A series of changes to the GPU driver stack have been announced.
Nvidia is now a single-processor processor company, and AMD has joined the fold with its own multi-core GPU.
Intel has moved to an Intel-like approach to its hardware, with its latest Skylake-X chip and its own Hyper-Threading technology.
And AMD has moved its own high-performance computing (HPC) architecture to the next-generation Zen architecture.
Intel and AMD announced the new GPU architecture at a press conference last week.
Here are the key changes to Nvidia’s driver stack: GPU: Intel’s Graphics Core Next (GCN) processor is no longer a standalone chip.
It has been merged into the next generation Intel® Core™ processor family.
Instead, it will become a part of a unified graphics platform called Graphics CoreNext (GCNE).
NVIDIA has added a new compute API called CUDA™ to the platform, which it calls NVIDIA Compute Engine (NVENC) for short.
NVENC provides accelerated computation across GPUs and GPUs integrated CPUs.
It is built on the new Pascal GPU architecture and the GPU pipeline that was first introduced in the Kepler architecture.
NVenc can also be used to process computations on GPUs that are embedded into software and embedded in hardware.
Intel is adding new compute features to its CPU core, called the Hyper-threading Technology.
These include a new Hyper-V hypervisor for virtualization, new virtualization capabilities for virtual machines, and new networking capabilities that are intended to support more virtualization workloads.
NVEnc enables virtualization to take advantage of the compute capabilities of the new generation of processors, which are based on the GPU core and the Hyperthreading technology, and it provides a new way to run multiple virtualization applications on the same CPU core.
AMD is adding a new HPC compute engine called AMD Radeon™ Open Compute that it calls Radeon™ Advanced Compute.
AMD’s compute engine will use the same hardware and architecture as the existing AMD Radeon compute architecture, and will be designed to be parallel, low-power and high-efficiency, and is expected to be available in the second half of 2019.
AMD and Nvidia are working on new graphics processing architectures.
AMD, along with Nvidia, is working on two new graphics architectures: GPU-accelerated GPUs, or GPUs that have multiple processing cores, and GPUs with compute units, or GPU clusters.
The first is called AMDGPU, which has been in development for some time and is designed for high-end graphics workloads where there are multiple workloads to process and where GPUs can have a large number of compute units.
AMDGPU is intended for a variety of workloads, and Nvidia has announced that it will release a high-powered GPU called Radeon RX Vega 56 in the first quarter of 2019, which will compete directly with AMDGPU.
NVIDIA has released the second architecture in its own pipeline called AMD GPU Boost, which is designed to support high-throughput, multi-threaded workloads in the graphics pipeline.
It will also work with other GPUs that share its compute and graphics pipelines, such as Intel’s Xeon Phi and AMD’s Vega.
AMD has also released Radeon RX Fury, which AMD calls the Vega Frontier Edition, a graphics card with more compute power and performance than the Vega 64.
It supports multiple GPUs, and there is a “full suite” of high-speed compute features for developers and enthusiasts.
NVIDIA is also making improvements to its PowerPlay, or Power Management, features.
PowerPlay is a system that enables applications to communicate with each other and with their graphics hardware, such that graphics applications can manage GPU-intensive tasks more efficiently and safely.
AMD will also release its next-gen graphics products, Radeon RX 580 and Radeon RX 570, which can support more than two GPUs simultaneously.
AMD released the Radeon RX 480 in November, which launched a number of new features.
These included the Radeon™ Vega 64, which was based on a new Graphics Core Processor family.
AMD also released the AMD Radeon Pro Vega 64 and Radeon Pro Pro Vega 56 GPUs.
The Radeon RX 500 series, which includes the Radeon Pro 500 and Radeon Vega 500, is aimed at enthusiasts and professional users.
NVIDIA’s first-generation GPU architecture, Maxwell, was designed to run next-to-no-cost compute workloads on consumer graphics cards.
Maxwell also provided more memory bandwidth for applications.
The GPU stack on the Radeon Vega 64 features two new compute cores, the “Vega 64” core and “Vesa 64” cores.
The Vega 64 core is a dual-core part that can run parallel compute workload and will support multiple compute units at the same time.
The “Vesa 64” is a two-core CPU core that can support a wider variety of compute workload on the Vega GPU.
AMD says the Vega 56 will be capable of up to 10 compute cores per GPU.
NVIDIA says the first Radeon Vega GPUs will support up to 64 compute