Chiplet, multi-chiplet system, and method for a chiplet

Abstract

The invention relates to a chiplet for processing data associated with at least one data source, wherein the chiplet is designed to exchange the data with at least one other device, for example at least one other chiplet, in accordance with a streaming protocol.

Description

[0001] Description

[0002] title

[0003] State of the art

[0004] The disclosure relates to a chiplet for processing data associated with at least one data source, for example a video data source.

[0005] The disclosure further relates to a multi-chiplet system.

[0006] The disclosure further relates to a method for a chiplet.

[0007] Disclosure of the invention

[0008] Some examples relate to a chiplet for processing data associated with at least one data source, wherein the chiplet is configured to exchange the data with at least one further device, for example, at least one further chiplet, according to a streaming protocol. In some examples, this enables an efficient exchange of information associated with, for example, image data or video data.

[0009] In some examples, a chiplet is an integrated circuit with a predeterminable functionality or partial functionality. In some examples, a chiplet can be combined with at least one other chiplet, for example, on a common substrate, for example, to form a multi-chiplet system.

[0010] In some examples, the data comprises at least one of the following elements: a) image data, or b) video data. In further examples, information, for example data, of types other than those exemplified above can also be exchanged by the chiplet according to the streaming protocol. In some examples, a streaming protocol is a protocol, for example a standard, that can be used to transmit data, for example multimedia data such as video and / or audio, and that enables, for example, continuous sending and / or receiving and / or evaluation, for example "playing" or processing, for example further processing, e.g. of content, e.g. without an entire data set, e.g. an entire file, having to be transmitted.

[0011] In some examples, exchanging according to the streaming protocol comprises at least one of the following elements: a) exchanging the data using an, for example, existing, for example standardized, streaming protocol, or b) exchanging the data based on a streaming protocol, for example extending at least one aspect of an already existing streaming protocol, for example according to some examples, for example with respect to the application for the chiplet.

[0012] In some examples, the chiplet is configured to process the data, for example at least temporarily during a data exchange or after the data exchange.

[0013] In some examples, the processing comprises at least one of the following elements: a) processing a part of the data, for example before, for example completely, receiving the data, for example according to the streaming protocol, and / or during sending the data, for example according to the streaming protocol, wherein, for example in the case of image data, the part of the data characterizes and / or comprises at least one of the following elements: a1) line, or a2) picture element, for example pixel, or a3) tile.

[0014] In other words, in some examples, it may be provided that the data exchange according to the streaming protocol comprises an exchange, for example sending or receiving, of several data units or parts, e.g. of data sets, in the case of image data or video frames, e.g. the lines or pixels or tiles (e.g. "tiles") mentioned as an example.

[0015] In some examples, parts of frames, for example, are therefore

[0016] Streaming protocol interchangeable, i.e. e.g. receivable by the chiplet, whereby, for example, after receipt of a respective part, e.g. of a frame, processing of information associated with the respective part of the frame can already take place by the chiplet, for example while, e.g. at the same time, one or more further parts of the same frame are being received by the chiplet. This has the advantage that the chiplet receiving the frame does not have to wait for the frame to be received in full before it can begin processing information from the frame. This means that in some examples, efficient processing can take place, for example which at least partially overlaps in time with, for example, the reception of data, which in some examples can, for example, reduce latencies.

[0017] In further examples, the streaming protocol can be used for the chiplet to receive multiple parts of an image, e.g., a digital image, one after the other. For example, a data source can transmit one line of the image or one tile of the image to the chiplet one after the other using the streaming protocol. For example, after receiving the first line of the image or the first tile of the image, the chiplet can process the information characterized by the first line or first tile, e.g., without the second line or tile having already been received by the chiplet, etc.

[0018] The principle mentioned above using several examples can, in some examples, also be applied analogously to other parts of image data or frames or video data, for example for receiving data and / or for sending data according to the streaming protocol.

[0019] In some examples, the chiplet is configured to perform at least one of the following elements: a) receiving the data from the at least one further device, for example from the at least one further chiplet, wherein for example the at least one further device is arranged on the same substrate as the chiplet, or b) receiving the data from at least one data source, wherein for example the at least one data source is not arranged on or the same substrate as the chiplet, or c) sending the data to the at least one further device, for example to the at least one further chiplet. In some examples, the chiplet has a memory, for example buffer memory, for example main memory, for example for at least temporarily storing at least part of the data.

[0020] In some examples, the chiplet comprises a communication system, for example a network on chip, for example a crossbar, for exchanging data within the chiplet, for example with a processor or processing core, and / or with at least one external device, for example an external memory device.

[0021] In some examples, the chiplet comprises at least one processing unit configured to process image data and / or video data, for example, specialized in processing image data and / or video data, for example, at least one image signal processor.

[0022] Some examples relate to a multi-chiplet system comprising a plurality of chiplets, wherein at least one chiplet of the plurality of chiplets is configured according to the disclosure.

[0023] In some examples, the multi-chiplet system comprises at least one of the following elements: a) a management chiplet, for example for managing at least one aspect of the multi-chiplet system, or b) a processing chiplet comprising and / or characterizing at least one processing device, for example a computing device, for example a processor device, for example a central processor device, for example a CPU, or c) a graphics processor chiplet, for example a GPU chiplet, or d) a K1 chiplet for executing aspects of artificial intelligence, for example aspects of a method for machine learning, wherein, for example, the K1 chiplet comprises a local storage device, for example for at least temporarily storing data for executing aspects of artificial intelligence, for example for parameters such asWeights or hyperparameters for at least one, for example artificial, neural network.

[0024] In some examples, a) the management chiplet and the Kl chiplet are designed to exchange data with each other according to a streaming protocol, for example via a first data connection designed to execute a streaming protocol or a method based on a streaming protocol, or b) the management chiplet and the graphics processor chiplet are designed to exchange data with each other according to a streaming protocol, for example via a second data connection designed to execute a streaming protocol or a method based on a streaming protocol, or c) the management chiplet and the processing chiplet are designed to exchange data with each other according to a streaming protocol, for example via a third data connection designed to execute a streaming protocol ora data connection designed according to a method based on a streaming protocol, or d) at least two of the chiplets are designed to exchange data with each other, for example directly with each other, for example without using the management chiplet, according to a streaming protocol, for example via a fourth data connection designed to execute a streaming protocol or a method based on a streaming protocol.

[0025] Some examples relate to a product, for example a control unit, for example for a vehicle, for example motor vehicle, comprising at least one of the following elements: a) chiplet according to the disclosure, or b) chiplet system according to the disclosure, or c) device according to the disclosure.

[0026] Some examples relate to a method for a chiplet to process data associated with at least one data source, comprising: exchanging data with at least one other device, e.g., at least one other chiplet, according to a streaming protocol.

[0027] Some examples relate to an apparatus for carrying out the method according to the disclosure.

[0028] Some examples relate to a computer-readable storage medium comprising instructions that, when executed by a computer, cause the computer to perform the method according to the disclosure.

[0029] Some examples relate to a computer program comprising instructions that, when executed by a computer, cause the computer to perform the method according to the disclosure. Some examples relate to a data carrier signal that transmits and / or characterizes the computer program according to the disclosure.

[0030] Some examples relate to a use of the chiplet according to the disclosure and / or the multi-chiplet system according to the disclosure and / or the product according to the disclosure and / or the method according to the disclosure and / or the device according to the disclosure and / or the computer-readable storage medium according to the disclosure, and / or the computer program according to the disclosure, and / or the data carrier signal according to the disclosure for at least one of the following elements: a) exchanging data with at least one chiplet according to a streaming protocol, or b) using aspects of a streaming protocol for data exchange within a multi-chiplet system, c) using aspects of a streaming protocol for data exchange between at least one component of a multi-chiplet system and at least one component or device external to the multi-chiplet system, or d) reducing latency,for example, when processing and / or reproducing image data and / or video data, or e) reducing memory bandwidth requirements with respect to at least one interface of the chiplet and / or with respect to at least one memory device for a chiplet and / or for a multi-chiplet system.

[0031] Further features, possible applications, and advantages of the invention will become apparent from the following description of examples of the invention, which are illustrated in the figures of the drawing. All described or illustrated features, individually or in any combination, constitute the subject matter of the invention, regardless of their summary in the claims or their references, as well as regardless of their wording or representation in the description or drawing.

[0032] The drawing shows:

[0033] Fig. 1 schematically shows a simplified block diagram,

[0034] Fig. 2 schematically shows a simplified flow diagram,

[0035] Fig. 3 schematically shows a simplified flow diagram, Fig. 4 schematically shows a simplified block diagram,

[0036] Fig. 5 schematically shows a simplified block diagram,

[0037] Fig. 6 schematically shows a simplified block diagram,

[0038] Fig. 7 schematically shows a simplified block diagram,

[0039] Fig. 8 shows a simplified block diagram,

[0040] Fig. 9 schematically shows aspects of uses.

[0041] Some examples, Fig. 1 , 2, relate to a chiplet 100 for processing data DAT-DQ associated with at least one data source DQ, DQ', DQ", DQ'", wherein the chiplet 100 is designed to exchange 200 the data DAT-DQ with at least one further device 100', for example at least one further chiplet 100', according to a streaming protocol SP. In some examples, this enables an efficient exchange of information associated with, for example, image data or video data.

[0042] In some examples, a chiplet is an integrated circuit with a predeterminable functionality or partial functionality. In some examples, a chiplet, e.g., chiplet 100 according to Fig. 1, can be combined with at least one other chiplet 100', for example, on a common substrate (not shown), for example, to form a multi-chiplet system (see below, e.g., Figs. 4, 6).

[0043] In some examples, Fig. 1 , the data DAT-DQ comprises at least one of the following elements: a) image data DAT-BILD, or b) video data DAT-VID. In further examples, information, such as data, of types other than those exemplified above may also be exchanged by the chiplet 100 according to the streaming protocol SP.

[0044] In some examples, Fig. 1 , at least one data source DQ can be in data communication with the chiplet 100, for example not integrated into the chiplet 100. In further examples, Fig. 1 , a similar situation applies to the data source DQ'", for example with respect to the further chiplet 100'. In some examples, Fig. 1 , at least one data source DQ' can be in data communication with the chiplet 100, for example integrated into the chiplet 100. In further examples, Fig. 1 , a similar situation applies to the data source DQ", for example with respect to the further chiplet 100'.

[0045] In some examples, a data connection DV, e.g., a bus system, may be used between the chiplets 100, 100' for data exchange 200 (Fig. 2) according to the streaming protocol SP. In some examples, the data connection DV may be of the type UCIe (Universal Chiplet Interconnect Express) or CXL (Compute Express Link), e.g.,

[0046] In some examples, a data exchange according to the streaming protocol SP can also take place with at least one further device, e.g. data sink DS, wherein, for example, the further device DS is not designed as a chiplet.

[0047] In some examples, Fig. 2, the exchange 200 according to the streaming protocol SP comprises at least one of the following elements: a) exchange 200a of the data using an, for example, existing, for example standardized, streaming protocol, or b) exchange 200b of the data based on a streaming protocol SP, wherein, for example, at least one aspect of an already existing streaming protocol is extended, for example according to some examples, for example with respect to the application for the chiplet 100.

[0048] In some examples, Fig. 1 , 2, the chiplet 100 is configured to process 200 the data DAT-DQ, for example at least temporarily during a data exchange 200 or after the data exchange 200, wherein, for example, processed data DAT-DQ' is obtained.

[0049] In some examples, Fig. 2, the processing 202 comprises at least one of the following elements: a) processing 202a of a part DAT-DQ_PART of the data DAT-DQ, for example before, for example completely, receiving the data, for example according to the streaming protocol SP, and / or during sending the data, for example according to the streaming protocol SP, wherein, for example in the case of image data DAT-IMAGE, the part DAT-DQ_PART of the data characterizes and / or comprises at least one of the following elements: a1) line, or a2) picture element, for example pixel, or a3) tile.

[0050] In other words, in some examples, it may be provided that the data exchange according to the streaming protocol SP comprises an exchange 200, for example sending and / or receiving, of several data units or parts, e.g. of data sets, in the case of image data or video frames, e.g. the lines or pixels or tiles mentioned as an example (e.g. "tiles").

[0051] In some examples, parts of frames, for example, are thus interchangeable according to the streaming protocol SP, i.e., can be received by the chiplet 100 (Fig. 1), for example. After the reception of a respective part, e.g., of a frame, processing 202a of information associated with the respective part of the frame can already take place by the chiplet 100, for example while, e.g., simultaneously, one or more further parts of the same frame are received by the chiplet 100. This has the advantage that the chiplet 100 receiving the frame does not have to wait for the complete reception of the frame before it can begin processing 202a of information in the frame. As a result, in some examples, efficient processing can be achieved, which, for example, at least partially overlaps in time with, for example, a reception (see, for example, block 200 according to Fig.2) of data, which can reduce latencies in some examples.

[0052] In further examples, Fig. 1 , the streaming protocol SP can be used for the chiplet 100 to receive multiple parts of an image, for example a digital image, for example one after the other.

[0053] For example, a data source DQ", 100' can transmit one line of the image or one tile of the image successively to the chiplet 100 using the streaming protocol SP, and, for example, after receiving the first line of the image or the first tile of the image, the chiplet 100 can process the information characterized by the first line or first tile, for example, without the second line or tile having already been received by the chiplet 100, etc.

[0054] The principle described above using several examples can, in some examples, also be applied analogously to other parts of image data or frames or video data, for example for receiving data and / or for sending data according to the streaming protocol.

[0055] In some examples, Fig. 3, the chiplet 100 (Fig. 1) is designed to carry out at least one of the following elements: a) receiving 210 the data DAT-DQ from the at least one further device, for example from the at least one further chiplet 100', wherein for example the at least one further device 100' is arranged on a same substrate as the chiplet 100, or b) receiving 212 the data DAT-DQ from at least one data source, wherein for example the at least one data source is not arranged on or the same substrate as the chiplet 100, or c) sending 214 the data DAT-DQ to the at least one further device 100', DS, for example to the at least one further chiplet 100'.

[0056] Fig. 4 schematically shows a simplified block diagram illustrating a transfer of data from a first chiplet 100a to a second chiplet 100b according to some examples. For example, the two chiplets 100a, 100b form a multi-chiplet system 1000. A common substrate of the two chiplets 100a is not depicted in Fig. 4.

[0057] In some examples, a data set D is present in the first chiplet 100a, which is to be transmitted to the second chiplet 100b. In some examples, individual parts D-T1, D-T2, D-T3, ... of the data set D are transmitted one after the other via a data connection DV from the first chiplet 100a to the second chiplet 100b using a streaming protocol SP, for example as long as or as often as necessary until the entire data set D has been received in the second chiplet 100b. Due to the transmission using the streaming protocol SP, the second chiplet 100b can, for example, process the first part D-T1 of the data set after receiving the first part D-T1, for example while the chiplet 100b waits for a transmission of the further parts D-T2, D-T3 or while the chiplet 100b receives the further parts D-T2, ...

[0058] In some examples (see Fig. 4), the data connection DV can be of the type U-Cle (Universal Chiplet Interconnect Express) or CXL (Compute Express Link), for example. Optionally, in some examples, at least one, e.g., additional, control line SL can be provided between the chiplets 100a, 100b, which can be used, e.g., for signaling when using the streaming protocol SP.

[0059] In some examples, Fig. 4, the data set D represents an image, and the parts D-T1 , DT-2, D-T3, ... characterize, for example, individual rows or pixels or groups of pixels, e.g., tiles, of the image.

[0060] In some examples, Fig. 5, the chiplet 100b has a memory 110, for example a buffer memory, for example working memory (e.g. RAM), for example for at least temporarily storing at least part of the data, for example the data received according to the streaming protocol via the data connection DV, for example from the chiplet 100a according to Fig. 4.

[0061] In some examples, Fig. 5, the buffer memory 110 is dimensioned such that it can hold at least a portion D-T1 of the data, e.g., it can store at least one line or one tile of, for example, a received image at least temporarily.

[0062] In some examples, Fig. 5, the chiplet 100b has a communication system 120, for example network on chip, for example crossbar, for exchanging data within the chiplet 100b, for example with a processor or processing core 130, and / or with at least one external device 140, for example an external memory device.

[0063] In some examples, a data path of a part D-T1 of a data set D, e.g., an image, that can be received and processed by the chiplet 100b may comprise, for example, the following stages: 1. Reception by means of the streaming protocol SP and buffering in the memory 110, 2. Transmission to the processor 130 by means of the chiplet-internal communication system 120, 3. Transmission from the processor 130 via the chiplet-internal communication system 120 to the external memory 140, e.g., as a part (not shown) of the data set D that is processed by the processor 130.

[0064] In some examples, Fig. 5, the chiplet 100b comprises at least one processing unit 130 configured for processing image data and / or video data, for example, specialized for processing image data and / or video data, for example, at least one image signal processor, for example, an image signal processor (ISP). In some examples, Fig. 5, the processing unit 130 may, for example, characterize or comprise one or more ISPs.

[0065] Some examples, Fig. 4, 6, relate to a multi-chiplet system 1000, 1000a comprising a plurality of chiplets, wherein at least one chiplet of the plurality of chiplets is configured according to the disclosure.

[0066] For example, in the multi-chiplet system 1000 according to Fig. 4, both chiplets 100a, 100b can be formed according to the disclosure.

[0067] In other examples, only one of a plurality of chiplets 100a, 100b, for example, chiplet 100b, of multi-chiplet system 1000 may be configured according to the disclosure. In this case, chiplet 100b configured according to the disclosure may perform a data exchange according to a streaming protocol, for example, with a unit arranged externally with respect to multi-chiplet system 1000.

[0068] In some examples, Fig. 6, the multi-chiplet system 1000a has at least one of the following elements: a) a management chiplet E1, for example for managing at least one aspect of the multi-chiplet system 1000a, or b) a processing chiplet E2 having and / or characterizing at least one processing device, for example a computing device, for example a processor device, for example a central processor device, for example a CPU, or c) a graphics processor chiplet, for example a GPU chiplet, E3 or d) a Kl-chiplet E4 for executing aspects of artificial intelligence, for example aspects of a method for machine learning, wherein, for example, the Kl-chiplet E4 has a local storage device E4a, for example for at least temporarily storing data for executing aspects of artificial intelligence, for example for parameters such asWeights or hyperparameters for at least one, for example artificial, neural network NN.

[0069] In some examples, Fig. 6, the management chiplet E5 comprises a plurality of processing devices E5a, E5b, for example ISPs respectively.

[0070] In some examples, Fig. 6, the management chiplet E5 has a memory controller E1a for connecting an external memory device E6. In some examples, a plurality of data sources DQ-1, DQ-2 are provided, which supply, for example, video data or image data, for example for processing by the multi-chip system 1000a. For example, reference symbol DQ-1 symbolizes a plurality of video data sources such as cameras, which supply, for example, a plurality of video data streams, for example four video data streams, for example each with a resolution of, for example, 3 megapixels and, for example, a frame rate of approximately 44 Hz. For example, reference symbol DQ-2 symbolizes a plurality of video data sources such as cameras, which supply, for example, a plurality of video data streams, for example five video data streams, for example each with a resolution of, for example, 8 megapixels and, for example, a frame rate of approximately 35 Hz.

[0071] In some examples, Fig. 6, the management chiplet E1 and the Kl chiplet E4 are designed to exchange data with each other according to a streaming protocol, for example via a first data connection DV-SP-1 designed to execute a streaming protocol or a method based on a streaming protocol.

[0072] In some examples, Fig. 6, the management chiplet E1 and the graphics processor chiplet E3 are designed to exchange data with each other according to a streaming protocol, for example via a second data connection DV-SP-2 designed to execute a streaming protocol or a method based on a streaming protocol.

[0073] In some examples, Fig. 6, the management chiplet E1 and the processing chiplet E2 are designed to exchange data with each other according to a streaming protocol, for example via a third data connection DV-SP-3 designed to execute a streaming protocol or a method based on a streaming protocol.

[0074] In some examples, Fig. 6, at least two of the chiplets, e.g., the CPU chiplet E2 and the GPU chiplet E3, are configured to exchange data with each other, for example directly with each other, for example without using the management chiplet E1, according to a streaming protocol, for example via a fourth data connection DV-SP-4 configured to execute a streaming protocol or a method based on a streaming protocol. In some examples, any combinations of the above examples are possible, for example, with regard to the data connections DV-SP-1, DV-SP-2, DV-SP-3, DV-SP-4.

[0075] For example, in the data connection DV-SP-1, the reference symbol "N" symbolizes a plurality of parts of, e.g., entire, data sets, i.e., a plurality of lines or tiles, e.g., each of a frame, which are exchanged according to the streaming protocol via the data connection DV-SP-1. Optionally, at least one, e.g., additional, control line (see also reference symbol SL in Fig. 4) can be provided.

[0076] In some examples, the data connections DV-SP-1, DV-SP-2, DV-SP-3, DV-SP-4 may each operate with or be based on the same streaming protocol SP.

[0077] In some examples, at least some of the data connections DV-SP-1, DV-SP-2, DV-SP-3, DV-SP-4 may, for example, operate with different streaming protocols or parameters relating to a data exchange according to the streaming protocol.

[0078] In some examples, Fig. 6, parts N-1, N-2, N-3, ... of the data that can be processed or are processed by the multi-chiplet system 1000a can be stored in the memory E6 at least temporarily.

[0079] Some examples, Fig. 7, relate to an apparatus 300 for carrying out the method according to the disclosure.

[0080] In further examples, Fig. 7, it is provided that the device 300 comprises: a computing device ("computer") 302 having at least one computing core 302a, a memory device 304 assigned to the computing device 302 for at least temporarily storing at least one of the following elements: a) data DAT (e.g., associated with information for using the streaming protocol SP and / or at least parts of the data transmitted and / or to be transmitted according to the streaming protocol SP), b) computer program PRG, for example, for executing the method according to the examples. In further examples, Fig. 7, the memory device 304 comprises a volatile memory (e.g., random access memory (RAM)) 304a, and / or a non-volatile (NVM) memory (e.g., flash EEPROM) 304b, or a combination thereof or with other memory types not explicitly mentioned.

[0081] Further examples, Fig. 7, relate to a computer-readable storage medium SM, comprising instructions PRG which, when executed by a computer 302, cause the computer 302 to carry out the method according to the examples.

[0082] Further examples, Fig. 7, relate to a computer program PRG comprising instructions which, when the program PRG is executed by a computer 302, cause the computer 302 to carry out the method according to the examples.

[0083] Further examples, Fig. 7, relate to a data carrier signal DCS that characterizes and / or transmits the computer program PRG according to the examples. The data carrier signal DCS can be received, for example, via an optional data interface 206 of the device 200. Likewise, the data D, for example, can be transmitted (sent and / or received) via the optional data interface 206. In further examples, the communication or data communication with the further unit 30 (Fig. 9) can also take place via the optional data interface 206.

[0084] In some examples, the device 300 or corresponding functionality is integrated into at least one chiplet 100, E1 and / or a multi-chiplet system.

[0085] Some examples, Fig. 8, relate to a product, for example a control unit, 1 , for example for a vehicle, for example motor vehicle, 10, comprising at least one of the following elements: a) chiplet 100 according to the disclosure, or b) chiplet system 1000 according to the disclosure, or c) device 300 according to the disclosure.

[0086] Some examples, Fig. 2, relate to a method for a chiplet 100 for processing data associated with at least one data source, comprising: exchanging 200 data DAT-DQ with at least one further device 100', for example at least one further chiplet, according to a streaming protocol SP. Some examples, Fig. 9, relate to a use 400 of the chiplet 100 according to the disclosure and / or the multi-chiplet system 1000, 1000a according to the disclosure and / or the product 10 according to the disclosure and / or the method according to the disclosure and / or the device 300 according to the disclosure and / or the computer-readable storage medium SM according to the disclosure, and / or the computer program PRG according to the disclosure, and / or the data carrier signal DCS according to the disclosure for at least one of the following elements: a) exchanging 401 data with at least one chiplet 100, 100', 100a, 100b, E1, E2, E3,E4 according to a streaming protocol SP, or b) using 402 aspects of a streaming protocol SP for data exchange within a multi-chiplet system 1000, 1000a, c) using 403 aspects of a streaming protocol SP for data exchange between at least one component of a multi-chiplet system 1000, 1000a and at least one component or device external to the multi-chiplet system 1000, 1000a, or d) reducing 404 a latency, for example during processing and / or playback of, for example, image data and / or video data, or e) reducing 405 memory bandwidth requirements with respect to at least one interface of the chiplet and / or with respect to at least one memory device E6 for a chiplet and / or for a multi-chiplet system.

Claims

Claims 1 . Chiplet (100) for processing data (DAT-DQ) associated with at least one data source (DQ; DQ'; DQ"; DQ'"), wherein the chiplet (100) is designed to exchange (200) the data (DAT-DQ) with at least one further device (100'; DS), for example at least one further chiplet (100'), according to a streaming protocol (SP).

2. Chiplet (100) according to claim 1, wherein the data (DAT-DQ) comprises at least one of the following elements: a) image data (DAT-BILD), or b) video data (DAT-VID).

3. Chiplet (100) according to at least one of the preceding claims, wherein the exchanging (200) according to the streaming protocol (SP) comprises at least one of the following elements: a) exchanging (200a) the data (DAT-DQ) using an existing, for example standardized, streaming protocol (SP), or b) exchanging (200b) the data (DAT-DQ) based on a streaming protocol (SP).

4. Chiplet (100) according to at least one of the preceding claims, wherein the chiplet (100) is designed to process the data (DAT-DQ) (202).

5. Chiplet (100) according to claim 4, wherein the processing (202) comprises at least one of the following elements: a) processing (202a) a part (DAT-DQ-PART) of the data (DAT-DQ), for example before, for example completely, receiving the data (DAT-DQ) and / or during sending the data (DAT-DQ), wherein, for example in the case of image data (DAT-IMAGE), the part (DAT-DQ-PART) of the data (DAT-DQ) characterizes and / or comprises at least one of the following elements: a1) line, or a2) picture element, for example pixel, or a3) tile 6. Chiplet (100) according to at least one of the preceding claims, wherein the chiplet (100) is designed to carry out at least one of the following elements: a) receiving (210) the data (DAT-DQ) from the at least one further device, for example from the at least one further chiplet (100'), or b) receiving (212) the data (DAT-DQ) from at least one data source (DQ; DQ'; DQ"; DQ'"), or c) sending (214) the data (DAT-DQ) to the at least one further device (100'), for example to the at least one further chiplet (100').

7. Chiplet (100) according to at least one of the preceding claims, comprising a memory (110), for example a buffer memory, for at least temporarily storing at least part of the data (DAT-DQ).

8. Chiplet (100) according to at least one of the preceding claims, comprising a communication system (120), for example network on chip, for example crossbar, for exchanging data within the chiplet (100), for example with a processor or computing core (130), and / or with at least one external device (140), for example an external memory device.

9. Chiplet (100) according to at least one of the preceding claims, comprising at least one processing unit (E5a, E5b) which is configured for processing image data (DAT-BILD) and / or video data (DAT-VID), for example is specialized in processing image data (DAT-BILD) and / or video data (DAT-VID), for example at least one image signal processor, for example image signal processor.

10. Multi-chiplet system (1000; 1000a), comprising several chiplets (100a, 100b; E1, E2, E3, E4), wherein at least one chiplet of the plurality of chiplets (100a, 100b; E1, E2, E3, E4) is designed according to at least one of the preceding claims.

11. Multi-chiplet system (1000; 1000a) according to claim 10, comprising at least one of the following elements: a) a management chiplet (E1), for example for managing at least one aspect of the multi-chiplet system (1000; 1000a), or b) a processing chiplet (E2) comprising and / or characterizing at least one processing device, for example a computing device, for example a processor device, for example a central processor device, for example a CPU, or c) a graphics processor chiplet, for example a GPU chiplet, or d) a K1 chiplet (E4) for executing aspects of artificial intelligence, for example aspects of a method for machine learning, wherein for example, the Kl chiplet (E4) has a local memory device (E4a), for example for at least temporarily storing data for the execution of aspects of artificial intelligence, for example for parameters such as weights or hyperparameters for at least one, for example artificial, neural network (NN).

12. Multi-chiplet system (1000; 1000a) according to claim 11, wherein a) the management chiplet (E1) and the KI chiplet (E4) are designed to exchange data with one another according to a streaming protocol (SP), for example via a first data connection (DV-SP-1) designed to execute a streaming protocol (SP) or a method based on a streaming protocol (SP), or wherein b) the management chiplet (E1) and the graphics processor chiplet (E3) are designed to exchange data with one another according to a streaming protocol (SP), for example via a second data connection (DV-SP-1) designed to execute a streaming protocol (SP) or a method based on a streaming protocol (SP).a data connection (DV-SP-2) designed for a method based on a streaming protocol (SP), or wherein c) the management chiplet (E1) and the processing chiplet (E2) are designed to exchange data with one another according to a streaming protocol (SP), for example via a third data connection (DV-SP-3) designed to execute a streaming protocol (SP) or a method based on a streaming protocol (SP), or wherein d) at least two of the chiplets (E2, E3, E4) are designed to exchange data with one another, for example directly with one another, for example without using the management chiplet (E1), according to a streaming protocol (SP), for example via a fourth data connection (DV-SP-4) designed to execute a streaming protocol (SP) or a method based on a streaming protocol (SP).

13. Product (1), for example a control unit, for example for a vehicle (10), for example a motor vehicle, comprising at least one of the following elements: a) chiplet (100) according to at least one of claims 1 to 9, or b) chiplet system (1000; 1000a) according to at least one of claims 10 to 12, or c) device (300) according to claim 12.

14. Method for a chiplet (100) for processing data (DAT-DQ) associated with at least one data source (DQ; DQ'; DQ"; DQ'"), comprising: exchanging (200) data with at least one further Device (100'; DS), for example at least one further chiplet (100'), according to a streaming protocol (SP).

15. Device (300) for carrying out the method according to claim 14.

16. A computer-readable storage medium (SM) comprising instructions (PRG) which, when executed by a computer (302), cause the computer to carry out the method according to claim 14.

17. A computer program (PRG) comprising instructions which, when the program (PRG) is executed by a computer (302), cause the computer (302) to carry out the method according to claim 14.

18. Data carrier signal (DCS) that transmits and / or characterizes the computer program (PRG) according to claim 17.

19. Use (400) of the chiplet (100) according to at least one of claims 1 to 9 and / or the multi-chiplet system (1000) according to at least one of claims 10 to 12 and / or the product (1) according to claim 13 and / or the method according to claim 14 and / or the device (300) according to claim 15 and / or the computer-readable storage medium (SM) according to claim 16, and / or the computer program (PRG) according to claim 17, and / or the data carrier signal (DCS) according to claim 18 for at least one of the following elements: a) exchanging (401) data with at least one chiplet (100) according to a streaming protocol (SP), or b) using (402) aspects of a streaming protocol (SP) for a data exchange within a multi-chiplet system (1000; 1000a), c) using (403) aspects of a Streaming protocol (SP) for data exchange between at least one component (100; E1) of a multi-chiplet system (1000;1000a) and at least one component or device external to the multi-chiplet system (1000; 1000a), or d) reducing (404) a latency, for example during processing and / or reproduction of, for example, image data and / or video data, or e) reducing (405) memory bandwidth requirements with respect to at least one interface of the chiplet (100) and / or with respect to at least one memory device for a chiplet (100) and / or for a multi-chiplet system (1000; 1000a), f) data processing, for example continuous data processing, for example image data processing, for example video data processing, in the automotive sector.;

Images