Back to results
US 2021/0279104 A1Published application
High bandwidth connection between processor dies
AssigneeIntel Corporation
InventorsALTUG KOKER, Abhishek R. Appu, Kiran C. Veernapu, Joydeep Ray, Balaji Vembu
Life of this patent
Estimating the term…
Abstract
Methods and apparatus relating to techniques for avoiding cache lookup for cold cache. In an example, an apparatus comprises logic, at least partially comprising hardware logic, to receive a completion acknowledgment from the plurality of graphics processing units and in response to a determination that the workload is finished, to terminate one or more communication connections on the interconnect bridge. Other embodiments are also disclosed and claimed.
Claims19
1-20. (canceled)
An apparatus, comprising:
- at least one processing unit (PU) disposed on a first PU die;
- at least a first graphics processing unit (GPU) disposed on a first GPU die and a second graphics processing unit (GPU) disposed on a second GPU die;
- and
- a communication network to provide communication connections between the first CPU die and the and the first GPU die;
- the processing unit to:
- receive graphics metadata from an application which is to execute on the at least one PU, wherein the graphics metadata indicates a level of graphics processing activity associated with the application;
- determine, from the graphics metadata, an amount of graphics processing resources to be dedicated to the application, the graphics processing resources comprising a plurality of graphics processing units located on the first GPU and the second GPU;
- configure one or more communication connections between the PU and the first GPU and the PU and the second GPU;
- transmit the graphics data for the application from the PU to the plurality of graphics processing units via the communication network;
- receive a completion acknowledgment from the plurality of graphics processing units;
- and
- in response to a determination that the workload is finished, to terminate one or more communication connections on the communication network.
9 more elements
The apparatus of claim 26, wherein:
- the first PU and the first GPU and the second GPU are disposed on a first layer of the semiconductor device;
- and
- the communication network traverses between the first PU and the first GPU and the second GPU on the first layer of the semiconductor package.
An electronic device, comprising:
- a printed circuit board;
- an input/output (I/O) module on the printed circuit board to provide an interface to a compute system;
- an apparatus mounted on the printed circuit board and comprising:
- at least one processing unit (PU) disposed on a first PU die;
- at least a first graphics processing unit (GPU) disposed on a first GPU die and a second graphics processing unit (GPU) disposed on a second GPU die;
- and
- a communication network to provide communication connections between the first CPU die and the and the first GPU die;
- the processing unit to:
- receive graphics metadata from an application which is to execute on the at least one PU, wherein the graphics metadata indicates a level of graphics processing activity associated with the application;
- determine, from the graphics metadata, an amount of graphics processing resources to be dedicated to the application, the graphics processing resources comprising a plurality of graphics processing units located on the first GPU and the second GPU;
- configure one or more communication connections between the PU and the first GPU and the PU and the second GPU;
- transmit the graphics data for the application from the PU to the plurality of graphics processing units via the communication network;
- receive a completion acknowledgment from the plurality of graphics processing units;
- and
- in response to a determination that the workload is finished, to terminate one or more communication connections on the communication network.
12 more elements
A method, comprising:
- receiving graphics metadata from an application which is to execute on at least one processing unit (PU) disposed on a first die, wherein the graphics metadata indicates a level of graphics processing activity associated with the application;
- determining, from the graphics metadata, an amount of graphics processing resources to be dedicated to the application, the graphics processing resources comprising a plurality of graphics processing units located on at least a first graphics processing unit (GPU) disposed on a first GPU die and a second graphics processing unit (GPU) disposed on a second GPU die, separate from the first GPU die;
- configuring one or more communication connections in a communication network to provide communication connections between the PU and the first GPU and the first GPU and the second GPU;
- transmitting the graphics data for the application from the CPU to the plurality of graphics processing units via the communication network;
- receiving a completion acknowledgment from the plurality of graphics processing units;
- and
- in response to a determination that the workload is finished, terminating one or more communication connections on the communication network.
4 more elements
Dependencies are inferred from the claim text.