Patent Application
20110002098
The present invention relates to chip heat dissipation, and more particularly to a method and circuit device for multiple chip heat dissipation on a single circuit board.
This Application claims priority to Chinese Patent Application 200910148722.0, filed Jul. 1, 2009.
As higher computing performance is continually sought, the capability of existing single processor or multiple processor cores cannot meet the requirements for increasingly complicated graphics processes. Currently, there are some solutions of multiple core processors or multiple graphics processing units (GPUs) on a single graphics card to meet the above requirements, such as the Gemini™ GPU technology developed by ATI Corporation and SLI™ technology developed by NVIDIA Corporation. Based on these technologies, more than one GPU can be arranged on a single graphics card to increase computing performance. However, a problem of a large quantity of heat being generated by the multiple GPUs will occur accompanying the high performance.
Referring to
The GPU_1201 and the GPU_2202 will also output an overheat alert signal from their pins OVERTEM_1 and OVERTEM_2 respectively when the GPU is overheated and needs to be shut down immediately. At this state, the core temperature of the GPU is usually 125° C. or above, for example. These overheat alert signals will be input into the OR_1205 for a logical OR operation, and the resultant signal will be input to the overheat protection unit 203, which herein is a latch circuit, to protect the GPU and the graphics card from overheating. Specifically, an enabling signal will be input to the overheat protection unit 203 when either the GPU_1201 or the GPU_2202 reach a core temperature that would cause overheating. The enabling signal will trigger the state of the overheat protection unit 203 to change and be latched, and thus a high level signal (referred to as OUTPUT signal herein) will be output from the OUTPUT pin of the overheat protection unit 203. Simultaneously, a control signal will be output from the SHUTDOWN pin 203b and then used to shut down the power supply for both of the GPUs. Only if the user inputs an enabling signal from RESET pin 203a, will the latched state of the overheat protection unit 203 be released and the GPUs allowed to resume operation. The OUTPUT signal will also be sent to the OR_2206 to make a logical OR operation with the PWM_1 signal, with the resultant signal used to control the fan to operate at a certain speed in order to cool down the GPUs.
It can be seen from the operation described above that the operation of the fan will be controlled by only the state of the GPU_1201 (via the PWM_1 signal) and the overheat state of both the GPU_1201 and the_GPU_2202 (via the OUTPUT signal). In a case where the GPU_1201 is working with a light load and the GPU_2202 is working with a heavy load but not overheated yet, the fan may rotate slowly or even stop due to a small duty cycle of the PWM_1 signal. However, the core temperature of the GPU_2202 may increase very rapidly since there is no way to cool it down. As a result, the GPU_2202 will reach its overheat alert temperature very quickly and then both of the GPUs will be shut down by the enabling of the overheat protection unit 203. The user has to reset the overheat protection unit 203 manually in order to resume work, which will bring a lot of inconvenience to the user if it happens very frequently. Therefore, there is a need for an improved circuit and method to start the fan to cool down the GPUs timely, preferably before any of GPUs reaches a state of overheating, in order to prevent the GPUs from shutting down frequently due to overheating.
This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.
In consideration of the above-identified shortcomings of the art, a circuit for controlling heat dissipation means for multiple units on a circuit board is provided. The circuit may comprise a first logical OR operation unit connected to said multiple units. The first logical OR operation unit performs a logical OR operation on a first set of signals output from each of the multiple units. A resultant signal is output from the first logical OR operation to control an overheat protection unit connected to the first logical OR operation unit. A signal within the first set of signals represents whether a unit from which the signal is output has reached an overheated status. The overheat protection unit shuts down the multiple units when any one of the multiple units is overheated.
A second logical OR operation unit is connected to the multiple units and the overheat protection unit. The second logical OR operation unit performs a logical OR operation on a second signal output from any one of the multiple units, a third set of signals output from each of the multiple units other than that from which the second signal is output and a fourth signal output from the overheat protection unit. The second logical OR operation unit also outputs a resultant signal to control operation of the heat dissipation means. A second signal represents a relationship between the workload and the core temperature of a unit from which the second signal is output. The third set of signals represents whether respective units from which signals within the third set of signals are output have reached an alert status and the fourth signal represents whether any of said multiple units has reached an overheated status.
According another aspect of the present invention, a method for controlling a heat dissipation means for multiple units on a circuit board is provided. The method may comprise selecting a first set of signals output by each of the multiple units, wherein a signal within the first set of signals represents whether a unit from which the signal is output has reached an overheated status. A logical OR operation is then performed with the selected set of first signals and a resulting signal is used to control an overheat protection unit. The overheat protection unit shuts down each of the_said multiple units when any one of said multiple units is overheated. Then a second signal output by one of said multiple units is selected. The second signal output represents a relationship between a workload and a core temperature of a unit from which the second signal is output. A third set of signals output from each of the multiple units other than that from which the second signal is output is also selected. The third set of signals represent whether respective units from which signals within the third set of signals are output have reached an alert status. A fourth signal output from the overheat protection unit is selected and the fourth signal represents whether any of the units has reached an overheated status. A logical OR operation is then performed with the second signal, the third set of signals and the fourth signal, with the resulting signal used to control operation of the heat dissipation means.
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention. In the drawings,
a shows a circuit diagram of an example circuit for controlling the fan of a dual GPU graphics card according to an embodiment of the present invention;
b is a diagram showing an example implementation of the control circuit shown in
In the following description, numerous specific details are set forth to provide a more thorough understanding of the present invention. However, it will be apparent to one of skill in the art that the present invention may be practiced without one or more of these specific details. In other instances, certain well-known features have not been described in order to avoid obscuring the present invention.
Referring to
The circuit according to the present embodiment has a fan control unit 304 controlled by a signal resulting from a logical OR operation of three signals via the logical OR operation unit OR_2306. These signals are a signal output from a pin PWM_1 of the GPU_1301 (representing the relationship between the core temperature and the workload of the GPU_1201 and preferably having a square wave shape), a signal output from a pin ALERT_2 of the GPU_2302 (referred to as the ALERT_2 signal and representing whether the GPU_2302 has reached an alert status), and a signal output from the pin OUTPUT of the overheat protection unit 303 (representing whether any of GPUs have reached an overheated status). For example, the input signal Fan Control Input of the fan control unit 304 will be:
Fan Control Input=(PWM—1 of GPU—1) logical OR (OUTPUT of Overheat Protection Unit) logical OR (ALERT—2 of GPU—2)
The ALERT_2 signal reflects a critical working state of the GPU_2302, which means the GPU_2302 is now in a state of full load and cannot endure any more load when the ALERT_2 signal is enabled. The GPU_2302 would then need to be cooled down immediately. This status usually corresponds to a GPU core temperature of 95° C., for example.
In controlling the fan control unit 304 in the manner described above, when both the GPU_1301 and the GPU_2302 are working with a normal load, neither the ALERT_2 signal nor the OUTPUT signal will be enabled. The fan control unit 304 will be controlled only by the PWM_1 signal of the GPU_1301. The higher the core temperature and the higher the workload of the GPU_1301, the larger the duty cycle of the square wave and the faster the fan will rotate. When the duty cycle reaches 100%, representing that the GPU_1301 may be in a state of full load, the fan will rotate at full speed in order to cool down the GPU_1 as quickly as possible. As the workload of the GPU_2 increases, the core temperature of the GPU_2 rises and when it reaches its alert temperature, an enabling ALERT_2 signal will be output and will result in an enabling signal generated at the output of the OR_2306. This resultant signal will then be sent to the fan control unit 304 to make the fan rotate at full speed to cool down the GPU_2. If either of the workload of the GPU_1301 or the GPU_2302 continues to increase, the core temperature will continue to rise and reach an overheat alert temperature. At this time, the overheat protection unit 303 will output an enabling signal at the pin SHUTDOWN to shut down both the GPU_1301 and the GPU_2302 to prevent them from overheating and becoming damaged. In this way, the GPUs will be cooled down in a timely fashion before they reach a temperature that would cause them to overheat. For example, they may be cooled down to 30° C. below the temperature that would cause them to overheat, thereby avoiding frequent shutdown of the GPUs due to rapid overheating.
Heat dissipation is achieved by introducing an ALERT_2 signal of the GPU_2302 and performing a logical OR operation with the existing signals to control the fan. The reason the ALERT_2 signal of the GPU_2302 is chosen rather than the PWM_2 signal of the GPU_2302 (which also reflects the change of the core temperature of the GPU as its workload varies) is because the PWM signals may be asynchronous between the different GPUs. If the fan control unit 304a is controlled by a signal resulting from a logical OR operation on the PWM_1 signal of the GPU_1301 and the PWM_2 signal of the GPU_2302, it may appear that when both of the GPU_1301 and the GPU_2302 work at a light load, they will both output a PWM signal having a low duty cycle, but the signal after the logical OR operation may have a large duty cycle due to an asynchronism between the signals. Thus the fan control unit 304 may mistakenly determine that the GPU has a high core temperature and should be cooled down immediately, resulting in the fan rotating at a high speed. Therefore, noise caused by the rotation of the fan will be introduced, which is disadvantageous for obtaining a quiet working status for the system.
b illustrates a diagram of an example implementation of the control circuit shown in
Similarly, in this embodiment the overheat protection unit 303 may be a low-level enabled circuit, and the OR_1305 may be implemented with diodes whose ON direction is opposite to the direction of the signal transmission in the circuit. As shown, each signal line input into the OR_1305 is provided by a diode respectively. If either the signal OVERTEP_1 of the GPU_1301 or the signal OVERTEP_2 of the GPU_2302 becomes low, the input of the overheat protection unit 303 will be brought to a low level. If either of the above two signals is high, its respective diode will be OFF and block the signal from being input to the overheat protection unit 303. Thus, the signal competition at the input of the overheat protection unit 303 can be avoided.
It will be appreciated by those of ordinary skill in the art that the diode used herein is only an exemplary implementation for the first and second logical OR operation units 305 and 306 and any other circuit form commonly known and used in the art may be also adopted.
Referring next to
Referring next to
Besides the multiple GPUs on a graphics card, the circuit and method as described above may be applied in various other chips, processors, circuit boards and add-in cards in cases where there is a need for heat dissipation of multiple units. The fan may also be replaced by any other means for heat dissipation.
It is noted that the foregoing examples have been provided merely for the purposes of explanation and are in no way to be construed as limiting of the present invention. While the invention has been described with reference to various embodiments, it is understood that the words which have been used herein are words of description and illustration, rather than words of limitation. Further, although embodiments have been described herein with reference to particular means and materials, the invention is not intended to be limited to the particulars disclosed herein; rather, the invention extends to all functionally equivalent structures, methods and uses, such as are within the scope of the appended claims. Those skilled in the art, having the benefit of the teachings of this specification, may effect numerous modifications thereto and changes may be made without departing from the scope and spirit of the invention in its aspects.
| Number | Date | Country | Kind |
|---|---|---|---|
| 200910148722.0 | Jul 2009 | CN | national |
