Скачать или смотреть 1.13 CPU-GPU Memory access (IBDP)

Memory access in the context of computing hardware
such as CPUs and GPUs
refers to how these processors
retrieve and
manipulate data stored in computer memory.
Each type of processor
handles memory access in its own way
based on its architectural design,
which impacts how well it performs.
CPUs utilize a memory hierarchy to manage data access in an efficient way.
This hierarchy typically includes several levels of caches (L1, L2, and sometimes L3).
This hierarchy is optimized to minimize memory latency(—the delay)
from issuing a memory request to receiving the data.
CPUs typically run in settings with several cores,
so they rely on cache coherence protocols.
These protocols
keep each core’s view of memory aligned,
which helps avoid data conflicts
and preserves the integrity of shared information.
Modern GPUs often use a unified memory architecture,
which allows them to access a large,
shared pool of memory which both the GPU and CPU can address.
GPUs are designed with high bandwidth memory.
These memory types are optimized
for the high-throughput requirements of GPU tasks,
enabling fast data transfer rates
that support the processing capabilities
of hundreds to thousands of parallel cores.
Unlike CPUs, which are optimized for low-latency access,
GPUs prioritize memory throughput.
To summarize memory access,
CPUs utilize low-latency memory
because they need to rapidly switch between tasks,
retrieve data from memory,
and execute operations
based on that data with minimal delay.
GPUs utilize high memory throughput
because they handle large volumes of data
and need to feed hundreds to thousands of parallel cores simultaneously.
IB Diploma Programme: Computer Science
The International Baccalaureate® (IB) Diploma Programme (DP)
Computer science.
#IBDP #InternationalBaccalaureate #ComputerScience #IBEducation #IBTeacher #TeachingAndLearning #EdTe
#IBDP #InternationalBaccalaureate #ComputerScience #IBEducation #IBTeacher #TeachingAndLearning #EdTech #LifelongLearning #IBContinuum #DigitalLearning