAM array; and

connecting the data amplifier circuit to the sense amplifier circuit, thereby causing the first set of data values to be provided to the DRAM array through the sense amplifier circuit.

46. A computer system comprising:

a central processing unit (CPU);

a first level cache memory integrated with the CPU;

a bus coupled to the CPU wherein data is transferred on the bus in response to a first clock signal having a first clock frequency, wherein data is transferred on the bus at a first data rate; and

a next level cache memory coupled to the bus, wherein the next level cache memory comprises at least one dynamic random access memory (DRAM) array and a control circuit for controlling data transfer between the bus and the DRAM array, wherein the control circuit derives a second clock signal from the first clock signal, and causes data to be transferred to and from the DRAM array in response to a second clock signal at a second data rate which is faster than the first data rate, the control circuit comprising a phase locked loop (PLL) circuit coupled to the bus, wherein the PLL circuit generates the second clock signal in response to the first clock signal, and wherein the second clock signal has the same frequency as the first clock signal.

47. A computer system comprising:

a central processing unit (CPU);

a first level cache memory integrated with the CPU;

a bus coupled to the CPU wherein data is transferred on the bus in response to a first clock signal having a first clock frequency, wherein data is transferred on the bus at a first data rate; and

a next level cache memory coupled to the bus, wherein the next level cache memory comprises at least one dynamic random access memory (DRAM) array and a control circuit for controlling data transfer between the bus and the DRAM array, wherein the control circuit derives a second clock signal from the first clock signal, and causes data to be transferred to and from the DRAM array in response to a second clock signal at a second data rate which is faster than the first data rate,

wherein the control circuit comprises a burst sequencer which accesses the DRAM array in an asynchronous, self-timed manner with respect to the first clock signal;

wherein the burst sequencer further comprises;

a sense amplifier circuit coupled to the DRAM array;

a data amplifier circuit having a plurality of data amplifiers and

a column selector coupled between the sense amplifier circuit and the data amplifier circuit, wherein the column selector can be selected to isolate the sense amplifier circuit from the data amplifier circuit;

wherein the burst sequencer further comprises;

a first delay circuit coupled to the sense amplifier circuit, wherein the first delay circuit enables the sense amplifier circuit in an asynchronous manner when the data values are available from the DRAM array; and

a second delay circuit coupled to the column selector, wherein the second delay circuit enables the column selector in an asynchronous manner when the data values are available in the send amplifier circuit.